diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
99 files changed, 16208 insertions, 6161 deletions
diff --git a/servers/rendering/renderer_rd/SCsub b/servers/rendering/renderer_rd/SCsub index 9c95f538ac..64e613ab91 100644 --- a/servers/rendering/renderer_rd/SCsub +++ b/servers/rendering/renderer_rd/SCsub @@ -5,4 +5,5 @@ Import("env") env.add_source_files(env.servers_sources, "*.cpp") SConscript("forward_clustered/SCsub") +SConscript("forward_mobile/SCsub") SConscript("shaders/SCsub") diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp index 0fdd864d47..b952ecbff0 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.cpp +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -86,13 +86,13 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { Vector<uint8_t> vertex_data; vertex_data.resize(sizeof(float) * icosphere_vertex_count * 3); - copymem(vertex_data.ptrw(), icosphere_vertices, vertex_data.size()); + memcpy(vertex_data.ptrw(), icosphere_vertices, vertex_data.size()); sphere_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); Vector<uint8_t> index_data; index_data.resize(sizeof(uint32_t) * icosphere_triangle_count * 3); - copymem(index_data.ptrw(), icosphere_triangle_indices, index_data.size()); + memcpy(index_data.ptrw(), icosphere_triangle_indices, index_data.size()); sphere_index_buffer = RD::get_singleton()->index_buffer_create(icosphere_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); @@ -130,13 +130,13 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { Vector<uint8_t> vertex_data; vertex_data.resize(sizeof(float) * cone_vertex_count * 3); - copymem(vertex_data.ptrw(), cone_vertices, vertex_data.size()); + memcpy(vertex_data.ptrw(), cone_vertices, vertex_data.size()); cone_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); Vector<uint8_t> index_data; index_data.resize(sizeof(uint32_t) * cone_triangle_count * 3); - copymem(index_data.ptrw(), cone_triangle_indices, index_data.size()); + memcpy(index_data.ptrw(), cone_triangle_indices, index_data.size()); cone_index_buffer = RD::get_singleton()->index_buffer_create(cone_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); @@ -184,13 +184,13 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { Vector<uint8_t> vertex_data; vertex_data.resize(sizeof(float) * box_vertex_count * 3); - copymem(vertex_data.ptrw(), box_vertices, vertex_data.size()); + memcpy(vertex_data.ptrw(), box_vertices, vertex_data.size()); box_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); Vector<uint8_t> index_data; index_data.resize(sizeof(uint32_t) * box_triangle_count * 3); - copymem(index_data.ptrw(), box_triangle_indices, index_data.size()); + memcpy(index_data.ptrw(), box_triangle_indices, index_data.size()); box_index_buffer = RD::get_singleton()->index_buffer_create(box_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); @@ -374,7 +374,7 @@ void ClusterBuilderRD::setup(Size2i p_screen_size, uint32_t p_max_elements, RID } } -void ClusterBuilderRD::begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y) { +void ClusterBuilderRD::begin(const Transform3D &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y) { view_xform = p_view_transform.affine_inverse(); projection = p_cam_projection; z_near = projection.get_z_near(); diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h index dc1707b534..c0c03eb26a 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.h +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -167,7 +167,7 @@ private: uint32_t render_element_count = 0; uint32_t render_element_max = 0; - Transform view_xform; + Transform3D view_xform; CameraMatrix adjusted_projection; CameraMatrix projection; float z_far = 0; @@ -220,9 +220,9 @@ private: public: void setup(Size2i p_screen_size, uint32_t p_max_elements, RID p_depth_buffer, RID p_depth_buffer_sampler, RID p_color_buffer); - void begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y); + void begin(const Transform3D &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y); - _FORCE_INLINE_ void add_light(LightType p_type, const Transform &p_transform, float p_radius, float p_spot_aperture) { + _FORCE_INLINE_ void add_light(LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) { if (p_type == LIGHT_TYPE_OMNI && cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT] == max_elements_by_type) { return; //max number elements reached } @@ -232,10 +232,10 @@ public: RenderElementData &e = render_elements[render_element_count]; - Transform xform = view_xform * p_transform; + Transform3D xform = view_xform * p_transform; float radius = xform.basis.get_uniform_scale(); - if (radius > 0.98 || radius < 1.02) { + if (radius < 0.98 || radius > 1.02) { xform.basis.orthonormalize(); } @@ -317,7 +317,7 @@ public: render_element_count++; } - _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform &p_transform, const Vector3 &p_half_extents) { + _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform3D &p_transform, const Vector3 &p_half_extents) { if (p_box_type == BOX_TYPE_DECAL && cluster_count_by_type[ELEMENT_TYPE_DECAL] == max_elements_by_type) { return; //max number elements reached } @@ -326,7 +326,7 @@ public: } RenderElementData &e = render_elements[render_element_count]; - Transform xform = view_xform * p_transform; + Transform3D xform = view_xform * p_transform; //extract scale and scale the matrix by it, makes things simpler Vector3 scale = p_half_extents; diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index bc304aedd8..3683622d3e 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -34,21 +34,11 @@ #include "core/math/math_defs.h" #include "core/os/os.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" #include "thirdparty/misc/cubemap_coeffs.h" -static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_basis, float *p_array) { - p_array[0] = p_basis.elements[0][0]; - p_array[1] = p_basis.elements[1][0]; - p_array[2] = p_basis.elements[2][0]; - p_array[3] = 0; - p_array[4] = p_basis.elements[0][1]; - p_array[5] = p_basis.elements[1][1]; - p_array[6] = p_basis.elements[2][1]; - p_array[7] = 0; - p_array[8] = p_basis.elements[0][2]; - p_array[9] = p_basis.elements[1][2]; - p_array[10] = p_basis.elements[2][2]; - p_array[11] = 0; +bool EffectsRD::get_prefer_raster_effects() { + return prefer_raster_effects; } static _FORCE_INLINE_ void store_camera(const CameraMatrix &p_mtx, float *p_array) { @@ -80,6 +70,28 @@ RID EffectsRD::_get_uniform_set_from_image(RID p_image) { return uniform_set; } +RID EffectsRD::_get_uniform_set_for_input(RID p_texture) { + if (input_to_uniform_set_cache.has(p_texture)) { + RID uniform_set = input_to_uniform_set_cache[p_texture]; + if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + return uniform_set; + } + } + + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_INPUT_ATTACHMENT; + u.binding = 0; + u.ids.push_back(p_texture); + uniforms.push_back(u); + // This is specific to our subpass shader + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, tonemap.shader.version_get_shader(tonemap.shader_version, TONEMAP_MODE_SUBPASS), 0); + + input_to_uniform_set_cache[p_texture] = uniform_set; + + return uniform_set; +} + RID EffectsRD::_get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) { if (texture_to_uniform_set_cache.has(p_texture)) { RID uniform_set = texture_to_uniform_set_cache[p_texture]; @@ -95,7 +107,7 @@ RID EffectsRD::_get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) u.ids.push_back(p_use_mipmaps ? default_mipmap_sampler : default_sampler); u.ids.push_back(p_texture); uniforms.push_back(u); - //anything with the same configuration (one texture in binding 0 for set 0), is good + // anything with the same configuration (one texture in binding 0 for set 0), is good RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, tonemap.shader.version_get_shader(tonemap.shader_version, 0), 0); texture_to_uniform_set_cache[p_texture] = uniform_set; @@ -226,7 +238,7 @@ RID EffectsRD::_get_compute_uniform_set_from_image_pair(RID p_texture1, RID p_te } void EffectsRD::copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_uv_rect, RD::DrawListID p_draw_list, bool p_flip_y, bool p_panorama) { - zeromem(©_to_fb.push_constant, sizeof(CopyToFbPushConstant)); + memset(©_to_fb.push_constant, 0, sizeof(CopyToFbPushConstant)); copy_to_fb.push_constant.use_section = true; copy_to_fb.push_constant.section[0] = p_uv_rect.position.x; @@ -247,7 +259,7 @@ void EffectsRD::copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_framebuffer } void EffectsRD::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_alpha_to_zero, bool p_srgb, RID p_secondary) { - zeromem(©_to_fb.push_constant, sizeof(CopyToFbPushConstant)); + memset(©_to_fb.push_constant, 0, sizeof(CopyToFbPushConstant)); if (p_flip_y) { copy_to_fb.push_constant.flip_y = true; @@ -275,7 +287,7 @@ void EffectsRD::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, } void EffectsRD::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_all_source, bool p_8_bit_dst, bool p_alpha_to_one) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); if (p_flip_y) { copy.push_constant.flags |= COPY_FLAG_FLIP_Y; } @@ -309,7 +321,7 @@ void EffectsRD::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const } void EffectsRD::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); copy.push_constant.section[0] = 0; copy.push_constant.section[1] = 0; @@ -329,7 +341,7 @@ void EffectsRD::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, } void EffectsRD::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y, float p_z_near, float p_z_far) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); if (p_flip_y) { copy.push_constant.flags |= COPY_FLAG_FLIP_Y; } @@ -353,7 +365,7 @@ void EffectsRD::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_ } void EffectsRD::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); if (p_flip_y) { copy.push_constant.flags |= COPY_FLAG_FLIP_Y; } @@ -375,7 +387,7 @@ void EffectsRD::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, } void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); copy.push_constant.section[0] = 0; copy.push_constant.section[1] = 0; @@ -397,7 +409,9 @@ void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i } void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back_texture, const Rect2i &p_region, bool p_8bit_dst) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the compute version of the gaussian blur with the mobile renderer."); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); uint32_t base_flags = 0; copy.push_constant.section[0] = p_region.position.x; @@ -430,7 +444,9 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back } void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_treshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the gaussian glow with the mobile renderer."); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); CopyMode copy_mode = p_first_pass && p_auto_exposure.is_valid() ? COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : COPY_MODE_GAUSSIAN_GLOW; uint32_t base_flags = 0; @@ -463,6 +479,57 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const RD::get_singleton()->compute_list_end(); } +void EffectsRD::gaussian_glow_raster(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_treshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of the gaussian glow with the clustered renderer."); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode blur_mode = p_first_pass && p_auto_exposure.is_valid() ? BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : BLUR_MODE_GAUSSIAN_GLOW; + uint32_t base_flags = 0; + + blur_raster.push_constant.pixel_size[0] = p_pixel_size.x; + blur_raster.push_constant.pixel_size[1] = p_pixel_size.y; + + blur_raster.push_constant.glow_strength = p_strength; + blur_raster.push_constant.glow_bloom = p_bloom; + blur_raster.push_constant.glow_hdr_threshold = p_hdr_bleed_treshold; + blur_raster.push_constant.glow_hdr_scale = p_hdr_bleed_scale; + blur_raster.push_constant.glow_exposure = p_exposure; + blur_raster.push_constant.glow_white = 0; //actually unused + blur_raster.push_constant.glow_luminance_cap = p_luminance_cap; + + blur_raster.push_constant.glow_auto_exposure_grey = p_auto_exposure_grey; //unused also + + //HORIZONTAL + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer_half, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer_half))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + if (p_auto_exposure.is_valid() && p_first_pass) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_auto_exposure), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags | BLUR_FLAG_HORIZONTAL | (p_first_pass ? BLUR_FLAG_GLOW_FIRST_PASS : 0); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + blur_mode = BLUR_MODE_GAUSSIAN_GLOW; + + //VERTICAL + draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_rd_texture_half), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags; + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); @@ -517,11 +584,10 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_output, p_blur_radius), 1); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_metallic, p_normal_roughness), 3); } else { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_output), 1); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_metallic), 3); } + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_metallic), 3); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_scale_normal), 2); RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); @@ -657,7 +723,7 @@ void EffectsRD::merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_bas } void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size) { - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); copy.push_constant.section[0] = 0; copy.push_constant.section[1] = 0; @@ -673,7 +739,27 @@ void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const S RD::get_singleton()->compute_list_end(); } -void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip) { +void EffectsRD::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of mipmap with the clustered renderer."); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode mode = BLUR_MIPMAP; + + blur_raster.push_constant.pixel_size[0] = 1.0 / float(p_size.x); + blur_raster.push_constant.pixel_size[1] = 1.0 / float(p_size.y); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip) { CopyToDPPushConstant push_constant; push_constant.screen_rect[0] = p_rect.position.x; push_constant.screen_rect[1] = p_rect.position.y; @@ -681,7 +767,9 @@ void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffe push_constant.screen_rect[3] = p_rect.size.height; push_constant.z_far = p_z_far; push_constant.z_near = p_z_near; - push_constant.z_flip = p_dp_flip; + push_constant.texel_size[0] = 1.0f / p_dst_size.x; + push_constant.texel_size[1] = 1.0f / p_dst_size.y; + push_constant.texel_size[0] *= p_dp_flip ? -1.0f : 1.0f; // Encode dp flip as x size sign RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cube_to_dp.pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); @@ -694,7 +782,7 @@ void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffe } void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings) { - zeromem(&tonemap.push_constant, sizeof(TonemapPushConstant)); + memset(&tonemap.push_constant, 0, sizeof(TonemapPushConstant)); tonemap.push_constant.use_bcs = p_settings.use_bcs; tonemap.push_constant.bcs[0] = p_settings.brightness; @@ -732,8 +820,13 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone tonemap.push_constant.pixel_size[0] = 1.0 / p_settings.texture_size.x; tonemap.push_constant.pixel_size[1] = 1.0 / p_settings.texture_size.y; + if (p_settings.view_count > 1) { + // Use MULTIVIEW versions + mode += 6; + } + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer), false, RD::get_singleton()->draw_list_get_current_pass())); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_color), 0); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_settings.exposure_texture), 1); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_settings.glow_texture, true), 2); @@ -745,7 +838,48 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone RD::get_singleton()->draw_list_end(); } +void EffectsRD::tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings) { + memset(&tonemap.push_constant, 0, sizeof(TonemapPushConstant)); + + tonemap.push_constant.use_bcs = p_settings.use_bcs; + tonemap.push_constant.bcs[0] = p_settings.brightness; + tonemap.push_constant.bcs[1] = p_settings.contrast; + tonemap.push_constant.bcs[2] = p_settings.saturation; + + ERR_FAIL_COND_MSG(p_settings.use_glow, "Glow is not supported when using subpasses."); + tonemap.push_constant.use_glow = p_settings.use_glow; + + int mode = p_settings.use_1d_color_correction ? TONEMAP_MODE_SUBPASS_1D_LUT : TONEMAP_MODE_SUBPASS; + if (p_settings.view_count > 1) { + // Use MULTIVIEW versions + mode += 6; + } + + tonemap.push_constant.tonemapper = p_settings.tonemap_mode; + tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; + tonemap.push_constant.exposure = p_settings.exposure; + tonemap.push_constant.white = p_settings.white; + tonemap.push_constant.auto_exposure_grey = p_settings.auto_exposure_grey; + + tonemap.push_constant.use_color_correction = p_settings.use_color_correction; + + tonemap.push_constant.use_debanding = p_settings.use_debanding; + + RD::get_singleton()->draw_list_bind_render_pipeline(p_subpass_draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, p_dst_format_id, false, RD::get_singleton()->draw_list_get_current_pass())); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_for_input(p_source_color), 0); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.exposure_texture), 1); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.glow_texture, true), 2); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.color_correction_texture), 3); + + RD::get_singleton()->draw_list_bind_index_array(p_subpass_draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(p_subpass_draw_list, &tonemap.push_constant, sizeof(TonemapPushConstant)); + RD::get_singleton()->draw_list_draw(p_subpass_draw_list, true); +} + void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of luminance reduction with the mobile renderer."); + luminance_reduce.push_constant.source_size[0] = p_source_size.x; luminance_reduce.push_constant.source_size[1] = p_source_size.y; luminance_reduce.push_constant.max_luminance = p_max_luminance; @@ -784,7 +918,41 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_end(); } -void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_halfsize_texture1, RID p_halfsize_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { +void EffectsRD::luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster version of luminance reduction with the clustered renderer."); + ERR_FAIL_COND_MSG(p_reduce.size() != p_fb.size(), "Incorrect frame buffer account for luminance reduction."); + + luminance_reduce_raster.push_constant.max_luminance = p_max_luminance; + luminance_reduce_raster.push_constant.min_luminance = p_min_luminance; + luminance_reduce_raster.push_constant.exposure_adjust = p_adjust; + + for (int i = 0; i < p_reduce.size(); i++) { + luminance_reduce_raster.push_constant.source_size[0] = i == 0 ? p_source_size.x : luminance_reduce_raster.push_constant.dest_size[0]; + luminance_reduce_raster.push_constant.source_size[1] = i == 0 ? p_source_size.y : luminance_reduce_raster.push_constant.dest_size[1]; + luminance_reduce_raster.push_constant.dest_size[0] = MAX(luminance_reduce_raster.push_constant.source_size[0] / 8, 1); + luminance_reduce_raster.push_constant.dest_size[1] = MAX(luminance_reduce_raster.push_constant.source_size[1] / 8, 1); + + bool final = !p_set && (luminance_reduce_raster.push_constant.dest_size[0] == 1) && (luminance_reduce_raster.push_constant.dest_size[1] == 1); + LuminanceReduceRasterMode mode = final ? LUMINANCE_REDUCE_FRAGMENT_FINAL : (i == 0 ? LUMINANCE_REDUCE_FRAGMENT_FIRST : LUMINANCE_REDUCE_FRAGMENT); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, luminance_reduce_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_fb[i]))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(i == 0 ? p_source_texture : p_reduce[i - 1]), 0); + if (final) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_prev_luminance), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &luminance_reduce_raster.push_constant, sizeof(LuminanceReduceRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } +} + +void EffectsRD::bokeh_dof(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of BOKEH DOF with the mobile renderer."); + bokeh.push_constant.blur_far_active = p_dof_far; bokeh.push_constant.blur_far_begin = p_dof_far_begin; bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; @@ -811,22 +979,22 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i // The alpha channel of the source color texture is filled with the expected circle size // If used for DOF far, the size is positive, if used for near, its negative. - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_GEN_BLUR_SIZE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BLUR_SIZE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.depth_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { //second pass - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL]); static const int quality_samples[4] = { 6, 12, 12, 24 }; @@ -835,18 +1003,18 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[0]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; - bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; bokeh.push_constant.half_size = true; bokeh.push_constant.blur_size *= 0.5; } else { //medium and high quality use full size - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_secondary_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.secondary_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -858,11 +1026,11 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i bokeh.push_constant.second_pass = true; if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture2), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[1]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[0]), 1); } else { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_secondary_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.secondary_texture), 1); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -873,25 +1041,25 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { //forth pass, upscale for low quality - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_COMPOSITE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture2), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[1]), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; bokeh.push_constant.half_size = false; bokeh.push_constant.second_pass = false; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); } } else { //circle //second pass - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_GEN_BOKEH_CIRCULAR]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BOKEH_CIRCULAR]); static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; @@ -900,11 +1068,11 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i //circle always runs in half size, otherwise too expensive - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[0]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; - bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; bokeh.push_constant.half_size = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -916,28 +1084,202 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i // upscale - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_COMPOSITE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[0]), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; bokeh.push_constant.half_size = false; bokeh.push_constant.second_pass = false; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); } RD::get_singleton()->compute_list_end(); } -void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass) { - RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.gather_uniform_set, 0); +void EffectsRD::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use blur DOF with the clustered renderer."); + + memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); + + bokeh.push_constant.orthogonal = p_cam_orthogonal; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.width; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.height; + bokeh.push_constant.z_far = p_cam_zfar; + bokeh.push_constant.z_near = p_cam_znear; + + bokeh.push_constant.second_pass = false; + bokeh.push_constant.half_size = false; + bokeh.push_constant.blur_size = p_dof_blur_amount; + + if (p_dof_far || p_dof_near) { + if (p_dof_far) { + bokeh.push_constant.blur_far_active = true; + bokeh.push_constant.blur_far_begin = p_dof_far_begin; + bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; + } + + if (p_dof_near) { + bokeh.push_constant.blur_near_active = true; + bokeh.push_constant.blur_near_begin = p_dof_near_begin; + bokeh.push_constant.blur_near_end = p_dof_near_begin - p_dof_near_size; + } + + { + // generate our depth data + RID framebuffer = p_buffers.base_weight_fb; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[BOKEH_GEN_BLUR_SIZE].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.depth_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + // double pass approach + BokehMode mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + + if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { + //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + bokeh.push_constant.blur_size *= 0.5; + } + + static const int quality_samples[4] = { 6, 12, 12, 24 }; + bokeh.push_constant.blur_scale = 0.5; + bokeh.push_constant.steps = quality_samples[p_quality]; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + // Pass 1 + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + // Pass 2 + if (!bokeh.push_constant.half_size) { + // do not output weight, we're writing back into our base buffer + mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX_NOWEIGHT : BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT; + } + bokeh.push_constant.second_pass = true; + + framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[1] : p_buffers.base_fb; + RID texture = bokeh.push_constant.half_size ? p_buffers.half_texture[0] : p_buffers.secondary_texture; + RID weight = bokeh.push_constant.half_size ? p_buffers.weight_texture[2] : p_buffers.weight_texture[1]; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(weight), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose pass + mode = BOKEH_COMPOSITE; + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.half_texture[1]), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[3]), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + } else { + // circular is a single pass approach + BokehMode mode = BOKEH_GEN_BOKEH_CIRCULAR; + + { + // circle always runs in half size, otherwise too expensive (though the code below does support making this optional) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + // bokeh.push_constant.blur_size *= 0.5; + } + + static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; + bokeh.push_constant.blur_scale = quality_scale[p_quality]; + bokeh.push_constant.steps = 0.0; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose + mode = BOKEH_COMPOSITE; + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.half_texture[0]), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[2]), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } else { + // Just copy it back (we use our blur raster shader here).. + draw_list = RD::get_singleton()->draw_list_begin(p_buffers.base_fb, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[BLUR_MODE_COPY].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_buffers.base_fb))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.secondary_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + } + } +} + +void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); if ((p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) && !p_adaptive_base_pass) { - RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.importance_map_uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 1); } for (int i = 0; i < 4; i++) { @@ -960,7 +1302,7 @@ void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> RD::get_singleton()->compute_list_add_barrier(p_compute_list); } -void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { +void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets, RID &r_downsample_uniform_set, RID &r_gather_uniform_set, RID &r_importance_map_uniform_set) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->draw_command_begin_label("SSAO"); /* FIRST PASS */ @@ -990,7 +1332,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(p_depth_mipmaps[3]); uniforms.push_back(u); } - ssao.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2); + r_downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2); } float depth_linearize_mul = -p_projection.matrix[3][2]; @@ -1025,7 +1367,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_depth_mipmaps[0]), 1); if (p_settings.quality > RS::ENV_SSAO_QUALITY_MEDIUM) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.downsample_uniform_set, 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, r_downsample_uniform_set, 2); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant)); @@ -1109,7 +1451,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(ssao.gather_constants_buffer); uniforms.push_back(u); } - ssao.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0), 0); + r_gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0), 0); } if (p_invalidate_uniform_sets) { @@ -1136,7 +1478,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(ssao.importance_map_load_counter); uniforms.push_back(u); } - ssao.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); + r_importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); } if (p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) { @@ -1147,7 +1489,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep ssao.importance_map_push_constant.power = p_settings.power; //base pass RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); - gather_ssao(compute_list, p_ao_pong_slices, p_settings, true); + gather_ssao(compute_list, p_ao_pong_slices, p_settings, true, r_gather_uniform_set, RID()); //generate importance map RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); @@ -1178,7 +1520,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); } - gather_ssao(compute_list, p_ao_slices, p_settings, false); + gather_ssao(compute_list, p_ao_slices, p_settings, false, r_gather_uniform_set, r_importance_map_uniform_set); RD::get_singleton()->draw_command_end_label(); // Gather SSAO } @@ -1198,8 +1540,9 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { if (pass < blur_passes - 2) { blur_pipeline = SSAO_BLUR_PASS_WIDE; + } else { + blur_pipeline = SSAO_BLUR_PASS_SMART; } - blur_pipeline = SSAO_BLUR_PASS_SMART; } for (int i = 0; i < 4; i++) { @@ -1293,8 +1636,10 @@ void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size RD::get_singleton()->compute_list_end(); } -void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { - zeromem(&roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); +void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap roughness with the mobile renderer."); + + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); roughness.push_constant.face_id = p_face_id > 9 ? 0 : p_face_id; roughness.push_constant.roughness = p_roughness; @@ -1303,10 +1648,10 @@ void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffe roughness.push_constant.face_size = p_size; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness.pipeline); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness.compute_pipeline); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_framebuffer), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); @@ -1318,11 +1663,37 @@ void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffe RD::get_singleton()->compute_list_end(); } +void EffectsRD::cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap roughness with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap roughness must process one side at a time."); + + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); + + roughness.push_constant.face_id = p_face_id; + roughness.push_constant.roughness = p_roughness; + roughness.push_constant.sample_count = p_sample_count; + roughness.push_constant.use_direct_write = p_roughness == 0.0; + roughness.push_constant.face_size = p_size; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, roughness.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap downsample with the mobile renderer."); + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = 0; // we render all 6 sides to each layer in one call RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cubemap_downsampler.pipeline); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cubemap_downsampler.compute_pipeline); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cubemap), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_cubemap), 1); @@ -1336,7 +1707,27 @@ void EffectsRD::cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, con RD::get_singleton()->compute_list_end(); } +void EffectsRD::cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap downsample with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap downsample must process one side at a time."); + + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = p_face_id; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cubemap_downsampler.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &cubemap_downsampler.push_constant, sizeof(CubemapDownsamplerPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap filter with the mobile renderer."); + Vector<RD::Uniform> uniforms; for (int i = 0; i < p_dest_cubemap.size(); i++) { RD::Uniform u; @@ -1348,12 +1739,12 @@ void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, if (RD::get_singleton()->uniform_set_is_valid(filter.image_uniform_set)) { RD::get_singleton()->free(filter.image_uniform_set); } - filter.image_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.shader.version_get_shader(filter.shader_version, 0), 2); + filter.image_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, 0), 2); int pipeline = p_use_array ? FILTER_MODE_HIGH_QUALITY_ARRAY : FILTER_MODE_HIGH_QUALITY; pipeline = filter.use_high_quality ? pipeline : pipeline + 1; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, filter.pipelines[pipeline]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, filter.compute_pipelines[pipeline]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cubemap, true), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.uniform_set, 1); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.image_uniform_set, 2); @@ -1365,55 +1756,42 @@ void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, RD::get_singleton()->compute_list_end(); } -void EffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_fog, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, const CameraMatrix &p_camera, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position) { - SkyPushConstant sky_push_constant; - - zeromem(&sky_push_constant, sizeof(SkyPushConstant)); - - sky_push_constant.proj[0] = p_camera.matrix[2][0]; - sky_push_constant.proj[1] = p_camera.matrix[0][0]; - sky_push_constant.proj[2] = p_camera.matrix[2][1]; - sky_push_constant.proj[3] = p_camera.matrix[1][1]; - sky_push_constant.position[0] = p_position.x; - sky_push_constant.position[1] = p_position.y; - sky_push_constant.position[2] = p_position.z; - sky_push_constant.multiplier = p_multiplier; - sky_push_constant.time = p_time; - store_transform_3x3(p_orientation, sky_push_constant.orientation); - - RenderingDevice::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_fb); +void EffectsRD::cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap filter with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap filter must process one side at a time."); - RD::DrawListID draw_list = p_list; + // TODO implement! + CubemapFilterRasterPushConstant push_constant; + push_constant.mip_level = p_mip_level; + push_constant.face_id = p_face_id; - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format)); - - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_samplers, 0); - if (p_uniform_set.is_valid()) { //material may not have uniform set - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_uniform_set, 1); - } - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_texture_set, 2); - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_fog, 3); + CubemapFilterMode mode = filter.use_high_quality ? FILTER_MODE_HIGH_QUALITY : FILTER_MODE_LOW_QUALITY; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, filter.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, filter.uniform_set, 1); RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); - RD::get_singleton()->draw_list_set_push_constant(draw_list, &sky_push_constant, sizeof(SkyPushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CubemapFilterRasterPushConstant)); RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); } -void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { +void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { ResolvePushConstant push_constant; push_constant.screen_size[0] = p_screen_size.x; push_constant.screen_size[1] = p_screen_size.y; push_constant.samples = p_samples; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[p_source_giprobe.is_valid() ? RESOLVE_MODE_GI_GIPROBE : RESOLVE_MODE_GI]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[p_source_voxel_gi.is_valid() ? RESOLVE_MODE_GI_VOXEL_GI : RESOLVE_MODE_GI]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_source_depth, p_source_normal_roughness), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_dest_depth, p_dest_normal_roughness), 1); - if (p_source_giprobe.is_valid()) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_giprobe), 2); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_giprobe), 3); + if (p_source_voxel_gi.is_valid()) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_voxel_gi), 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_voxel_gi), 3); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); @@ -1423,6 +1801,24 @@ void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RI RD::get_singleton()->compute_list_end(p_barrier); } +void EffectsRD::resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { + ResolvePushConstant push_constant; + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.samples = p_samples; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[RESOLVE_MODE_DEPTH]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_depth), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_depth), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(p_barrier); +} + void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { Sort::PushConstant push_constant; push_constant.total_elements = p_size; @@ -1492,8 +1888,35 @@ void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { RD::get_singleton()->compute_list_end(); } -EffectsRD::EffectsRD() { - { // Initialize copy +EffectsRD::EffectsRD(bool p_prefer_raster_effects) { + prefer_raster_effects = p_prefer_raster_effects; + + if (prefer_raster_effects) { + // init blur shader (on compute use copy shader) + + Vector<String> blur_modes; + blur_modes.push_back("\n#define MODE_MIPMAP\n"); // BLUR_MIPMAP + blur_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); // BLUR_MODE_GAUSSIAN_BLUR + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n"); // BLUR_MODE_GAUSSIAN_GLOW + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n#define GLOW_USE_AUTO_EXPOSURE\n"); // BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE + blur_modes.push_back("\n#define MODE_COPY\n"); // BLUR_MODE_COPY + + blur_raster.shader.initialize(blur_modes); + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + blur_raster.shader_version = blur_raster.shader.version_create(); + + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].setup(blur_raster.shader.version_get_shader(blur_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + + } else { + // not used in clustered + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].clear(); + } + } + + if (!prefer_raster_effects) { // Initialize copy Vector<String> copy_modes; copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n#define DST_IMAGE_8BIT\n"); @@ -1510,11 +1933,22 @@ EffectsRD::EffectsRD() { copy_modes.push_back("\n#define MODE_CUBEMAP_ARRAY_TO_PANORAMA\n"); copy.shader.initialize(copy_modes); - zeromem(©.push_constant, sizeof(CopyPushConstant)); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + if (prefer_raster_effects) { + // disable shaders we can't use + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_COPY, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_COPY_8BIT, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_GLOW, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, false); + } + copy.shader_version = copy.shader.version_create(); for (int i = 0; i < COPY_MODE_MAX; i++) { - copy.pipelines[i] = RD::get_singleton()->compute_pipeline_create(copy.shader.version_get_shader(copy.shader_version, i)); + if (copy.shader.is_variant_enabled(i)) { + copy.pipelines[i] = RD::get_singleton()->compute_pipeline_create(copy.shader.version_get_shader(copy.shader_version, i)); + } } } { @@ -1538,11 +1972,22 @@ EffectsRD::EffectsRD() { // Initialize roughness Vector<String> cubemap_roughness_modes; cubemap_roughness_modes.push_back(""); - roughness.shader.initialize(cubemap_roughness_modes); - roughness.shader_version = roughness.shader.version_create(); + if (prefer_raster_effects) { + roughness.raster_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.raster_shader.version_create(); + + roughness.raster_pipeline.setup(roughness.raster_shader.version_get_shader(roughness.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); - roughness.pipeline = RD::get_singleton()->compute_pipeline_create(roughness.shader.version_get_shader(roughness.shader_version, 0)); + } else { + roughness.compute_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.compute_shader.version_create(); + + roughness.compute_pipeline = RD::get_singleton()->compute_pipeline_create(roughness.compute_shader.version_get_shader(roughness.shader_version, 0)); + roughness.raster_pipeline.clear(); + } } { @@ -1552,17 +1997,53 @@ EffectsRD::EffectsRD() { tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n"); tonemap_modes.push_back("\n#define USE_1D_LUT\n"); tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define SUBPASS\n#define USE_1D_LUT\n"); + + // multiview versions of our shaders + tonemap_modes.push_back("\n#define MULTIVIEW\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n#define USE_1D_LUT\n"); tonemap.shader.initialize(tonemap_modes); + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + tonemap.shader.set_variant_enabled(TONEMAP_MODE_NORMAL_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_1D_LUT_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, false); + } + tonemap.shader_version = tonemap.shader.version_create(); for (int i = 0; i < TONEMAP_MODE_MAX; i++) { - tonemap.pipelines[i].setup(tonemap.shader.version_get_shader(tonemap.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + if (tonemap.shader.is_variant_enabled(i)) { + tonemap.pipelines[i].setup(tonemap.shader.version_get_shader(tonemap.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + tonemap.pipelines[i].clear(); + } } } - { + if (prefer_raster_effects) { + Vector<String> luminance_reduce_modes; + luminance_reduce_modes.push_back("\n#define FIRST_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FIRST + luminance_reduce_modes.push_back("\n"); // LUMINANCE_REDUCE_FRAGMENT + luminance_reduce_modes.push_back("\n#define FINAL_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FINAL + + luminance_reduce_raster.shader.initialize(luminance_reduce_modes); + memset(&luminance_reduce_raster.push_constant, 0, sizeof(LuminanceReduceRasterPushConstant)); + luminance_reduce_raster.shader_version = luminance_reduce_raster.shader.version_create(); + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].setup(luminance_reduce_raster.shader.version_get_shader(luminance_reduce_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + } else { // Initialize luminance_reduce Vector<String> luminance_reduce_modes; luminance_reduce_modes.push_back("\n#define READ_TEXTURE\n"); @@ -1576,6 +2057,10 @@ EffectsRD::EffectsRD() { for (int i = 0; i < LUMINANCE_REDUCE_MAX; i++) { luminance_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(luminance_reduce.shader.version_get_shader(luminance_reduce.shader_version, i)); } + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].clear(); + } } { @@ -1594,25 +2079,43 @@ EffectsRD::EffectsRD() { cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); } - { - // Initialize bokeh - Vector<String> bokeh_modes; - bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_CIRCULAR\n"); - bokeh_modes.push_back("\n#define MODE_COMPOSITE_BOKEH\n"); + // Initialize bokeh + Vector<String> bokeh_modes; + bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_CIRCULAR\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_COMPOSITE_BOKEH\n"); + if (prefer_raster_effects) { + bokeh.raster_shader.initialize(bokeh_modes); - bokeh.shader.initialize(bokeh_modes); + bokeh.shader_version = bokeh.raster_shader.version_create(); - bokeh.shader_version = bokeh.shader.version_create(); + const int att_count[BOKEH_MAX] = { 1, 2, 1, 2, 1, 2, 1 }; + for (int i = 0; i < BOKEH_MAX; i++) { + RD::PipelineColorBlendState blend_state = (i == BOKEH_COMPOSITE) ? RD::PipelineColorBlendState::create_blend(att_count[i]) : RD::PipelineColorBlendState::create_disabled(att_count[i]); + bokeh.raster_pipelines[i].setup(bokeh.raster_shader.version_get_shader(bokeh.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } else { + bokeh.compute_shader.initialize(bokeh_modes); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_BOX_NOWEIGHT, false); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, false); + bokeh.shader_version = bokeh.compute_shader.version_create(); for (int i = 0; i < BOKEH_MAX; i++) { - bokeh.pipelines[i] = RD::get_singleton()->compute_pipeline_create(bokeh.shader.version_get_shader(bokeh.shader_version, i)); + if (bokeh.compute_shader.is_variant_enabled(i)) { + bokeh.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(bokeh.compute_shader.version_get_shader(bokeh.shader_version, i)); + } + } + + for (int i = 0; i < BOKEH_MAX; i++) { + bokeh.raster_pipelines[i].clear(); } } - { + if (!prefer_raster_effects) { // Initialize ssao RD::SamplerState sampler; @@ -1756,7 +2259,7 @@ EffectsRD::EffectsRD() { ERR_FAIL_COND(pipeline != SSAO_MAX); } - { + if (!prefer_raster_effects) { // Initialize roughness limiter Vector<String> shader_modes; shader_modes.push_back(""); @@ -1772,11 +2275,21 @@ EffectsRD::EffectsRD() { //Initialize cubemap downsampler Vector<String> cubemap_downsampler_modes; cubemap_downsampler_modes.push_back(""); - cubemap_downsampler.shader.initialize(cubemap_downsampler_modes); - cubemap_downsampler.shader_version = cubemap_downsampler.shader.version_create(); + if (prefer_raster_effects) { + cubemap_downsampler.raster_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.raster_shader.version_create(); - cubemap_downsampler.pipeline = RD::get_singleton()->compute_pipeline_create(cubemap_downsampler.shader.version_get_shader(cubemap_downsampler.shader_version, 0)); + cubemap_downsampler.raster_pipeline.setup(cubemap_downsampler.raster_shader.version_get_shader(cubemap_downsampler.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + cubemap_downsampler.compute_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.compute_shader.version_create(); + + cubemap_downsampler.compute_pipeline = RD::get_singleton()->compute_pipeline_create(cubemap_downsampler.compute_shader.version_get_shader(cubemap_downsampler.shader_version, 0)); + cubemap_downsampler.raster_pipeline.clear(); + } } { @@ -1788,12 +2301,6 @@ EffectsRD::EffectsRD() { cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n"); cubemap_filter_modes.push_back("\n#define USE_HIGH_QUALITY\n#define USE_TEXTURE_ARRAY\n"); cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n#define USE_TEXTURE_ARRAY\n"); - filter.shader.initialize(cubemap_filter_modes); - filter.shader_version = filter.shader.version_create(); - - for (int i = 0; i < FILTER_MODE_MAX; i++) { - filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(filter.shader.version_get_shader(filter.shader_version, i)); - } if (filter.use_high_quality) { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(high_quality_coeffs)); @@ -1803,18 +2310,54 @@ EffectsRD::EffectsRD() { RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); } - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 0; - u.ids.push_back(filter.coefficient_buffer); - uniforms.push_back(u); + if (prefer_raster_effects) { + filter.raster_shader.initialize(cubemap_filter_modes); + + // array variants are not supported in raster + filter.raster_shader.set_variant_enabled(FILTER_MODE_HIGH_QUALITY_ARRAY, false); + filter.raster_shader.set_variant_enabled(FILTER_MODE_LOW_QUALITY_ARRAY, false); + + filter.shader_version = filter.raster_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + if (filter.raster_shader.is_variant_enabled(i)) { + filter.raster_pipelines[i].setup(filter.raster_shader.version_get_shader(filter.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + filter.raster_pipelines[i].clear(); + } + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.ids.push_back(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.raster_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); + } else { + filter.compute_shader.initialize(cubemap_filter_modes); + filter.shader_version = filter.compute_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + filter.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(filter.compute_shader.version_get_shader(filter.shader_version, i)); + filter.raster_pipelines[i].clear(); + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.ids.push_back(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); } - filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); } - { + if (!prefer_raster_effects) { Vector<String> specular_modes; specular_modes.push_back("\n#define MODE_MERGE\n"); specular_modes.push_back("\n#define MODE_MERGE\n#define MODE_SSR\n"); @@ -1850,71 +2393,74 @@ EffectsRD::EffectsRD() { } } - { - Vector<String> ssr_modes; - ssr_modes.push_back("\n"); - ssr_modes.push_back("\n#define MODE_ROUGH\n"); + if (!prefer_raster_effects) { + { + Vector<String> ssr_modes; + ssr_modes.push_back("\n"); + ssr_modes.push_back("\n#define MODE_ROUGH\n"); - ssr.shader.initialize(ssr_modes); + ssr.shader.initialize(ssr_modes); - ssr.shader_version = ssr.shader.version_create(); + ssr.shader_version = ssr.shader.version_create(); - for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { - ssr.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i)); + for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { + ssr.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i)); + } } - } - { - Vector<String> ssr_filter_modes; - ssr_filter_modes.push_back("\n"); - ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); + { + Vector<String> ssr_filter_modes; + ssr_filter_modes.push_back("\n"); + ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); - ssr_filter.shader.initialize(ssr_filter_modes); + ssr_filter.shader.initialize(ssr_filter_modes); - ssr_filter.shader_version = ssr_filter.shader.version_create(); + ssr_filter.shader_version = ssr_filter.shader.version_create(); - for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { - ssr_filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i)); + for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { + ssr_filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i)); + } } - } - { - Vector<String> ssr_scale_modes; - ssr_scale_modes.push_back("\n"); + { + Vector<String> ssr_scale_modes; + ssr_scale_modes.push_back("\n"); - ssr_scale.shader.initialize(ssr_scale_modes); + ssr_scale.shader.initialize(ssr_scale_modes); - ssr_scale.shader_version = ssr_scale.shader.version_create(); + ssr_scale.shader_version = ssr_scale.shader.version_create(); - ssr_scale.pipeline = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0)); - } + ssr_scale.pipeline = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0)); + } - { - Vector<String> sss_modes; - sss_modes.push_back("\n#define USE_11_SAMPLES\n"); - sss_modes.push_back("\n#define USE_17_SAMPLES\n"); - sss_modes.push_back("\n#define USE_25_SAMPLES\n"); + { + Vector<String> sss_modes; + sss_modes.push_back("\n#define USE_11_SAMPLES\n"); + sss_modes.push_back("\n#define USE_17_SAMPLES\n"); + sss_modes.push_back("\n#define USE_25_SAMPLES\n"); - sss.shader.initialize(sss_modes); + sss.shader.initialize(sss_modes); - sss.shader_version = sss.shader.version_create(); + sss.shader_version = sss.shader.version_create(); - for (int i = 0; i < sss_modes.size(); i++) { - sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + for (int i = 0; i < sss_modes.size(); i++) { + sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + } } - } - { - Vector<String> resolve_modes; - resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n"); - resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define GIPROBE_RESOLVE\n"); + { + Vector<String> resolve_modes; + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define VOXEL_GI_RESOLVE\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_DEPTH\n"); - resolve.shader.initialize(resolve_modes); + resolve.shader.initialize(resolve_modes); - resolve.shader_version = resolve.shader.version_create(); + resolve.shader_version = resolve.shader.version_create(); - for (int i = 0; i < RESOLVE_MODE_MAX; i++) { - resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i)); + for (int i = 0; i < RESOLVE_MODE_MAX; i++) { + resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i)); + } } } @@ -1980,30 +2526,41 @@ EffectsRD::~EffectsRD() { RD::get_singleton()->free(index_buffer); //array gets freed as dependency RD::get_singleton()->free(filter.coefficient_buffer); - RD::get_singleton()->free(ssao.mirror_sampler); - RD::get_singleton()->free(ssao.gather_constants_buffer); - RD::get_singleton()->free(ssao.importance_map_load_counter); - - bokeh.shader.version_free(bokeh.shader_version); - copy.shader.version_free(copy.shader_version); + if (prefer_raster_effects) { + blur_raster.shader.version_free(blur_raster.shader_version); + bokeh.raster_shader.version_free(blur_raster.shader_version); + luminance_reduce_raster.shader.version_free(luminance_reduce_raster.shader_version); + roughness.raster_shader.version_free(roughness.shader_version); + cubemap_downsampler.raster_shader.version_free(cubemap_downsampler.shader_version); + filter.raster_shader.version_free(filter.shader_version); + } else { + bokeh.compute_shader.version_free(bokeh.shader_version); + luminance_reduce.shader.version_free(luminance_reduce.shader_version); + roughness.compute_shader.version_free(roughness.shader_version); + cubemap_downsampler.compute_shader.version_free(cubemap_downsampler.shader_version); + filter.compute_shader.version_free(filter.shader_version); + } + if (!prefer_raster_effects) { + copy.shader.version_free(copy.shader_version); + resolve.shader.version_free(resolve.shader_version); + specular_merge.shader.version_free(specular_merge.shader_version); + ssao.blur_shader.version_free(ssao.blur_shader_version); + ssao.gather_shader.version_free(ssao.gather_shader_version); + ssao.downsample_shader.version_free(ssao.downsample_shader_version); + ssao.interleave_shader.version_free(ssao.interleave_shader_version); + ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); + roughness_limiter.shader.version_free(roughness_limiter.shader_version); + ssr.shader.version_free(ssr.shader_version); + ssr_filter.shader.version_free(ssr_filter.shader_version); + ssr_scale.shader.version_free(ssr_scale.shader_version); + sss.shader.version_free(sss.shader_version); + + RD::get_singleton()->free(ssao.mirror_sampler); + RD::get_singleton()->free(ssao.gather_constants_buffer); + RD::get_singleton()->free(ssao.importance_map_load_counter); + } copy_to_fb.shader.version_free(copy_to_fb.shader_version); cube_to_dp.shader.version_free(cube_to_dp.shader_version); - cubemap_downsampler.shader.version_free(cubemap_downsampler.shader_version); - filter.shader.version_free(filter.shader_version); - luminance_reduce.shader.version_free(luminance_reduce.shader_version); - resolve.shader.version_free(resolve.shader_version); - roughness.shader.version_free(roughness.shader_version); - roughness_limiter.shader.version_free(roughness_limiter.shader_version); sort.shader.version_free(sort.shader_version); - specular_merge.shader.version_free(specular_merge.shader_version); - ssao.blur_shader.version_free(ssao.blur_shader_version); - ssao.gather_shader.version_free(ssao.gather_shader_version); - ssao.downsample_shader.version_free(ssao.downsample_shader_version); - ssao.interleave_shader.version_free(ssao.interleave_shader_version); - ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); - ssr.shader.version_free(ssr.shader_version); - ssr_filter.shader.version_free(ssr_filter.shader_version); - ssr_scale.shader.version_free(ssr_scale.shader_version); - sss.shader.version_free(sss.shader_version); tonemap.shader.version_free(tonemap.shader_version); } diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index 1ba25e301b..c8d4cb7ad4 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -33,14 +33,20 @@ #include "core/math/camera_matrix.h" #include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/blur_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/bokeh_dof.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/copy.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/copy_to_fb.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cube_to_dp.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_filter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/luminance_reduce.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/resolve.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/roughness_limiter.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl.gen.h" @@ -55,10 +61,55 @@ #include "servers/rendering/renderer_rd/shaders/ssao_interleave.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/tonemap.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" #include "servers/rendering_server.h" class EffectsRD { +private: + bool prefer_raster_effects; + + enum BlurRasterMode { + BLUR_MIPMAP, + + BLUR_MODE_GAUSSIAN_BLUR, + BLUR_MODE_GAUSSIAN_GLOW, + BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, + BLUR_MODE_COPY, + + BLUR_MODE_MAX + }; + + enum { + BLUR_FLAG_HORIZONTAL = (1 << 0), + BLUR_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 1), + BLUR_FLAG_GLOW_FIRST_PASS = (1 << 2), + }; + + struct BlurRasterPushConstant { + float pixel_size[2]; + uint32_t flags; + uint32_t pad; + + //glow + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_grey; + }; + + struct BlurRaster { + BlurRasterPushConstant push_constant; + BlurRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[BLUR_MODE_MAX]; + } blur_raster; + enum CopyMode { COPY_MODE_GAUSSIAN_COPY, COPY_MODE_GAUSSIAN_COPY_8BIT, @@ -160,9 +211,11 @@ class EffectsRD { struct CubemapRoughness { CubemapRoughnessPushConstant push_constant; - CubemapRoughnessShaderRD shader; + CubemapRoughnessShaderRD compute_shader; + CubemapRoughnessRasterShaderRD raster_shader; RID shader_version; - RID pipeline; + RID compute_pipeline; + PipelineCacheRD raster_pipeline; } roughness; enum TonemapMode { @@ -170,33 +223,43 @@ class EffectsRD { TONEMAP_MODE_BICUBIC_GLOW_FILTER, TONEMAP_MODE_1D_LUT, TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT, + TONEMAP_MODE_SUBPASS, + TONEMAP_MODE_SUBPASS_1D_LUT, + + TONEMAP_MODE_NORMAL_MULTIVIEW, + TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, + TONEMAP_MODE_1D_LUT_MULTIVIEW, + TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, + TONEMAP_MODE_SUBPASS_MULTIVIEW, + TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, + TONEMAP_MODE_MAX }; struct TonemapPushConstant { - float bcs[3]; - uint32_t use_bcs; + float bcs[3]; // 12 - 12 + uint32_t use_bcs; // 4 - 16 - uint32_t use_glow; - uint32_t use_auto_exposure; - uint32_t use_color_correction; - uint32_t tonemapper; + uint32_t use_glow; // 4 - 20 + uint32_t use_auto_exposure; // 4 - 24 + uint32_t use_color_correction; // 4 - 28 + uint32_t tonemapper; // 4 - 32 - uint32_t glow_texture_size[2]; - float glow_intensity; - uint32_t pad3; + uint32_t glow_texture_size[2]; // 8 - 40 + float glow_intensity; // 4 - 44 + uint32_t pad3; // 4 - 48 - uint32_t glow_mode; - float glow_levels[7]; + uint32_t glow_mode; // 4 - 52 + float glow_levels[7]; // 28 - 80 - float exposure; - float white; - float auto_exposure_grey; - uint32_t pad2; + float exposure; // 4 - 84 + float white; // 4 - 88 + float auto_exposure_grey; // 4 - 92 + uint32_t pad2; // 4 - 96 - float pixel_size[2]; - uint32_t use_fxaa; - uint32_t use_debanding; + float pixel_size[2]; // 8 - 104 + uint32_t use_fxaa; // 4 - 108 + uint32_t use_debanding; // 4 - 112 }; /* tonemap actually writes to a framebuffer, which is @@ -232,11 +295,33 @@ class EffectsRD { RID pipelines[LUMINANCE_REDUCE_MAX]; } luminance_reduce; + enum LuminanceReduceRasterMode { + LUMINANCE_REDUCE_FRAGMENT_FIRST, + LUMINANCE_REDUCE_FRAGMENT, + LUMINANCE_REDUCE_FRAGMENT_FINAL, + LUMINANCE_REDUCE_FRAGMENT_MAX + }; + + struct LuminanceReduceRasterPushConstant { + int32_t source_size[2]; + int32_t dest_size[2]; + float exposure_adjust; + float min_luminance; + float max_luminance; + float pad[1]; + }; + + struct LuminanceReduceFragment { + LuminanceReduceRasterPushConstant push_constant; + LuminanceReduceRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[LUMINANCE_REDUCE_FRAGMENT_MAX]; + } luminance_reduce_raster; + struct CopyToDPPushConstant { float z_far; float z_near; - uint32_t z_flip; - uint32_t pad; + float texel_size[2]; float screen_rect[4]; }; @@ -274,7 +359,9 @@ class EffectsRD { enum BokehMode { BOKEH_GEN_BLUR_SIZE, BOKEH_GEN_BOKEH_BOX, + BOKEH_GEN_BOKEH_BOX_NOWEIGHT, BOKEH_GEN_BOKEH_HEXAGONAL, + BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, BOKEH_GEN_BOKEH_CIRCULAR, BOKEH_COMPOSITE, BOKEH_MAX @@ -282,9 +369,11 @@ class EffectsRD { struct Bokeh { BokehPushConstant push_constant; - BokehDofShaderRD shader; + BokehDofShaderRD compute_shader; + BokehDofRasterShaderRD raster_shader; RID shader_version; - RID pipelines[BOKEH_MAX]; + RID compute_pipelines[BOKEH_MAX]; + PipelineCacheRD raster_pipelines[BOKEH_MAX]; } bokeh; enum SSAOMode { @@ -378,12 +467,10 @@ class EffectsRD { SSAODownsamplePushConstant downsample_push_constant; SsaoDownsampleShaderRD downsample_shader; RID downsample_shader_version; - RID downsample_uniform_set; SSAOGatherPushConstant gather_push_constant; SsaoShaderRD gather_shader; RID gather_shader_version; - RID gather_uniform_set; RID gather_constants_buffer; bool gather_initialized = false; @@ -391,7 +478,6 @@ class EffectsRD { SsaoImportanceMapShaderRD importance_map_shader; RID importance_map_shader_version; RID importance_map_load_counter; - RID importance_map_uniform_set; RID counter_uniform_set; SSAOBlurPushConstant blur_push_constant; @@ -422,15 +508,17 @@ class EffectsRD { struct CubemapDownsamplerPushConstant { uint32_t face_size; - float pad[3]; + uint32_t face_id; + float pad[2]; }; struct CubemapDownsampler { CubemapDownsamplerPushConstant push_constant; - CubemapDownsamplerShaderRD shader; + CubemapDownsamplerShaderRD compute_shader; + CubemapDownsamplerRasterShaderRD raster_shader; RID shader_version; - RID pipeline; - + RID compute_pipeline; + PipelineCacheRD raster_pipeline; } cubemap_downsampler; enum CubemapFilterMode { @@ -441,10 +529,19 @@ class EffectsRD { FILTER_MODE_MAX, }; + struct CubemapFilterRasterPushConstant { + uint32_t mip_level; + uint32_t face_id; + float pad[2]; + }; + struct CubemapFilter { - CubemapFilterShaderRD shader; + CubemapFilterShaderRD compute_shader; + CubemapFilterRasterShaderRD raster_shader; RID shader_version; - RID pipelines[FILTER_MODE_MAX]; + RID compute_pipelines[FILTER_MODE_MAX]; + PipelineCacheRD raster_pipelines[FILTER_MODE_MAX]; + RID uniform_set; RID image_uniform_set; RID coefficient_buffer; @@ -452,15 +549,6 @@ class EffectsRD { } filter; - struct SkyPushConstant { - float orientation[12]; - float proj[4]; - float position[3]; - float multiplier; - float time; - float pad[3]; - }; - enum SpecularMergeMode { SPECULAR_MERGE_ADD, SPECULAR_MERGE_SSR, @@ -585,7 +673,8 @@ class EffectsRD { enum ResolveMode { RESOLVE_MODE_GI, - RESOLVE_MODE_GI_GIPROBE, + RESOLVE_MODE_GI_VOXEL_GI, + RESOLVE_MODE_DEPTH, RESOLVE_MODE_MAX }; @@ -621,6 +710,7 @@ class EffectsRD { RID index_array; Map<RID, RID> texture_to_uniform_set_cache; + Map<RID, RID> input_to_uniform_set_cache; Map<RID, RID> image_to_uniform_set_cache; @@ -654,6 +744,7 @@ class EffectsRD { Map<TextureSamplerPair, RID> texture_sampler_to_compute_uniform_set_cache; RID _get_uniform_set_from_image(RID p_texture); + RID _get_uniform_set_for_input(RID p_texture); RID _get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); RID _get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); RID _get_compute_uniform_set_from_texture_and_sampler(RID p_texture, RID p_sampler); @@ -661,6 +752,8 @@ class EffectsRD { RID _get_compute_uniform_set_from_image_pair(RID p_texture, RID p_texture2); public: + bool get_prefer_raster_effects(); + void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false, bool p_srgb = false, RID p_secondary = RID()); void copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_all_source = false, bool p_8_bit_dst = false, bool p_alpha_to_one = false); void copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array); @@ -670,12 +763,36 @@ public: void gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back_texture, const Rect2i &p_region, bool p_8bit_dst = false); void set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst = false); void gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_treshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); + void gaussian_glow_raster(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_treshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); - void cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + void cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + void cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip); + void make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size); + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip); void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); - void bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); + + struct BokehBuffers { + // bokeh buffers + + // textures + Size2i base_texture_size; + RID base_texture; + RID depth_texture; + RID secondary_texture; + RID half_texture[2]; + + // raster only + RID base_fb; + RID secondary_fb; // with weights + RID half_fb[2]; // with weights + RID base_weight_fb; + RID weight_texture[4]; + }; + + void bokeh_dof(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); struct TonemapSettings { bool use_glow = false; @@ -714,6 +831,7 @@ public: bool use_fxaa = false; bool use_debanding = false; Vector2i texture_size; + uint32_t view_count = 1; }; struct SSAOSettings { @@ -737,24 +855,27 @@ public: }; void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings); + void tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings); - void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass); - void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets); + void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set); + void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets, RID &r_downsample_uniform_set, RID &r_gather_uniform_set, RID &r_importance_map_uniform_set); void roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve); void cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size); + void cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size); void cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array); - void render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_fog, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, const CameraMatrix &p_camera, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position); + void cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level); void screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RS::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera); void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection); void sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RS::SubSurfaceScatteringQuality p_quality); - void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); void sort_buffer(RID p_uniform_set, int p_size); - EffectsRD(); + EffectsRD(bool p_prefer_raster_effects); ~EffectsRD(); }; diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp index cdff3139eb..611f7c6494 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -93,8 +93,8 @@ void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_specular() } } -void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_giprobe() { - if (!giprobe_buffer.is_valid()) { +void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_voxelgi() { + if (!voxelgi_buffer.is_valid()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8G8_UINT; tf.width = width; @@ -105,41 +105,41 @@ void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_giprobe() RD::TextureFormat tf_aa = tf; tf_aa.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; tf_aa.samples = texture_samples; - giprobe_buffer_msaa = RD::get_singleton()->texture_create(tf_aa, RD::TextureView()); + voxelgi_buffer_msaa = RD::get_singleton()->texture_create(tf_aa, RD::TextureView()); } else { tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; } tf.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT; - giprobe_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + voxelgi_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); Vector<RID> fb; if (msaa != RS::VIEWPORT_MSAA_DISABLED) { fb.push_back(depth_msaa); fb.push_back(normal_roughness_buffer_msaa); - fb.push_back(giprobe_buffer_msaa); + fb.push_back(voxelgi_buffer_msaa); } else { fb.push_back(depth); fb.push_back(normal_roughness_buffer); - fb.push_back(giprobe_buffer); + fb.push_back(voxelgi_buffer); } - depth_normal_roughness_giprobe_fb = RD::get_singleton()->framebuffer_create(fb); + depth_normal_roughness_voxelgi_fb = RD::get_singleton()->framebuffer_create(fb); } } void RenderForwardClustered::RenderBufferDataForwardClustered::clear() { - if (giprobe_buffer != RID()) { - RD::get_singleton()->free(giprobe_buffer); - giprobe_buffer = RID(); + if (voxelgi_buffer != RID()) { + RD::get_singleton()->free(voxelgi_buffer); + voxelgi_buffer = RID(); - if (giprobe_buffer_msaa.is_valid()) { - RD::get_singleton()->free(giprobe_buffer_msaa); - giprobe_buffer_msaa = RID(); + if (voxelgi_buffer_msaa.is_valid()) { + RD::get_singleton()->free(voxelgi_buffer_msaa); + voxelgi_buffer_msaa = RID(); } - depth_normal_roughness_giprobe_fb = RID(); + depth_normal_roughness_voxelgi_fb = RID(); } if (color_msaa.is_valid()) { @@ -183,9 +183,11 @@ void RenderForwardClustered::RenderBufferDataForwardClustered::clear() { } } -void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa) { +void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { clear(); + ERR_FAIL_COND_MSG(p_view_count != 1, "Multiple views is currently not supported in this renderer, please use the mobile renderer for VR support"); + msaa = p_msaa; width = p_width; @@ -284,8 +286,6 @@ void RenderForwardClustered::_allocate_normal_roughness_texture(RenderBufferData fb.push_back(rb->normal_roughness_buffer_msaa); rb->depth_normal_roughness_fb = RD::get_singleton()->framebuffer_create(fb); } - - _render_buffers_clear_uniform_set(rb); } RendererSceneRenderRD::RenderBufferData *RenderForwardClustered::_create_render_buffer_data() { @@ -318,11 +318,11 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p RID prev_pipeline_rd; RID prev_xforms_uniform_set; - bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); + bool shadow_pass = (p_pass_mode == PASS_MODE_SHADOW) || (p_pass_mode == PASS_MODE_SHADOW_DP); SceneState::PushConstant push_constant; - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL) { push_constant.uv_offset = Math::make_half_float(p_params->uv_offset.y) << 16; push_constant.uv_offset |= Math::make_half_float(p_params->uv_offset.x); } else { @@ -339,14 +339,26 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p SceneShaderForwardClustered::ShaderData *shader; void *mesh_surface; - if (shadow_pass || p_params->pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too + if (shadow_pass || p_pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too material_uniform_set = surf->material_uniform_set_shadow; shader = surf->shader_shadow; mesh_surface = surf->surface_shadow; } else { - material_uniform_set = surf->material_uniform_set; - shader = surf->shader; +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif mesh_surface = surf->surface; } @@ -357,7 +369,7 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p //find cull variant SceneShaderForwardClustered::ShaderData::CullVariant cull_variant; - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL || p_params->pass_mode == PASS_MODE_SDF || ((p_params->pass_mode == PASS_MODE_SHADOW || p_params->pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { + if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || p_pass_mode == PASS_MODE_SDF || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { cull_variant = SceneShaderForwardClustered::ShaderData::CULL_VARIANT_DOUBLE_SIDED; } else { bool mirror = surf->owner->mirror; @@ -372,14 +384,30 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p SceneShaderForwardClustered::ShaderVersion shader_version = SceneShaderForwardClustered::SHADER_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. - switch (p_params->pass_mode) { + uint32_t pipeline_specialization = 0; + + if (p_pass_mode == PASS_MODE_COLOR || p_pass_mode == PASS_MODE_COLOR_TRANSPARENT || p_pass_mode == PASS_MODE_COLOR_SPECULAR) { + if (element_info.uses_softshadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_SOFT_SHADOWS; + } + if (element_info.uses_projector) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_PROJECTOR; + } + + if (p_params->use_directional_soft_shadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS; + } + } + + switch (p_pass_mode) { case PASS_MODE_COLOR: case PASS_MODE_COLOR_TRANSPARENT: { if (element_info.uses_lightmap) { shader_version = SceneShaderForwardClustered::SHADER_VERSION_LIGHTMAP_COLOR_PASS; - } else if (element_info.uses_forward_gi) { - shader_version = SceneShaderForwardClustered::SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI; } else { + if (element_info.uses_forward_gi) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_FORWARD_GI; + } shader_version = SceneShaderForwardClustered::SHADER_VERSION_COLOR_PASS; } } break; @@ -400,8 +428,8 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { shader_version = SceneShaderForwardClustered::SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS; } break; - case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: { - shader_version = SceneShaderForwardClustered::SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { + shader_version = SceneShaderForwardClustered::SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI; } break; case PASS_MODE_DEPTH_MATERIAL: { shader_version = SceneShaderForwardClustered::SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL; @@ -440,7 +468,7 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p prev_index_array_rd = index_array_rd; } - RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe); + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, 0, pipeline_specialization); if (pipeline_rd != prev_pipeline_rd) { // checking with prev shader does not make so much sense, as @@ -455,8 +483,8 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p } if (material_uniform_set != prev_material_uniform_set) { - //update uniform set - if (material_uniform_set.is_valid()) { + // Update uniform set. + if (RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); } @@ -466,6 +494,10 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS) { + instance_count /= surf->owner->trail_steps; + } + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); i += element_info.repeat - 1; //skip equal elements } @@ -496,8 +528,8 @@ void RenderForwardClustered::_render_list(RenderingDevice::DrawListID p_draw_lis case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); } break; - case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: { - _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { + _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); } break; case PASS_MODE_DEPTH_MATERIAL: { _render_list_template<PASS_MODE_DEPTH_MATERIAL>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); @@ -534,21 +566,21 @@ void RenderForwardClustered::_render_list_with_threads(RenderListParameters *p_p } } -void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { - //CameraMatrix projection = p_cam_projection; +void RenderForwardClustered::_setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { + //CameraMatrix projection = p_render_data->cam_projection; //projection.flip_y(); // Vulkan and modern APIs use Y-Down CameraMatrix correction; correction.set_depth_correction(p_flip_y); - CameraMatrix projection = correction * p_cam_projection; + CameraMatrix projection = correction * p_render_data->cam_projection; //store camera into ubo RendererStorageRD::store_camera(projection, scene_state.ubo.projection_matrix); RendererStorageRD::store_camera(projection.inverse(), scene_state.ubo.inv_projection_matrix); - RendererStorageRD::store_transform(p_cam_transform, scene_state.ubo.camera_matrix); - RendererStorageRD::store_transform(p_cam_transform.affine_inverse(), scene_state.ubo.inv_camera_matrix); + RendererStorageRD::store_transform(p_render_data->cam_transform, scene_state.ubo.camera_matrix); + RendererStorageRD::store_transform(p_render_data->cam_transform.affine_inverse(), scene_state.ubo.inv_camera_matrix); - scene_state.ubo.z_far = p_zfar; - scene_state.ubo.z_near = p_znear; + scene_state.ubo.z_far = p_render_data->z_far; + scene_state.ubo.z_near = p_render_data->z_near; scene_state.ubo.pancake_shadows = p_pancake_shadows; @@ -557,26 +589,21 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ RendererStorageRD::store_soft_shadow_kernel(penumbra_shadow_kernel_get(), scene_state.ubo.penumbra_shadow_kernel); RendererStorageRD::store_soft_shadow_kernel(soft_shadow_kernel_get(), scene_state.ubo.soft_shadow_kernel); - scene_state.ubo.directional_penumbra_shadow_samples = directional_penumbra_shadow_samples_get(); - scene_state.ubo.directional_soft_shadow_samples = directional_soft_shadow_samples_get(); - scene_state.ubo.penumbra_shadow_samples = penumbra_shadow_samples_get(); - scene_state.ubo.soft_shadow_samples = soft_shadow_samples_get(); - Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; - scene_state.ubo.cluster_shift = get_shift_from_power_of_2(p_cluster_size); - scene_state.ubo.max_cluster_element_count_div_32 = p_max_cluster_elements / 32; + scene_state.ubo.cluster_shift = get_shift_from_power_of_2(p_render_data->cluster_size); + scene_state.ubo.max_cluster_element_count_div_32 = p_render_data->cluster_max_elements / 32; { - uint32_t cluster_screen_width = (p_screen_size.width - 1) / p_cluster_size + 1; - uint32_t cluster_screen_height = (p_screen_size.height - 1) / p_cluster_size + 1; + uint32_t cluster_screen_width = (p_screen_size.width - 1) / p_render_data->cluster_size + 1; + uint32_t cluster_screen_height = (p_screen_size.height - 1) / p_render_data->cluster_size + 1; scene_state.ubo.cluster_type_size = cluster_screen_width * cluster_screen_height * (scene_state.ubo.max_cluster_element_count_div_32 + 32); scene_state.ubo.cluster_width = cluster_screen_width; } - if (p_shadow_atlas.is_valid()) { - Vector2 sas = shadow_atlas_get_size(p_shadow_atlas); + if (p_render_data->shadow_atlas.is_valid()) { + Vector2 sas = shadow_atlas_get_size(p_render_data->shadow_atlas); scene_state.ubo.shadow_atlas_pixel_size[0] = 1.0 / sas.x; scene_state.ubo.shadow_atlas_pixel_size[1] = 1.0 / sas.y; } @@ -592,22 +619,22 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ scene_state.ubo.volumetric_fog_enabled = false; scene_state.ubo.fog_enabled = false; - if (p_render_buffers.is_valid()) { - RenderBufferDataForwardClustered *render_buffers = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffers); + if (p_render_data->render_buffers.is_valid()) { + RenderBufferDataForwardClustered *render_buffers = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_data->render_buffers); if (render_buffers->msaa != RS::VIEWPORT_MSAA_DISABLED) { scene_state.ubo.gi_upscale_for_msaa = true; } - if (render_buffers_has_volumetric_fog(p_render_buffers)) { + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers)) { scene_state.ubo.volumetric_fog_enabled = true; - float fog_end = render_buffers_get_volumetric_fog_end(p_render_buffers); + float fog_end = render_buffers_get_volumetric_fog_end(p_render_data->render_buffers); if (fog_end > 0.0) { scene_state.ubo.volumetric_fog_inv_length = 1.0 / fog_end; } else { scene_state.ubo.volumetric_fog_inv_length = 1.0; } - float fog_detail_spread = render_buffers_get_volumetric_fog_detail_spread(p_render_buffers); //reverse lookup + float fog_detail_spread = render_buffers_get_volumetric_fog_detail_spread(p_render_data->render_buffers); //reverse lookup if (fog_detail_spread > 0.0) { scene_state.ubo.volumetric_fog_detail_spread = 1.0 / fog_detail_spread; } else { @@ -616,26 +643,26 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ } } #if 0 - if (p_render_buffers.is_valid() && render_buffers_is_sdfgi_enabled(p_render_buffers)) { - scene_state.ubo.sdfgi_cascade_count = render_buffers_get_sdfgi_cascade_count(p_render_buffers); - scene_state.ubo.sdfgi_probe_axis_size = render_buffers_get_sdfgi_cascade_probe_count(p_render_buffers); + if (p_render_data->render_buffers.is_valid() && render_buffers_is_sdfgi_enabled(p_render_data->render_buffers)) { + scene_state.ubo.sdfgi_cascade_count = render_buffers_get_sdfgi_cascade_count(p_render_data->render_buffers); + scene_state.ubo.sdfgi_probe_axis_size = render_buffers_get_sdfgi_cascade_probe_count(p_render_data->render_buffers); scene_state.ubo.sdfgi_cascade_probe_size[0] = scene_state.ubo.sdfgi_probe_axis_size - 1; //float version for performance scene_state.ubo.sdfgi_cascade_probe_size[1] = scene_state.ubo.sdfgi_probe_axis_size - 1; scene_state.ubo.sdfgi_cascade_probe_size[2] = scene_state.ubo.sdfgi_probe_axis_size - 1; - float csize = render_buffers_get_sdfgi_cascade_size(p_render_buffers); + float csize = render_buffers_get_sdfgi_cascade_size(p_render_data->render_buffers); scene_state.ubo.sdfgi_probe_to_uvw = 1.0 / float(scene_state.ubo.sdfgi_cascade_probe_size[0]); float occ_bias = 0.0; scene_state.ubo.sdfgi_occlusion_bias = occ_bias / csize; - scene_state.ubo.sdfgi_use_occlusion = render_buffers_is_sdfgi_using_occlusion(p_render_buffers); - scene_state.ubo.sdfgi_energy = render_buffers_get_sdfgi_energy(p_render_buffers); + scene_state.ubo.sdfgi_use_occlusion = render_buffers_is_sdfgi_using_occlusion(p_render_data->render_buffers); + scene_state.ubo.sdfgi_energy = render_buffers_get_sdfgi_energy(p_render_data->render_buffers); float cascade_voxel_size = (csize / scene_state.ubo.sdfgi_cascade_probe_size[0]); float occlusion_clamp = (cascade_voxel_size - 0.5) / cascade_voxel_size; scene_state.ubo.sdfgi_occlusion_clamp[0] = occlusion_clamp; scene_state.ubo.sdfgi_occlusion_clamp[1] = occlusion_clamp; scene_state.ubo.sdfgi_occlusion_clamp[2] = occlusion_clamp; - scene_state.ubo.sdfgi_normal_bias = (render_buffers_get_sdfgi_normal_bias(p_render_buffers) / csize) * scene_state.ubo.sdfgi_cascade_probe_size[0]; + scene_state.ubo.sdfgi_normal_bias = (render_buffers_get_sdfgi_normal_bias(p_render_data->render_buffers) / csize) * scene_state.ubo.sdfgi_cascade_probe_size[0]; //vec2 tex_pixel_size = 1.0 / vec2(ivec2( (OCT_SIZE+2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE+2) * params.probe_axis_size ) ); //vec3 probe_uv_offset = (ivec3(OCT_SIZE+2,OCT_SIZE+2,(OCT_SIZE+2) * params.probe_axis_size)) * tex_pixel_size.xyx; @@ -656,14 +683,14 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ for (uint32_t i = 0; i < scene_state.ubo.sdfgi_cascade_count; i++) { SceneState::UBO::SDFGICascade &c = scene_state.ubo.sdfgi_cascades[i]; - Vector3 pos = render_buffers_get_sdfgi_cascade_offset(p_render_buffers, i); - pos -= p_cam_transform.origin; //make pos local to camera, to reduce numerical error + Vector3 pos = render_buffers_get_sdfgi_cascade_offset(p_render_data->render_buffers, i); + pos -= p_render_data->cam_transform.origin; //make pos local to camera, to reduce numerical error c.position[0] = pos.x; c.position[1] = pos.y; c.position[2] = pos.z; - c.to_probe = 1.0 / render_buffers_get_sdfgi_cascade_probe_size(p_render_buffers, i); + c.to_probe = 1.0 / render_buffers_get_sdfgi_cascade_probe_size(p_render_data->render_buffers, i); - Vector3i probe_ofs = render_buffers_get_sdfgi_cascade_probe_offset(p_render_buffers, i); + Vector3i probe_ofs = render_buffers_get_sdfgi_cascade_probe_offset(p_render_data->render_buffers, i); c.probe_world_offset[0] = probe_ofs.x; c.probe_world_offset[1] = probe_ofs.y; c.probe_world_offset[2] = probe_ofs.z; @@ -680,18 +707,18 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ scene_state.ubo.use_reflection_cubemap = false; scene_state.ubo.ssao_enabled = false; - } else if (is_environment(p_environment)) { - RS::EnvironmentBG env_bg = environment_get_background(p_environment); - RS::EnvironmentAmbientSource ambient_src = environment_get_ambient_source(p_environment); + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG env_bg = environment_get_background(p_render_data->environment); + RS::EnvironmentAmbientSource ambient_src = environment_get_ambient_source(p_render_data->environment); - float bg_energy = environment_get_bg_energy(p_environment); + float bg_energy = environment_get_bg_energy(p_render_data->environment); scene_state.ubo.ambient_light_color_energy[3] = bg_energy; - scene_state.ubo.ambient_color_sky_mix = environment_get_ambient_sky_contribution(p_environment); + scene_state.ubo.ambient_color_sky_mix = environment_get_ambient_sky_contribution(p_render_data->environment); //ambient if (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && (env_bg == RS::ENV_BG_CLEAR_COLOR || env_bg == RS::ENV_BG_COLOR)) { - Color color = env_bg == RS::ENV_BG_CLEAR_COLOR ? p_default_bg_color : environment_get_bg_color(p_environment); + Color color = env_bg == RS::ENV_BG_CLEAR_COLOR ? p_default_bg_color : environment_get_bg_color(p_render_data->environment); color = color.to_linear(); scene_state.ubo.ambient_light_color_energy[0] = color.r * bg_energy; @@ -700,15 +727,15 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ scene_state.ubo.use_ambient_light = true; scene_state.ubo.use_ambient_cubemap = false; } else { - float energy = environment_get_ambient_light_energy(p_environment); - Color color = environment_get_ambient_light_color(p_environment); + float energy = environment_get_ambient_light_energy(p_render_data->environment); + Color color = environment_get_ambient_light_color(p_render_data->environment); color = color.to_linear(); scene_state.ubo.ambient_light_color_energy[0] = color.r * energy; scene_state.ubo.ambient_light_color_energy[1] = color.g * energy; scene_state.ubo.ambient_light_color_energy[2] = color.b * energy; - Basis sky_transform = environment_get_sky_orientation(p_environment); - sky_transform = sky_transform.inverse() * p_cam_transform.basis; + Basis sky_transform = environment_get_sky_orientation(p_render_data->environment); + sky_transform = sky_transform.inverse() * p_render_data->cam_transform.basis; RendererStorageRD::store_transform_3x3(sky_transform, scene_state.ubo.radiance_inverse_xform); scene_state.ubo.use_ambient_cubemap = (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ambient_src == RS::ENV_AMBIENT_SOURCE_SKY; @@ -716,43 +743,43 @@ void RenderForwardClustered::_setup_environment(RID p_environment, RID p_render_ } //specular - RS::EnvironmentReflectionSource ref_src = environment_get_reflection_source(p_environment); + RS::EnvironmentReflectionSource ref_src = environment_get_reflection_source(p_render_data->environment); if ((ref_src == RS::ENV_REFLECTION_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ref_src == RS::ENV_REFLECTION_SOURCE_SKY) { scene_state.ubo.use_reflection_cubemap = true; } else { scene_state.ubo.use_reflection_cubemap = false; } - scene_state.ubo.ssao_enabled = p_opaque_render_buffers && environment_is_ssao_enabled(p_environment); - scene_state.ubo.ssao_ao_affect = environment_get_ssao_ao_affect(p_environment); - scene_state.ubo.ssao_light_affect = environment_get_ssao_light_affect(p_environment); + scene_state.ubo.ssao_enabled = p_opaque_render_buffers && environment_is_ssao_enabled(p_render_data->environment); + scene_state.ubo.ssao_ao_affect = environment_get_ssao_ao_affect(p_render_data->environment); + scene_state.ubo.ssao_light_affect = environment_get_ssao_light_affect(p_render_data->environment); - Color ao_color = environment_get_ao_color(p_environment).to_linear(); + Color ao_color = environment_get_ao_color(p_render_data->environment).to_linear(); scene_state.ubo.ao_color[0] = ao_color.r; scene_state.ubo.ao_color[1] = ao_color.g; scene_state.ubo.ao_color[2] = ao_color.b; scene_state.ubo.ao_color[3] = ao_color.a; - scene_state.ubo.fog_enabled = environment_is_fog_enabled(p_environment); - scene_state.ubo.fog_density = environment_get_fog_density(p_environment); - scene_state.ubo.fog_height = environment_get_fog_height(p_environment); - scene_state.ubo.fog_height_density = environment_get_fog_height_density(p_environment); + scene_state.ubo.fog_enabled = environment_is_fog_enabled(p_render_data->environment); + scene_state.ubo.fog_density = environment_get_fog_density(p_render_data->environment); + scene_state.ubo.fog_height = environment_get_fog_height(p_render_data->environment); + scene_state.ubo.fog_height_density = environment_get_fog_height_density(p_render_data->environment); if (scene_state.ubo.fog_height_density >= 0.0001) { scene_state.ubo.fog_height_density = 1.0 / scene_state.ubo.fog_height_density; } - scene_state.ubo.fog_aerial_perspective = environment_get_fog_aerial_perspective(p_environment); + scene_state.ubo.fog_aerial_perspective = environment_get_fog_aerial_perspective(p_render_data->environment); - Color fog_color = environment_get_fog_light_color(p_environment).to_linear(); - float fog_energy = environment_get_fog_light_energy(p_environment); + Color fog_color = environment_get_fog_light_color(p_render_data->environment).to_linear(); + float fog_energy = environment_get_fog_light_energy(p_render_data->environment); scene_state.ubo.fog_light_color[0] = fog_color.r * fog_energy; scene_state.ubo.fog_light_color[1] = fog_color.g * fog_energy; scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy; - scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_environment); + scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_render_data->environment); } else { - if (p_reflection_probe.is_valid() && storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_reflection_probe))) { + if (p_render_data->reflection_probe.is_valid() && storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { scene_state.ubo.use_ambient_light = false; } else { scene_state.ubo.use_ambient_light = true; @@ -797,13 +824,16 @@ void RenderForwardClustered::_update_instance_data_buffer(RenderListType p_rende RD::get_singleton()->buffer_update(scene_state.instance_buffer[p_render_list], 0, sizeof(SceneState::InstanceData) * scene_state.instance_data[p_render_list].size(), scene_state.instance_data[p_render_list].ptr(), RD::BARRIER_MASK_RASTER); } } -void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { +void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, int *p_render_info, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { RenderList *rl = &render_list[p_render_list]; uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); scene_state.instance_data[p_render_list].resize(p_offset + element_total); rl->element_info.resize(p_offset + element_total); + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] += element_total; + } uint32_t repeats = 0; GeometryInstanceSurfaceDataCache *prev_surface = nullptr; for (uint32_t i = 0; i < element_total; i++) { @@ -815,7 +845,7 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u if (inst->store_transform_cache) { RendererStorageRD::store_transform(inst->transform, instance_data.transform); } else { - RendererStorageRD::store_transform(Transform(), instance_data.transform); + RendererStorageRD::store_transform(Transform3D(), instance_data.transform); } instance_data.flags = inst->flags_cache; @@ -829,7 +859,7 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u bool cant_repeat = instance_data.flags & INSTANCE_DATA_FLAG_MULTIMESH || inst->mesh_instance.is_valid(); - if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2) { + if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2 && repeats < RenderElementInfo::MAX_REPEATS) { //this element is the same as the previous one, count repeats to draw it using instancing repeats++; } else { @@ -839,6 +869,9 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u } } repeats = 1; + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME]++; + } } RenderElementInfo &element_info = rl->element_info[p_offset + i]; @@ -846,6 +879,8 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u element_info.lod_index = surface->sort.lod_index; element_info.uses_forward_gi = surface->sort.uses_forward_gi; element_info.uses_lightmap = surface->sort.uses_lightmap; + element_info.uses_softshadow = surface->sort.uses_softshadow; + element_info.uses_projector = surface->sort.uses_projector; if (cant_repeat) { prev_surface = nullptr; @@ -865,7 +900,12 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u } } -void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi, const Plane &p_lod_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_append) { +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} +void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi, bool p_using_opaque_gi, bool p_append) { if (p_render_list == RENDER_LIST_OPAQUE) { scene_state.used_sss = false; scene_state.used_screen_texture = false; @@ -874,9 +914,9 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con } uint32_t lightmap_captures_used = 0; - Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); - near_plane.d += p_cam_projection.get_z_near(); - float z_max = p_cam_projection.get_z_far() - p_cam_projection.get_z_near(); + Plane near_plane(p_render_data->cam_transform.origin, -p_render_data->cam_transform.basis.get_axis(Vector3::AXIS_Z)); + near_plane.d += p_render_data->cam_projection.get_z_near(); + float z_max = p_render_data->cam_projection.get_z_far() - p_render_data->cam_projection.get_z_near(); RenderList *rl = &render_list[p_render_list]; _update_dirty_geometry_instances(); @@ -890,8 +930,8 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con //fill list - for (int i = 0; i < (int)p_instances.size(); i++) { - GeometryInstanceForwardClustered *inst = static_cast<GeometryInstanceForwardClustered *>(p_instances[i]); + for (int i = 0; i < (int)p_render_data->instances->size(); i++) { + GeometryInstanceForwardClustered *inst = static_cast<GeometryInstanceForwardClustered *>((*p_render_data->instances)[i]); Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); inst->depth = near_plane.distance_to(support_min); @@ -899,6 +939,9 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con uint32_t flags = inst->base_flags; //fill flags if appropriate + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } bool uses_lightmap = false; bool uses_gi = false; @@ -941,19 +984,19 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con uses_lightmap = true; } - } else if (!low_end) { + } else { if (p_using_opaque_gi) { flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS; } - if (inst->gi_probes[0].is_valid()) { + if (inst->voxel_gi_instances[0].is_valid()) { uint32_t probe0_index = 0xFFFF; uint32_t probe1_index = 0xFFFF; - for (uint32_t j = 0; j < scene_state.giprobes_used; j++) { - if (scene_state.giprobe_ids[j] == inst->gi_probes[0]) { + for (uint32_t j = 0; j < scene_state.voxelgis_used; j++) { + if (scene_state.voxelgi_ids[j] == inst->voxel_gi_instances[0]) { probe0_index = j; - } else if (scene_state.giprobe_ids[j] == inst->gi_probes[1]) { + } else if (scene_state.voxelgi_ids[j] == inst->voxel_gi_instances[1]) { probe1_index = j; } } @@ -964,7 +1007,7 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con } inst->gi_offset_cache = probe0_index | (probe1_index << 16); - flags |= INSTANCE_DATA_FLAG_USE_GIPROBE; + flags |= INSTANCE_DATA_FLAG_USE_VOXEL_GI; uses_gi = true; } else { if (p_using_sdfgi && inst->can_sdfgi) { @@ -985,13 +1028,13 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con // LOD - if (p_screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { + if (p_render_data->screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { //lod - Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_lod_plane.normal); - Vector3 lod_support_max = inst->transformed_aabb.get_support(p_lod_plane.normal); + Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_render_data->lod_camera_plane.normal); + Vector3 lod_support_max = inst->transformed_aabb.get_support(p_render_data->lod_camera_plane.normal); - float distance_min = p_lod_plane.distance_to(lod_support_min); - float distance_max = p_lod_plane.distance_to(lod_support_max); + float distance_min = p_render_data->lod_camera_plane.distance_to(lod_support_min); + float distance_max = p_render_data->lod_camera_plane.distance_to(lod_support_max); float distance = 0.0; @@ -1004,17 +1047,41 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con distance = -distance_max; } - surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_lod_distance_multiplier, p_screen_lod_threshold); + uint32_t indices; + surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, &indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } } else { surf->sort.lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } } // ADD Element if (p_pass_mode == PASS_MODE_COLOR) { - if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { rl->add_element(surf); } - if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { render_list[RENDER_LIST_ALPHA].add_element(surf); if (uses_gi) { surf->sort.uses_forward_gi = 1; @@ -1059,14 +1126,14 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con } } -void RenderForwardClustered::_setup_giprobes(const PagedArray<RID> &p_giprobes) { - scene_state.giprobes_used = MIN(p_giprobes.size(), uint32_t(MAX_GI_PROBES)); - for (uint32_t i = 0; i < scene_state.giprobes_used; i++) { - scene_state.giprobe_ids[i] = p_giprobes[i]; +void RenderForwardClustered::_setup_voxelgis(const PagedArray<RID> &p_voxelgis) { + scene_state.voxelgis_used = MIN(p_voxelgis.size(), uint32_t(MAX_VOXEL_GI_INSTANCESS)); + for (uint32_t i = 0; i < scene_state.voxelgis_used; i++) { + scene_state.voxelgi_ids[i] = p_voxelgis[i]; } } -void RenderForwardClustered::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform) { +void RenderForwardClustered::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform) { scene_state.lightmaps_used = 0; for (int i = 0; i < (int)p_lightmaps.size(); i++) { if (i >= (int)scene_state.max_lightmaps) { @@ -1088,28 +1155,24 @@ void RenderForwardClustered::_setup_lightmaps(const PagedArray<RID> &p_lightmaps } } -void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { +void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { + ERR_FAIL_COND_MSG(p_render_data->view_count != 1, "Multiview is currently not supported in the clustered renderer. Please use the mobile renderer for VR."); + RenderBufferDataForwardClustered *render_buffer = nullptr; - if (p_render_buffer.is_valid()) { - render_buffer = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffer); + if (p_render_data->render_buffers.is_valid()) { + render_buffer = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_data->render_buffers); } - RendererSceneEnvironmentRD *env = get_environment(p_environment); + RendererSceneEnvironmentRD *env = get_environment(p_render_data->environment); + static const int texture_multisamples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; //first of all, make a new render pass //fill up ubo RENDER_TIMESTAMP("Setup 3D Scene"); - float lod_distance_multiplier = p_cam_projection.get_lod_multiplier(); - Plane lod_camera_plane(p_cam_transform.get_origin(), -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); - - if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { - p_screen_lod_threshold = 0.0; - } - //scene_state.ubo.subsurface_scatter_width = subsurface_scatter_size; - Vector2 vp_he = p_cam_projection.get_viewport_half_extents(); + Vector2 vp_he = p_render_data->cam_projection.get_viewport_half_extents(); scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; scene_state.ubo.directional_light_count = 0; @@ -1125,7 +1188,8 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform bool using_separate_specular = false; bool using_ssr = false; bool using_sdfgi = false; - bool using_giprobe = false; + bool using_voxelgi = false; + bool reverse_cull = false; if (render_buffer) { screen_size.x = render_buffer->width; @@ -1133,29 +1197,29 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform opaque_framebuffer = render_buffer->color_fb; - if (!low_end && p_gi_probes.size() > 0) { - using_giprobe = true; + if (p_render_data->voxel_gi_instances->size() > 0) { + using_voxelgi = true; } - if (!p_environment.is_valid() && using_giprobe) { - depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE; + if (!p_render_data->environment.is_valid() && using_voxelgi) { + depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI; - } else if (p_environment.is_valid() && (environment_is_ssr_enabled(p_environment) || environment_is_sdfgi_enabled(p_environment) || using_giprobe)) { - if (environment_is_sdfgi_enabled(p_environment)) { - depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also giprobe + } else if (p_render_data->environment.is_valid() && (environment_is_ssr_enabled(p_render_data->environment) || environment_is_sdfgi_enabled(p_render_data->environment) || using_voxelgi)) { + if (environment_is_sdfgi_enabled(p_render_data->environment)) { + depth_pass_mode = using_voxelgi ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also voxelgi using_sdfgi = true; } else { - depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; + depth_pass_mode = using_voxelgi ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; } - if (environment_is_ssr_enabled(p_environment)) { + if (environment_is_ssr_enabled(p_render_data->environment)) { render_buffer->ensure_specular(); using_separate_specular = true; using_ssr = true; opaque_specular_framebuffer = render_buffer->color_specular_fb; } - } else if (p_environment.is_valid() && (environment_is_ssao_enabled(p_environment) || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER)) { + } else if (p_render_data->environment.is_valid() && (environment_is_ssao_enabled(p_render_data->environment) || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER)) { depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS; } @@ -1168,10 +1232,10 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform depth_framebuffer = render_buffer->depth_normal_roughness_fb; depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0)); } break; - case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: { + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { _allocate_normal_roughness_texture(render_buffer); - render_buffer->ensure_giprobe(); - depth_framebuffer = render_buffer->depth_normal_roughness_giprobe_fb; + render_buffer->ensure_voxelgi(); + depth_framebuffer = render_buffer->depth_normal_roughness_voxelgi_fb; depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0)); depth_pass_clear.push_back(Color(0, 0, 0, 0)); } break; @@ -1180,39 +1244,42 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } alpha_framebuffer = opaque_framebuffer; - } else if (p_reflection_probe.is_valid()) { - uint32_t resolution = reflection_probe_instance_get_resolution(p_reflection_probe); + } else if (p_render_data->reflection_probe.is_valid()) { + uint32_t resolution = reflection_probe_instance_get_resolution(p_render_data->reflection_probe); screen_size.x = resolution; screen_size.y = resolution; - opaque_framebuffer = reflection_probe_instance_get_framebuffer(p_reflection_probe, p_reflection_probe_pass); - depth_framebuffer = reflection_probe_instance_get_depth_framebuffer(p_reflection_probe, p_reflection_probe_pass); + opaque_framebuffer = reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); + depth_framebuffer = reflection_probe_instance_get_depth_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); alpha_framebuffer = opaque_framebuffer; - if (storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_reflection_probe))) { - p_environment = RID(); //no environment on interiors + if (storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + p_render_data->environment = RID(); //no environment on interiors + env = nullptr; } + + reverse_cull = true; // for some reason our views are inverted } else { ERR_FAIL(); //bug? } RD::get_singleton()->draw_command_begin_label("Render Setup"); - _setup_lightmaps(p_lightmaps, p_cam_transform); - _setup_giprobes(p_gi_probes); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + _setup_lightmaps(*p_render_data->lightmaps, p_render_data->cam_transform); + _setup_voxelgis(*p_render_data->voxel_gi_instances); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) - _fill_render_list(RENDER_LIST_OPAQUE, p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe, lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR, using_sdfgi, using_sdfgi || using_voxelgi); render_list[RENDER_LIST_OPAQUE].sort_by_key(); - render_list[RENDER_LIST_ALPHA].sort_by_depth(); - _fill_instance_data(RENDER_LIST_OPAQUE); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_instance_data(RENDER_LIST_OPAQUE, p_render_data->render_info ? p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE] : (int *)nullptr); _fill_instance_data(RENDER_LIST_ALPHA); RD::get_singleton()->draw_command_end_label(); - bool using_sss = !low_end && render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; + bool using_sss = render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; if (using_sss) { using_separate_specular = true; @@ -1229,26 +1296,26 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW) { clear_color = Color(0, 0, 0, 1); //in overdraw mode, BG should always be black - } else if (is_environment(p_environment)) { - RS::EnvironmentBG bg_mode = environment_get_background(p_environment); - float bg_energy = environment_get_bg_energy(p_environment); + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG bg_mode = environment_get_background(p_render_data->environment); + float bg_energy = environment_get_bg_energy(p_render_data->environment); switch (bg_mode) { case RS::ENV_BG_CLEAR_COLOR: { clear_color = p_default_bg_color; clear_color.r *= bg_energy; clear_color.g *= bg_energy; clear_color.b *= bg_energy; - if (render_buffers_has_volumetric_fog(p_render_buffer) || environment_is_fog_enabled(p_environment)) { + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers) || environment_is_fog_enabled(p_render_data->environment)) { draw_sky_fog_only = true; storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); } } break; case RS::ENV_BG_COLOR: { - clear_color = environment_get_bg_color(p_environment); + clear_color = environment_get_bg_color(p_render_data->environment); clear_color.r *= bg_energy; clear_color.g *= bg_energy; clear_color.b *= bg_energy; - if (render_buffers_has_volumetric_fog(p_render_buffer) || environment_is_fog_enabled(p_environment)) { + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers) || environment_is_fog_enabled(p_render_data->environment)) { draw_sky_fog_only = true; storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); } @@ -1268,21 +1335,21 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } } // setup sky if used for ambient, reflections, or background - if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_environment) == RS::ENV_AMBIENT_SOURCE_SKY) { + if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_render_data->environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_render_data->environment) == RS::ENV_AMBIENT_SOURCE_SKY) { RENDER_TIMESTAMP("Setup Sky"); RD::get_singleton()->draw_command_begin_label("Setup Sky"); - CameraMatrix projection = p_cam_projection; - if (p_reflection_probe.is_valid()) { + CameraMatrix projection = p_render_data->cam_projection; + if (p_render_data->reflection_probe.is_valid()) { CameraMatrix correction; correction.set_depth_correction(true); - projection = correction * p_cam_projection; + projection = correction * p_render_data->cam_projection; } - sky.setup(env, p_render_buffer, projection, p_cam_transform, screen_size, this); + sky.setup(env, p_render_data->render_buffers, projection, p_render_data->cam_transform, screen_size, this); RID sky_rid = env->sky; if (sky_rid.is_valid()) { - sky.update(env, projection, p_cam_transform, time); + sky.update(env, projection, p_render_data->cam_transform, time); radiance_texture = sky.sky_get_radiance_texture_rd(sky_rid); } else { // do not try to draw sky if invalid @@ -1294,15 +1361,15 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform clear_color = p_default_bg_color; } - bool debug_giprobes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION; + bool debug_voxelgis = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_EMISSION; bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; - bool depth_pre_pass = !low_end && depth_framebuffer.is_valid(); + bool depth_pre_pass = depth_framebuffer.is_valid(); - bool using_ssao = depth_pre_pass && p_render_buffer.is_valid() && p_environment.is_valid() && environment_is_ssao_enabled(p_environment); + bool using_ssao = depth_pre_pass && p_render_data->render_buffers.is_valid() && p_render_data->environment.is_valid() && environment_is_ssao_enabled(p_render_data->environment); bool continue_depth = false; if (depth_pre_pass) { //depth pre pass - bool needs_pre_resolve = _needs_post_prepass_render(using_sdfgi || using_giprobe); + bool needs_pre_resolve = _needs_post_prepass_render(p_render_data, using_sdfgi || using_voxelgi); if (needs_pre_resolve) { RENDER_TIMESTAMP("GI + Render Depth Pre-Pass (parallel)"); } else { @@ -1313,34 +1380,33 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform RD::get_singleton()->draw_list_begin(depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, depth_pass_clear); RD::get_singleton()->draw_list_end(); //start compute processes here, so they run at the same time as depth pre-pass - _post_prepass_render(using_sdfgi || using_giprobe); + _post_prepass_render(p_render_data, using_sdfgi || using_voxelgi); } RD::get_singleton()->draw_command_begin_label("Render Depth Pre-Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, nullptr, RID()); - bool finish_depth = using_ssao || using_sdfgi || using_giprobe; - RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + bool finish_depth = using_ssao || using_sdfgi || using_voxelgi; + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, depth_pass_mode, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, depth_framebuffer, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, needs_pre_resolve ? Vector<Color>() : depth_pass_clear); RD::get_singleton()->draw_command_end_label(); if (needs_pre_resolve) { - _pre_resolve_render(using_sdfgi || using_giprobe); + _pre_resolve_render(p_render_data, using_sdfgi || using_voxelgi); } if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { RENDER_TIMESTAMP("Resolve Depth Pre-Pass"); RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass"); - if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) { + if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI) { if (needs_pre_resolve) { RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); } - static int texture_samples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; - storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_giprobe ? render_buffer->giprobe_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_giprobe ? render_buffer->giprobe_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_samples[render_buffer->msaa]); + storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_voxelgi ? render_buffer->voxelgi_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_voxelgi ? render_buffer->voxelgi_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } else if (finish_depth) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); + storage->get_effects()->resolve_depth(render_buffer->depth_msaa, render_buffer->depth, Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } RD::get_singleton()->draw_command_end_label(); } @@ -1348,24 +1414,24 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform continue_depth = !finish_depth; } - _pre_opaque_render(using_ssao, using_sdfgi || using_giprobe, render_buffer ? render_buffer->normal_roughness_buffer : RID(), render_buffer ? render_buffer->giprobe_buffer : RID()); + _pre_opaque_render(p_render_data, using_ssao, using_sdfgi || using_voxelgi, render_buffer ? render_buffer->normal_roughness_buffer : RID(), render_buffer ? render_buffer->voxelgi_buffer : RID()); RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); - scene_state.ubo.directional_light_count = _get_render_state_directional_light_count(); + scene_state.ubo.directional_light_count = p_render_data->directional_light_count; - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); RENDER_TIMESTAMP("Render Opaque Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss; bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss; { - bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only || debug_giprobes || debug_sdfgi_probes); - bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only || debug_giprobes || debug_sdfgi_probes); + bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only || debug_voxelgis || debug_sdfgi_probes); + bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only || debug_voxelgis || debug_sdfgi_probes); //regular forward for now Vector<Color> c; @@ -1379,7 +1445,7 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer; - RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); if (will_continue_color && using_separate_specular) { // close the specular framebuffer, as it's no longer used @@ -1390,34 +1456,34 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform RD::get_singleton()->draw_command_end_label(); - if (debug_giprobes) { - //debug giprobes + if (debug_voxelgis) { + //debug voxelgis bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only); CameraMatrix dc; dc.set_depth_correction(true); - CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); + CameraMatrix cm = (dc * p_render_data->cam_projection) * CameraMatrix(p_render_data->cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); - RD::get_singleton()->draw_command_begin_label("Debug GIProbes"); - for (int i = 0; i < (int)p_gi_probes.size(); i++) { - gi.debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0); + RD::get_singleton()->draw_command_begin_label("Debug VoxelGIs"); + for (int i = 0; i < (int)p_render_data->voxel_gi_instances->size(); i++) { + gi.debug_voxel_gi((*p_render_data->voxel_gi_instances)[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_EMISSION, 1.0); } RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } if (debug_sdfgi_probes) { - //debug giprobes + //debug voxelgis bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only); CameraMatrix dc; dc.set_depth_correction(true); - CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); + CameraMatrix cm = (dc * p_render_data->cam_projection) * CameraMatrix(p_render_data->cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); RD::get_singleton()->draw_command_begin_label("Debug SDFGI"); - _debug_sdfgi_probes(p_render_buffer, draw_list, opaque_framebuffer, cm); + _debug_sdfgi_probes(p_render_data->render_buffers, draw_list, opaque_framebuffer, cm); RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1425,14 +1491,14 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform if (draw_sky || draw_sky_fog_only) { RENDER_TIMESTAMP("Render Sky"); - CameraMatrix projection = p_cam_projection; - if (p_reflection_probe.is_valid()) { + CameraMatrix projection = p_render_data->cam_projection; + if (p_render_data->reflection_probe.is_valid()) { CameraMatrix correction; correction.set_depth_correction(true); - projection = correction * p_cam_projection; + projection = correction * p_render_data->cam_projection; } RD::get_singleton()->draw_command_begin_label("Draw Sky"); - sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, projection, p_cam_transform, time); + sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, 1, &projection, p_render_data->cam_transform, time); RD::get_singleton()->draw_command_end_label(); } @@ -1444,21 +1510,21 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); + storage->get_effects()->resolve_depth(render_buffer->depth_msaa, render_buffer->depth, Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } if (using_separate_specular) { if (using_sss) { RENDER_TIMESTAMP("Sub Surface Scattering"); RD::get_singleton()->draw_command_begin_label("Process Sub Surface Scattering"); - _process_sss(p_render_buffer, p_cam_projection); + _process_sss(p_render_data->render_buffers, p_render_data->cam_projection); RD::get_singleton()->draw_command_end_label(); } if (using_ssr) { RENDER_TIMESTAMP("Screen Space Reflection"); RD::get_singleton()->draw_command_begin_label("Process Screen Space Reflections"); - _process_ssr(p_render_buffer, render_buffer->color_fb, render_buffer->normal_roughness_buffer, render_buffer->specular, render_buffer->specular, Color(0, 0, 0, 1), p_environment, p_cam_projection, render_buffer->msaa == RS::VIEWPORT_MSAA_DISABLED); + _process_ssr(p_render_data->render_buffers, render_buffer->color_fb, render_buffer->normal_roughness_buffer, render_buffer->specular, render_buffer->specular, Color(0, 0, 0, 1), p_render_data->environment, p_render_data->cam_projection, render_buffer->msaa == RS::VIEWPORT_MSAA_DISABLED); RD::get_singleton()->draw_command_end_label(); } else { //just mix specular back @@ -1467,16 +1533,26 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } } + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } + + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } + RENDER_TIMESTAMP("Render Transparent Pass"); RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); - rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); { - RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); } @@ -1489,6 +1565,14 @@ void RenderForwardClustered::_render_scene(RID p_render_buffer, const Transform } RD::get_singleton()->draw_command_end_label(); + + if (p_render_data->render_buffers.is_valid()) { + _debug_draw_cluster(p_render_data->render_buffers); + + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(p_render_data); + } } void RenderForwardClustered::_render_shadow_begin() { @@ -1499,26 +1583,41 @@ void RenderForwardClustered::_render_shadow_begin() { render_list[RENDER_LIST_SECONDARY].clear(); scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); } -void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { + +void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RendererScene::RenderInfo *p_render_info) { uint32_t shadow_pass_index = scene_state.shadow_passes.size(); SceneState::ShadowPass shadow_pass; + RenderDataRD render_data; + render_data.cam_projection = p_projection; + render_data.cam_transform = p_transform; + render_data.z_far = p_zfar; + render_data.z_near = 0.0; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + render_data.lod_camera_plane = p_camera_plane; + render_data.lod_distance_multiplier = p_lod_distance_multiplier; + render_data.render_info = p_render_info; + scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; - _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), !p_flip_y, Color(), 0, p_zfar, false, p_use_pancake, shadow_pass_index); + _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { - p_screen_lod_threshold = 0.0; + render_data.screen_lod_threshold = 0.0; + } else { + render_data.screen_lod_threshold = p_screen_lod_threshold; } PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); - _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_projection, p_transform, false, false, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, true); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, false, false, true); uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); - _fill_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); + _fill_instance_data(RENDER_LIST_SECONDARY, p_render_info ? p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW] : (int *)nullptr, render_list_from, render_list_size, false); { //regular forward for now @@ -1534,8 +1633,8 @@ void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const Page shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete shadow_pass.camera_plane = p_camera_plane; - shadow_pass.screen_lod_threshold = p_screen_lod_threshold; - shadow_pass.lod_distance_multiplier = p_lod_distance_multiplier; + shadow_pass.screen_lod_threshold = render_data.screen_lod_threshold; + shadow_pass.lod_distance_multiplier = render_data.lod_distance_multiplier; shadow_pass.framebuffer = p_framebuffer; shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); @@ -1553,7 +1652,7 @@ void RenderForwardClustered::_render_shadow_process() { for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { //render passes need to be configured after instance buffer is done, since they need the latest version SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; - shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>(), false, i); + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID(), false, i); } RD::get_singleton()->draw_command_end_label(); @@ -1563,7 +1662,7 @@ void RenderForwardClustered::_render_shadow_end(uint32_t p_barrier) { for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; - RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, false, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); } @@ -1573,57 +1672,73 @@ void RenderForwardClustered::_render_shadow_end(uint32_t p_barrier) { RD::get_singleton()->draw_command_end_label(); } -void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { +void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { RENDER_TIMESTAMP("Setup Render Collider Heightfield"); RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + RenderDataRD render_data; + render_data.cam_projection = p_cam_projection; + render_data.cam_transform = p_cam_transform; + render_data.z_near = 0.0; + render_data.z_far = p_cam_projection.get_z_far(); + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); + _setup_environment(&render_data, true, Vector2(1, 1), true, Color(), false, false); PassMode pass_mode = PASS_MODE_SHADOW; - _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); RENDER_TIMESTAMP("Render Collider Heightfield"); { //regular forward for now - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, false, rp_uniform_set); _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); } RD::get_singleton()->draw_command_end_label(); } -void RenderForwardClustered::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RenderForwardClustered::_render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering Material"); RD::get_singleton()->draw_command_begin_label("Render Material"); + RenderDataRD render_data; + render_data.cam_projection = p_cam_projection; + render_data.cam_transform = p_cam_transform; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = false; - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1644,24 +1759,29 @@ void RenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p RD::get_singleton()->draw_command_begin_label("Render UV2"); + RenderDataRD render_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = true; - _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, true); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set, true); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1693,6 +1813,7 @@ void RenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //first wireframe, for pseudo conservative } render_list_params.uv_offset = Vector2(); + render_list_params.force_wireframe = false; _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //second regular triangles RD::get_singleton()->draw_list_end(); @@ -1706,13 +1827,18 @@ void RenderForwardClustered::_render_sdfgi(RID p_render_buffers, const Vector3i RD::get_singleton()->draw_command_begin_label("Render SDFGI Voxel"); + RenderDataRD render_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + _update_render_base_uniform_set(); RenderBufferDataForwardClustered *render_buffer = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffers); ERR_FAIL_COND(!render_buffer); PassMode pass_mode = PASS_MODE_SDF; - _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); _fill_instance_data(RENDER_LIST_SECONDARY); @@ -1744,28 +1870,26 @@ void RenderForwardClustered::_render_sdfgi(RID p_render_buffers, const Vector3i fb_size.x = p_size[right_axis]; fb_size.y = p_size[up_axis]; - Transform cam_xform; - cam_xform.origin = center + axis * half_extents; - cam_xform.basis.set_axis(0, right); - cam_xform.basis.set_axis(1, up); - cam_xform.basis.set_axis(2, axis); + render_data.cam_transform.origin = center + axis * half_extents; + render_data.cam_transform.basis.set_axis(0, right); + render_data.cam_transform.basis.set_axis(1, up); + render_data.cam_transform.basis.set_axis(2, axis); - //print_line("pass: " + itos(i) + " xform " + cam_xform); + //print_line("pass: " + itos(i) + " xform " + render_data.cam_transform); float h_size = half_extents[right_axis]; float v_size = half_extents[up_axis]; float d_size = half_extents[i] * 2.0; - CameraMatrix camera_proj; - camera_proj.set_orthogonal(-h_size, h_size, -v_size, v_size, 0, d_size); + render_data.cam_projection.set_orthogonal(-h_size, h_size, -v_size, v_size, 0, d_size); //print_line("pass: " + itos(i) + " cam hsize: " + rtos(h_size) + " vsize: " + rtos(v_size) + " dsize " + rtos(d_size)); - Transform to_bounds; + Transform3D to_bounds; to_bounds.origin = p_bounds.position; to_bounds.basis.scale(p_bounds.size); - RendererStorageRD::store_transform(to_bounds.affine_inverse() * cam_xform, scene_state.ubo.sdf_to_bounds); + RendererStorageRD::store_transform(to_bounds.affine_inverse() * render_data.cam_transform, scene_state.ubo.sdf_to_bounds); - _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); @@ -1775,7 +1899,7 @@ void RenderForwardClustered::_render_sdfgi(RID p_render_buffers, const Vector3i E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); } - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, false); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set, false); _render_list_with_threads(&render_list_params, E->get(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); } @@ -1831,13 +1955,67 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 4; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); @@ -1845,35 +2023,35 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 5; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 6; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 7; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 8; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 9; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -1881,7 +2059,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 10; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -1889,7 +2067,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 11; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -1898,15 +2076,15 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 12; + u.binding = 14; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } - if (!low_end) { + { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 13; + u.binding = 15; u.ids.push_back(sdfgi_get_ubo()); uniforms.push_back(u); } @@ -1915,13 +2093,13 @@ void RenderForwardClustered::_update_render_base_uniform_set() { } } -RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas, int p_index) { +RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas, int p_index) { //there should always be enough uniform buffers for render passes, otherwise bugs ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); RenderBufferDataForwardClustered *rb = nullptr; - if (p_render_buffers.is_valid()) { - rb = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffers); + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_data->render_buffers); } //default render buffer and scene state uniform set @@ -1961,7 +2139,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend } { - RID ref_texture = p_reflection_atlas.is_valid() ? reflection_atlas_get_texture(p_reflection_atlas) : RID(); + RID ref_texture = (p_render_data && p_render_data->reflection_atlas.is_valid()) ? reflection_atlas_get_texture(p_render_data->reflection_atlas) : RID(); RD::Uniform u; u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; @@ -1978,8 +2156,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture; - if (p_shadow_atlas.is_valid()) { - texture = shadow_atlas_get_texture(p_shadow_atlas); + if (p_render_data && p_render_data->shadow_atlas.is_valid()) { + texture = shadow_atlas_get_texture(p_render_data->shadow_atlas); } if (!texture.is_valid()) { texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); @@ -2005,8 +2183,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend u.ids.resize(scene_state.max_lightmaps); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { - if (i < p_lightmaps.size()) { - RID base = lightmap_instance_get_lightmap(p_lightmaps[i]); + if (p_render_data && i < p_render_data->lightmaps->size()) { + RID base = lightmap_instance_get_lightmap((*p_render_data->lightmaps)[i]); RID texture = storage->lightmap_get_texture(base); RID rd_texture = storage->texture_get_rd_texture(texture); u.ids.write[i] = rd_texture; @@ -2021,11 +2199,11 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.resize(MAX_GI_PROBES); + u.ids.resize(MAX_VOXEL_GI_INSTANCESS); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); - for (int i = 0; i < MAX_GI_PROBES; i++) { - if (i < (int)p_gi_probes.size()) { - RID tex = gi.gi_probe_instance_get_texture(p_gi_probes[i]); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { + if (p_render_data && i < (int)p_render_data->voxel_gi_instances->size()) { + RID tex = gi.voxel_gi_instance_get_texture((*p_render_data->voxel_gi_instances)[i]); if (!tex.is_valid()) { tex = default_tex; } @@ -2042,7 +2220,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : scene_shader.default_vec4_xform_buffer; + RID cb = (p_render_data && p_render_data->cluster_buffer.is_valid()) ? p_render_data->cluster_buffer : scene_shader.default_vec4_xform_buffer; u.ids.push_back(cb); uniforms.push_back(u); } @@ -2051,7 +2229,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + RID dbt = rb ? render_buffers_get_back_depth_texture(p_render_data->render_buffers) : RID(); + RID texture = (dbt.is_valid()) ? dbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); uniforms.push_back(u); } @@ -2059,13 +2238,13 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_buffers) : RID(); + RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_data->render_buffers) : RID(); RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } - if (!low_end) { + { { RD::Uniform u; u.binding = 11; @@ -2079,7 +2258,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID aot = rb ? render_buffers_get_ao_texture(p_render_buffers) : RID(); + RID aot = rb ? render_buffers_get_ao_texture(p_render_data->render_buffers) : RID(); RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); @@ -2089,7 +2268,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID ambient_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_ambient_texture(p_render_buffers) : RID(); + RID ambient_buffer = rb ? render_buffers_get_gi_ambient_texture(p_render_data->render_buffers) : RID(); RID texture = ambient_buffer.is_valid() ? ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); @@ -2099,7 +2278,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 14; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID reflection_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_reflection_texture(p_render_buffers) : RID(); + RID reflection_buffer = rb ? render_buffers_get_gi_reflection_texture(p_render_data->render_buffers) : RID(); RID texture = reflection_buffer.is_valid() ? reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); @@ -2109,8 +2288,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend u.binding = 15; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID t; - if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { - t = render_buffers_get_sdfgi_irradiance_probes(p_render_buffers); + if (rb && render_buffers_is_sdfgi_enabled(p_render_data->render_buffers)) { + t = render_buffers_get_sdfgi_irradiance_probes(p_render_data->render_buffers); } else { t = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); } @@ -2121,8 +2300,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 16; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { - u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers)); + if (rb && render_buffers_is_sdfgi_enabled(p_render_data->render_buffers)) { + u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_data->render_buffers)); } else { u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); } @@ -2132,7 +2311,7 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 17; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(rb ? render_buffers_get_gi_probe_buffer(p_render_buffers) : render_buffers_get_default_gi_probe_buffer()); + u.ids.push_back(rb ? render_buffers_get_voxel_gi_buffer(p_render_data->render_buffers) : render_buffers_get_default_voxel_gi_buffer()); uniforms.push_back(u); } { @@ -2140,8 +2319,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend u.binding = 18; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID vfog = RID(); - if (rb && render_buffers_has_volumetric_fog(p_render_buffers)) { - vfog = render_buffers_get_volumetric_fog_texture(p_render_buffers); + if (rb && render_buffers_has_volumetric_fog(p_render_data->render_buffers)) { + vfog = render_buffers_get_volumetric_fog_texture(p_render_data->render_buffers); if (vfog.is_null()) { vfog = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); } @@ -2241,13 +2420,13 @@ RID RenderForwardClustered::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_te } { - // No GIProbes + // No VoxelGIs RD::Uniform u; u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.resize(MAX_GI_PROBES); + u.ids.resize(MAX_VOXEL_GI_INSTANCESS); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); - for (int i = 0; i < MAX_GI_PROBES; i++) { + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { u.ids.write[i] = default_tex; } @@ -2298,15 +2477,6 @@ RID RenderForwardClustered::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_te return sdfgi_pass_uniform_set; } -void RenderForwardClustered::_render_buffers_clear_uniform_set(RenderBufferDataForwardClustered *rb) { -} - -void RenderForwardClustered::_render_buffers_uniform_set_changed(RID p_render_buffers) { - RenderBufferDataForwardClustered *rb = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffers); - - _render_buffers_clear_uniform_set(rb); -} - RID RenderForwardClustered::_render_buffers_get_normal_texture(RID p_render_buffers) { RenderBufferDataForwardClustered *rb = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_buffers); @@ -2376,9 +2546,13 @@ void RenderForwardClustered::_geometry_instance_add_surface_with_material(Geomet flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; } + if (p_material->shader_data->uses_particle_trails) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS; + } + SceneShaderForwardClustered::MaterialData *material_shadow = nullptr; void *surface_shadow = nullptr; - if (!p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { + if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; material_shadow = (SceneShaderForwardClustered::MaterialData *)storage->material_get_data(scene_shader.default_material, RendererStorageRD::SHADER_TYPE_3D); @@ -2423,12 +2597,14 @@ void RenderForwardClustered::_geometry_instance_add_surface_with_material(Geomet sdcache->sort.sort_key2 = 0; sdcache->sort.surface_index = p_surface; - sdcache->sort.material_id_low = p_material_id & 0x3FFF; - sdcache->sort.material_id_hi = p_material_id >> 14; + sdcache->sort.material_id_low = p_material_id & 0xFFFF; + sdcache->sort.material_id_hi = p_material_id >> 16; sdcache->sort.shader_id = p_shader_id; sdcache->sort.geometry_id = p_mesh.get_local_index(); //only meshes can repeat anyway sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; sdcache->sort.priority = p_material->priority; + sdcache->sort.uses_projector = ginstance->using_projectors; + sdcache->sort.uses_softshadow = ginstance->using_softshadows; } void RenderForwardClustered::_geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { @@ -2547,7 +2723,7 @@ void RenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geome } } - ginstance->instance_count = storage->particles_get_amount(ginstance->data->base); + ginstance->instance_count = storage->particles_get_amount(ginstance->data->base, ginstance->trail_steps); } break; @@ -2557,46 +2733,32 @@ void RenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geome //Fill push constant + ginstance->base_flags = 0; + bool store_transform = true; if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; - uint32_t stride; if (storage->multimesh_get_transform_format(ginstance->data->base) == RS::MULTIMESH_TRANSFORM_2D) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; - stride = 2; - } else { - stride = 3; } if (storage->multimesh_uses_colors(ginstance->data->base)) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; - stride += 1; } if (storage->multimesh_uses_custom_data(ginstance->data->base)) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; - stride += 1; } - ginstance->base_flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); ginstance->transforms_uniform_set = storage->multimesh_get_3d_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); } else if (ginstance->data->base_type == RS::INSTANCE_PARTICLES) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; - uint32_t stride; - if (false) { // 2D particles - ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; - stride = 2; - } else { - stride = 3; - } ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; - stride += 1; - ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; - stride += 1; - ginstance->base_flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); + //for particles, stride is the trail size + ginstance->base_flags |= (ginstance->trail_steps << INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT); if (!storage->particles_is_using_local_coords(ginstance->data->base)) { store_transform = false; @@ -2605,7 +2767,6 @@ void RenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geome } else if (ginstance->data->base_type == RS::INSTANCE_MESH) { if (storage->skeleton_is_valid(ginstance->data->skeleton)) { - ginstance->base_flags |= INSTANCE_DATA_FLAG_SKELETON; ginstance->transforms_uniform_set = storage->skeleton_get_3d_uniform_set(ginstance->data->skeleton, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); if (ginstance->data->dirty_dependencies) { storage->skeleton_update_dependency(ginstance->data->skeleton, &ginstance->data->dependency_tracker); @@ -2616,8 +2777,8 @@ void RenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geome ginstance->store_transform_cache = store_transform; ginstance->can_sdfgi = false; - if (!lightmap_instance_is_valid(ginstance->lightmap_instance) && !low_end) { - if (ginstance->gi_probes[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { + if (!lightmap_instance_is_valid(ginstance->lightmap_instance)) { + if (ginstance->voxel_gi_instances[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { ginstance->can_sdfgi = true; } } @@ -2640,6 +2801,7 @@ void RenderForwardClustered::_geometry_instance_dependency_changed(RendererStora switch (p_notification) { case RendererStorage::DEPENDENCY_CHANGED_MATERIAL: case RendererStorage::DEPENDENCY_CHANGED_MESH: + case RendererStorage::DEPENDENCY_CHANGED_PARTICLES: case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH: case RendererStorage::DEPENDENCY_CHANGED_SKELETON_DATA: { static_cast<RenderForwardClustered *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); @@ -2703,7 +2865,7 @@ void RenderForwardClustered::geometry_instance_set_mesh_instance(GeometryInstanc ginstance->mesh_instance = p_mesh_instance; _geometry_instance_mark_dirty(ginstance); } -void RenderForwardClustered::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { +void RenderForwardClustered::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->transform = p_transform; @@ -2716,6 +2878,7 @@ void RenderForwardClustered::geometry_instance_set_transform(GeometryInstance *p float max_scale = MAX(model_scale_vec.x, MAX(model_scale_vec.y, model_scale_vec.z)); float min_scale = MIN(model_scale_vec.x, MIN(model_scale_vec.y, model_scale_vec.z)); + ginstance->non_uniform_scale = max_scale >= 0.0 && (min_scale / max_scale) < 0.9; ginstance->lod_model_scale = max_scale; @@ -2753,7 +2916,7 @@ void RenderForwardClustered::geometry_instance_set_lightmap_capture(GeometryInst ginstance->lightmap_sh = geometry_instance_lightmap_sh.alloc(); } - copymem(ginstance->lightmap_sh->sh, p_sh9, sizeof(Color) * 9); + memcpy(ginstance->lightmap_sh->sh, p_sh9, sizeof(Color) * 9); } else { if (ginstance->lightmap_sh != nullptr) { geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); @@ -2799,7 +2962,7 @@ void RenderForwardClustered::geometry_instance_free(GeometryInstance *p_geometry } uint32_t RenderForwardClustered::geometry_instance_get_pair_mask() { - return (1 << RS::INSTANCE_GI_PROBE); + return (1 << RS::INSTANCE_VOXEL_GI); } void RenderForwardClustered::geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) { } @@ -2808,33 +2971,84 @@ void RenderForwardClustered::geometry_instance_pair_reflection_probe_instances(G void RenderForwardClustered::geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) { } -Transform RenderForwardClustered::geometry_instance_get_transform(GeometryInstance *p_instance) { +Transform3D RenderForwardClustered::geometry_instance_get_transform(GeometryInstance *p_instance) { GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_instance); - ERR_FAIL_COND_V(!ginstance, Transform()); + ERR_FAIL_COND_V(!ginstance, Transform3D()); return ginstance->transform; } + AABB RenderForwardClustered::geometry_instance_get_aabb(GeometryInstance *p_instance) { GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_instance); ERR_FAIL_COND_V(!ginstance, AABB()); return ginstance->data->aabb; } -void RenderForwardClustered::geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count) { +void RenderForwardClustered::geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) { GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - if (p_gi_probe_instance_count > 0) { - ginstance->gi_probes[0] = p_gi_probe_instances[0]; + if (p_voxel_gi_instance_count > 0) { + ginstance->voxel_gi_instances[0] = p_voxel_gi_instances[0]; } else { - ginstance->gi_probes[0] = RID(); + ginstance->voxel_gi_instances[0] = RID(); } - if (p_gi_probe_instance_count > 1) { - ginstance->gi_probes[1] = p_gi_probe_instances[1]; + if (p_voxel_gi_instance_count > 1) { + ginstance->voxel_gi_instances[1] = p_voxel_gi_instances[1]; } else { - ginstance->gi_probes[1] = RID(); + ginstance->voxel_gi_instances[1] = RID(); } } +void RenderForwardClustered::geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->using_projectors = p_projector; + ginstance->using_softshadows = p_softshadow; + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardClustered::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_FILTER; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_FILTER; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + _base_uniforms_changed(); //also need this +} + RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : RendererSceneRenderRD(p_storage) { singleton = this; @@ -2843,20 +3057,16 @@ RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : { String defines; - if (low_end) { - defines += "\n#define LOW_END_MODE \n"; - } - defines += "\n#define MAX_ROUGHNESS_LOD " + itos(get_roughness_layers() - 1) + ".0\n"; if (is_using_radiance_cubemap_array()) { defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; } defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; - defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(get_max_directional_lights()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; { //lightmaps - scene_state.max_lightmaps = low_end ? 2 : MAX_LIGHTMAPS; + scene_state.max_lightmaps = MAX_LIGHTMAPS; defines += "\n#define MAX_LIGHTMAP_TEXTURES " + itos(scene_state.max_lightmaps) + "\n"; defines += "\n#define MAX_LIGHTMAPS " + itos(scene_state.max_lightmaps) + "\n"; @@ -2872,10 +3082,12 @@ RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : defines += "\n#define MATERIAL_UNIFORM_SET " + itos(MATERIAL_UNIFORM_SET) + "\n"; } - scene_shader.init(p_storage, defines, low_end); + scene_shader.init(p_storage, defines); } render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); } RenderForwardClustered::~RenderForwardClustered() { diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h index 72e84a6f24..676f633d33 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -51,10 +51,19 @@ class RenderForwardClustered : public RendererSceneRenderRD { }; enum { + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 6, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 7, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 8, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 9, + SPEC_CONSTANT_DECAL_FILTER = 10, + SPEC_CONSTANT_PROJECTOR_FILTER = 11, + }; + + enum { SDFGI_MAX_CASCADES = 8, - MAX_GI_PROBES = 8, + MAX_VOXEL_GI_INSTANCESS = 8, MAX_LIGHTMAPS = 8, - MAX_GI_PROBES_PER_INSTANCE = 2, + MAX_VOXEL_GI_INSTANCESS_PER_INSTANCE = 2, INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 }; @@ -79,7 +88,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { RID depth; RID specular; RID normal_roughness_buffer; - RID giprobe_buffer; + RID voxelgi_buffer; RS::ViewportMSAA msaa; RD::TextureSamples texture_samples; @@ -89,11 +98,11 @@ class RenderForwardClustered : public RendererSceneRenderRD { RID specular_msaa; RID normal_roughness_buffer_msaa; RID roughness_buffer_msaa; - RID giprobe_buffer_msaa; + RID voxelgi_buffer_msaa; RID depth_fb; RID depth_normal_roughness_fb; - RID depth_normal_roughness_giprobe_fb; + RID depth_normal_roughness_voxelgi_fb; RID color_fb; RID color_specular_fb; RID specular_only_fb; @@ -101,14 +110,14 @@ class RenderForwardClustered : public RendererSceneRenderRD { RID render_sdfgi_uniform_set; void ensure_specular(); - void ensure_giprobe(); + void ensure_voxelgi(); void clear(); - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa); + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); ~RenderBufferDataForwardClustered(); }; - virtual RenderBufferData *_create_render_buffer_data(); + virtual RenderBufferData *_create_render_buffer_data() override; void _allocate_normal_roughness_texture(RenderBufferDataForwardClustered *rb); RID render_base_uniform_set; @@ -117,14 +126,12 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint64_t lightmap_texture_array_version = 0xFFFFFFFF; - virtual void _base_uniforms_changed(); - void _render_buffers_clear_uniform_set(RenderBufferDataForwardClustered *rb); - virtual void _render_buffers_uniform_set_changed(RID p_render_buffers); - virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); + virtual void _base_uniforms_changed() override; + virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) override; void _update_render_base_uniform_set(); RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); - RID _setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas = false, int p_index = 0); + RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); enum PassMode { PASS_MODE_COLOR, @@ -134,7 +141,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { PASS_MODE_SHADOW_DP, PASS_MODE_DEPTH, PASS_MODE_DEPTH_NORMAL_ROUGHNESS, - PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI, PASS_MODE_DEPTH_MATERIAL, PASS_MODE_SDF, }; @@ -158,8 +165,9 @@ class RenderForwardClustered : public RendererSceneRenderRD { RD::FramebufferFormatID framebuffer_format = 0; uint32_t element_offset = 0; uint32_t barrier = RD::BARRIER_MASK_ALL; + bool use_directional_soft_shadow = false; - RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, bool p_use_directional_soft_shadows, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { elements = p_elements; element_info = p_element_info; element_count = p_element_count; @@ -174,6 +182,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { screen_lod_threshold = p_screen_lod_threshold; element_offset = p_element_offset; barrier = p_barrier; + use_directional_soft_shadow = p_use_directional_soft_shadows; } }; @@ -186,26 +195,26 @@ class RenderForwardClustered : public RendererSceneRenderRD { }; enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 5, INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 6, INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 7, INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 8, INSTANCE_DATA_FLAG_USE_LIGHTMAP = 1 << 9, INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP = 1 << 10, - INSTANCE_DATA_FLAG_USE_GIPROBE = 1 << 11, + INSTANCE_DATA_FLAG_USE_VOXEL_GI = 1 << 11, INSTANCE_DATA_FLAG_MULTIMESH = 1 << 12, INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D = 1 << 13, INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR = 1 << 14, INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, - INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT = 16, - INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_MASK = 0x7, - INSTANCE_DATA_FLAG_SKELETON = 1 << 19, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, }; struct SceneState { + // This struct is loaded into Set 1 - Binding 0, populated at start of rendering a frame, must match with shader code struct UBO { float projection_matrix[16]; float inv_projection_matrix[16]; - float camera_matrix[16]; float inv_camera_matrix[16]; @@ -222,11 +231,6 @@ class RenderForwardClustered : public RendererSceneRenderRD { float penumbra_shadow_kernel[128]; float soft_shadow_kernel[128]; - uint32_t directional_penumbra_shadow_samples; - uint32_t directional_soft_shadow_samples; - uint32_t penumbra_shadow_samples; - uint32_t soft_shadow_samples; - float ambient_light_color_energy[4]; float ambient_color_sky_mix; @@ -319,8 +323,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint32_t max_lightmap_captures; RID lightmap_capture_buffer; - RID giprobe_ids[MAX_GI_PROBES]; - uint32_t giprobes_used = 0; + RID voxelgi_ids[MAX_VOXEL_GI_INSTANCESS]; + uint32_t voxelgis_used = 0; bool used_screen_texture = false; bool used_normal_texture = false; @@ -350,12 +354,15 @@ class RenderForwardClustered : public RendererSceneRenderRD { static RenderForwardClustered *singleton; - void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); - void _setup_giprobes(const PagedArray<RID> &p_giprobes); - void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform); + void _setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); + void _setup_voxelgis(const PagedArray<RID> &p_voxelgis); + void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform); struct RenderElementInfo { - uint32_t repeat : 22; + enum { MAX_REPEATS = (1 << 20) - 1 }; + uint32_t repeat : 20; + uint32_t uses_projector : 1; + uint32_t uses_softshadow : 1; uint32_t uses_lightmap : 1; uint32_t uses_forward_gi : 1; uint32_t lod_index : 8; @@ -373,8 +380,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint32_t render_list_thread_threshold = 500; void _update_instance_data_buffer(RenderListType p_render_list); - void _fill_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); - void _fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, const Plane &p_lod_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, bool p_append = false); + void _fill_instance_data(RenderListType p_render_list, int *p_render_info = nullptr, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, bool p_append = false); Map<Size2i, RID> sdfgi_framebuffer_size_cache; @@ -398,17 +405,20 @@ class RenderForwardClustered : public RendererSceneRenderRD { FLAG_USES_DEPTH_TEXTURE = 8192, FLAG_USES_NORMAL_TEXTURE = 16384, FLAG_USES_DOUBLE_SIDED_SHADOWS = 32768, + FLAG_USES_PARTICLE_TRAILS = 65536, }; union { struct { uint64_t lod_index : 8; - uint64_t surface_index : 10; + uint64_t surface_index : 8; uint64_t geometry_id : 32; - uint64_t material_id_low : 14; + uint64_t material_id_low : 16; - uint64_t material_id_hi : 18; + uint64_t material_id_hi : 16; uint64_t shader_id : 32; + uint64_t uses_softshadow : 1; + uint64_t uses_projector : 1; uint64_t uses_forward_gi : 1; uint64_t uses_lightmap : 1; uint64_t depth_layer : 4; @@ -453,12 +463,15 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint32_t layer_mask = 1; RID transforms_uniform_set; uint32_t instance_count = 0; + uint32_t trail_steps = 1; RID mesh_instance; bool can_sdfgi = false; + bool using_projectors = false; + bool using_softshadows = false; //used during setup uint32_t base_flags = 0; - Transform transform; - RID gi_probes[MAX_GI_PROBES_PER_INSTANCE]; + Transform3D transform; + RID voxel_gi_instances[MAX_VOXEL_GI_INSTANCESS_PER_INSTANCE]; RID lightmap_instance; GeometryInstanceLightmapSH *lightmap_sh = nullptr; GeometryInstanceSurfaceDataCache *surface_caches = nullptr; @@ -551,7 +564,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { } }; - void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha + void sort_by_reverse_depth_and_priority() { //used for alpha SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; sorter.sort(elements.ptr(), elements.size()); @@ -564,47 +577,51 @@ class RenderForwardClustered : public RendererSceneRenderRD { RenderList render_list[RENDER_LIST_MAX]; + virtual void _update_shader_quality_settings() override; + protected: - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; - virtual void _render_shadow_begin(); - virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true); - virtual void _render_shadow_process(); - virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + virtual void _render_shadow_begin() override; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) override; + virtual void _render_shadow_process() override; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) override; - virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances); + virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) override; public: - virtual GeometryInstance *geometry_instance_create(RID p_base); - virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton); - virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override); - virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials); - virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance); - virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb); - virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask); - virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias); - virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index); - virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9); - virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset); - virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable); - - virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance); - virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance); - - virtual void geometry_instance_free(GeometryInstance *p_geometry_instance); - - virtual uint32_t geometry_instance_get_pair_mask(); - virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count); - virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count); - virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count); - virtual void geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count); - - virtual bool free(RID p_rid); + virtual GeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) override; + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) override; + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) override; + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) override; + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) override; + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) override; + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) override; + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) override; + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) override; + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) override; + + virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) override; + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) override; + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) override; + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override; + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) override; + virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override; + + virtual void geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) override; + + virtual bool free(RID p_rid) override; RenderForwardClustered(RendererStorageRD *p_storage); ~RenderForwardClustered(); diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp index cb8c6e0cf3..be18a73989 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp @@ -30,6 +30,7 @@ #include "scene_shader_forward_clustered.h" #include "core/config/project_settings.h" +#include "core/math/math_defs.h" #include "render_forward_clustered.h" using namespace RendererSceneRenderImplementation; @@ -73,10 +74,14 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { uses_time = false; writes_modelview_or_projection = false; uses_world_coordinates = false; + uses_particle_trails = false; int depth_drawi = DEPTH_DRAW_OPAQUE; ShaderCompilerRD::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompilerRD::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompilerRD::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompilerRD::STAGE_FRAGMENT; actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); @@ -98,6 +103,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { actions.render_mode_flags["unshaded"] = &unshaded; actions.render_mode_flags["wireframe"] = &wireframe; + actions.render_mode_flags["particle_trails"] = &uses_particle_trails; actions.usage_flag_pointers["ALPHA"] = &uses_alpha; actions.render_mode_flags["depth_prepass_alpha"] = &uses_depth_pre_pass; @@ -125,8 +131,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = shader_singleton->shader.version_create(); @@ -141,14 +146,19 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { for (int i = 0; i < gen_code.defines.size(); i++) { print_line(gen_code.defines[i]); } + + Map<String, String>::Element * el = gen_code.code.front(); + while (el) { + print_line("\n**code " + el->key() + ":\n" + el->value()); + + el = el->next(); + } + print_line("\n**uniforms:\n" + gen_code.uniforms); - print_line("\n**vertex_globals:\n" + gen_code.vertex_global); - print_line("\n**vertex_code:\n" + gen_code.vertex); - print_line("\n**fragment_globals:\n" + gen_code.fragment_global); - print_line("\n**fragment_code:\n" + gen_code.fragment); - print_line("\n**light_code:\n" + gen_code.light); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT]); #endif - shader_singleton->shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.fragment, gen_code.defines); + shader_singleton->shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX], gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT], gen_code.defines); ERR_FAIL_COND(!shader_singleton->shader.version_is_valid(version)); ubo_size = gen_code.uniform_total_size; @@ -276,7 +286,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { multisample_state.enable_alpha_to_one = true; } - if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { blend_state = blend_state_blend; if (depth_draw == DEPTH_DRAW_OPAQUE) { depth_stencil.enable_depth_write = false; //alpha does not draw depth @@ -294,13 +304,13 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { continue; // do not use this version (will error if using it is attempted) } } else { - if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { blend_state = blend_state_opaque; } else if (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP) { //none, leave empty } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS) { blend_state = blend_state_depth_normal_roughness; - } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE) { + } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI) { blend_state = blend_state_depth_normal_roughness_giprobe; } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) { blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way @@ -309,13 +319,11 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { } else { //specular write blend_state = blend_state_opaque_specular; - depth_stencil.enable_depth_test = false; - depth_stencil.enable_depth_write = false; } } RID shader_variant = shader_singleton->shader.version_get_shader(version, k); - pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); } } } @@ -399,7 +407,8 @@ RS::ShaderNativeSourceCode SceneShaderForwardClustered::ShaderData::get_native_s return shader_singleton->shader.version_get_native_source_code(version); } -SceneShaderForwardClustered::ShaderData::ShaderData() { +SceneShaderForwardClustered::ShaderData::ShaderData() : + shader_list_element(this) { valid = false; uses_screen_texture = false; } @@ -415,6 +424,7 @@ SceneShaderForwardClustered::ShaderData::~ShaderData() { RendererStorageRD::ShaderData *SceneShaderForwardClustered::_create_shader_func() { ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); return shader_data; } @@ -426,94 +436,14 @@ void SceneShaderForwardClustered::MaterialData::set_next_pass(RID p_pass) { next_pass = p_pass; } -void SceneShaderForwardClustered::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool SceneShaderForwardClustered::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardClustered::MATERIAL_UNIFORM_SET); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardClustered::MATERIAL_UNIFORM_SET, RD::BARRIER_MASK_RASTER); } SceneShaderForwardClustered::MaterialData::~MaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *SceneShaderForwardClustered::_create_material_func(ShaderData *p_shader) { @@ -535,42 +465,30 @@ SceneShaderForwardClustered::~SceneShaderForwardClustered() { RD::get_singleton()->free(default_vec4_xform_buffer); RD::get_singleton()->free(shadow_sampler); - storage->free(wireframe_material_shader); storage->free(overdraw_material_shader); storage->free(default_shader); - storage->free(wireframe_material); storage->free(overdraw_material); storage->free(default_material); } -void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const String p_defines, bool p_is_low_end) { +void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const String p_defines) { storage = p_storage; { Vector<String> shader_versions; - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n"); - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n"); - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n"); - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n#define MODE_RENDER_GIPROBE\n"); - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); - shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_SDF\n"); - shader_versions.push_back(""); - shader_versions.push_back("\n#define USE_FORWARD_GI\n"); - shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n"); - shader_versions.push_back("\n#define USE_LIGHTMAP\n"); - shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n#define USE_LIGHTMAP\n"); - shader.initialize(shader_versions, p_defines); + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_DEPTH_PASS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n"); // SHADER_VERSION_DEPTH_PASS_DP + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n#define MODE_RENDER_VOXEL_GI\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); // SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_SDF\n"); // SHADER_VERSION_DEPTH_PASS_WITH_SDF + shader_versions.push_back(""); // SHADER_VERSION_COLOR_PASS + shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n"); // SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR + shader_versions.push_back("\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS + shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR - if (p_is_low_end) { - //disable the high end versions - shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS, false); - shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE, false); - shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_SDF, false); - shader.set_variant_enabled(SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI, false); - shader.set_variant_enabled(SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR, false); - shader.set_variant_enabled(SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR, false); - } + shader.initialize(shader_versions, p_defines); } storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_3D, _create_shader_funcs); @@ -608,6 +526,9 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin //builtins actions.renames["TIME"] = "scene_data.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); actions.renames["VIEWPORT_SIZE"] = "scene_data.viewport_size"; actions.renames["FRAGCOORD"] = "gl_FragCoord"; @@ -628,7 +549,6 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.renames["SSS_STRENGTH"] = "sss_strength"; actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; - actions.renames["SSS_TRANSMITTANCE_CURVE"] = "transmittance_curve"; actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; actions.renames["BACKLIGHT"] = "backlight"; actions.renames["AO"] = "ao"; @@ -652,6 +572,11 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.renames["CUSTOM2"] = "custom2_attrib"; actions.renames["CUSTOM3"] = "custom3_attrib"; + // not implemented but need these just in case code is in the shaders + actions.renames["VIEW_INDEX"] = "0"; + actions.renames["VIEW_MONO_LEFT"] = "0"; + actions.renames["VIEW_RIGHT"] = "1"; + //for light actions.renames["VIEW"] = "view"; actions.renames["LIGHT_COLOR"] = "light_color"; @@ -709,6 +634,7 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.render_mode_defines["ensure_correct_normals"] = "#define ENSURE_CORRECT_NORMALS\n"; actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); @@ -716,7 +642,6 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; } - actions.render_mode_defines["diffuse_oren_nayar"] = "#define DIFFUSE_OREN_NAYAR\n"; actions.render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; actions.render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; @@ -748,7 +673,7 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; actions.global_buffer_array_variable = "global_variables.data"; - actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; + actions.instance_uniform_index_variable = "instances.data[instance_index].instance_uniforms_ofs"; compiler.initialize(actions); } @@ -757,32 +682,52 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin //default material and shader default_shader = storage->shader_allocate(); storage->shader_initialize(default_shader); - storage->shader_set_code(default_shader, "shader_type spatial; void vertex() { ROUGHNESS = 0.8; } void fragment() { ALBEDO=vec3(0.6); ROUGHNESS=0.8; METALLIC=0.2; } \n"); + storage->shader_set_code(default_shader, R"( +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); default_material = storage->material_allocate(); storage->material_initialize(default_material); storage->material_set_shader(default_material, default_shader); MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); - if (!p_is_low_end) { - default_shader_sdfgi_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_DEPTH_PASS_WITH_SDF); - } + default_shader_sdfgi_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_DEPTH_PASS_WITH_SDF); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; } { overdraw_material_shader = storage->shader_allocate(); storage->shader_initialize(overdraw_material_shader); - storage->shader_set_code(overdraw_material_shader, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.2; }"); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + storage->shader_set_code(overdraw_material_shader, R"( +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); overdraw_material = storage->material_allocate(); storage->material_initialize(overdraw_material); storage->material_set_shader(overdraw_material, overdraw_material_shader); - wireframe_material_shader = storage->shader_allocate(); - storage->shader_initialize(wireframe_material_shader); - storage->shader_set_code(wireframe_material_shader, "shader_type spatial;\nrender_mode wireframe,unshaded;\n void fragment() { ALBEDO=vec3(0.0,0.0,0.0); }"); - wireframe_material = storage->material_allocate(); - storage->material_initialize(wireframe_material); - storage->material_set_shader(wireframe_material, wireframe_material_shader); + MaterialData *md = (MaterialData *)storage->material_get_data(overdraw_material, RendererStorageRD::SHADER_TYPE_3D); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; } { @@ -805,3 +750,16 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin shadow_sampler = RD::get_singleton()->sampler_create(sampler); } } + +void SceneShaderForwardClustered::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h index 368340e258..8d75f30a20 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h @@ -48,17 +48,24 @@ public: SHADER_VERSION_DEPTH_PASS, SHADER_VERSION_DEPTH_PASS_DP, SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS, - SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI, SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, SHADER_VERSION_DEPTH_PASS_WITH_SDF, SHADER_VERSION_COLOR_PASS, - SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI, SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR, SHADER_VERSION_LIGHTMAP_COLOR_PASS, SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR, + SHADER_VERSION_MAX }; + enum ShaderSpecializations { + SHADER_SPECIALIZATION_FORWARD_GI = 1 << 0, + SHADER_SPECIALIZATION_PROJECTOR = 1 << 1, + SHADER_SPECIALIZATION_SOFT_SHADOWS = 1 << 2, + SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS = 1 << 3, + }; + struct ShaderData : public RendererStorageRD::ShaderData { enum BlendMode { //used internally BLEND_MODE_MIX, @@ -126,6 +133,7 @@ public: bool uses_discard; bool uses_roughness; bool uses_normal; + bool uses_particle_trails; bool unshaded; bool uses_vertex; @@ -152,10 +160,13 @@ public: virtual Variant get_default_parameter(const StringName &p_parameter) const; virtual RS::ShaderNativeSourceCode get_native_source_code() const; + SelfList<ShaderData> shader_list_element; ShaderData(); virtual ~ShaderData(); }; + SelfList<ShaderData>::List shader_list; + RendererStorageRD::ShaderData *_create_shader_func(); static RendererStorageRD::ShaderData *_create_shader_funcs() { return static_cast<SceneShaderForwardClustered *>(singleton)->_create_shader_func(); @@ -164,17 +175,14 @@ public: struct MaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; ShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; uint64_t last_pass = 0; uint32_t index = 0; RID next_pass; uint8_t priority; virtual void set_render_priority(int p_priority); virtual void set_next_pass(RID p_pass); - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~MaterialData(); }; @@ -190,8 +198,6 @@ public: RID default_material; RID overdraw_material_shader; RID overdraw_material; - RID wireframe_material_shader; - RID wireframe_material; RID default_shader_rd; RID default_shader_sdfgi_rd; @@ -200,10 +206,18 @@ public: RID shadow_sampler; + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; SceneShaderForwardClustered(); ~SceneShaderForwardClustered(); - void init(RendererStorageRD *p_storage, const String p_defines, bool p_is_low_end); + void init(RendererStorageRD *p_storage, const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); }; } // namespace RendererSceneRenderImplementation diff --git a/servers/rendering/renderer_rd/forward_mobile/SCsub b/servers/rendering/renderer_rd/forward_mobile/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp new file mode 100644 index 0000000000..2064d9c5c5 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -0,0 +1,2679 @@ +/*************************************************************************/ +/* render_forward_mobile.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "render_forward_mobile.h" +#include "core/config/project_settings.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_server_default.h" + +using namespace RendererSceneRenderImplementation; + +RenderForwardMobile::ForwardID RenderForwardMobile::_allocate_forward_id(ForwardIDType p_type) { + int32_t index = -1; + for (uint32_t i = 0; i < forward_id_allocators[p_type].allocations.size(); i++) { + if (forward_id_allocators[p_type].allocations[i] == false) { + index = i; + break; + } + } + + if (index == -1) { + index = forward_id_allocators[p_type].allocations.size(); + forward_id_allocators[p_type].allocations.push_back(true); + forward_id_allocators[p_type].map.push_back(0xFF); + } else { + forward_id_allocators[p_type].allocations[index] = true; + } + + return index; +} +void RenderForwardMobile::_free_forward_id(ForwardIDType p_type, ForwardID p_id) { + ERR_FAIL_INDEX(p_id, (ForwardID)forward_id_allocators[p_type].allocations.size()); + forward_id_allocators[p_type].allocations[p_id] = false; +} + +void RenderForwardMobile::_map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) { + forward_id_allocators[p_type].map[p_id] = p_index; +} + +/* Render buffer */ + +void RenderForwardMobile::RenderBufferDataForwardMobile::clear() { + if (color_msaa.is_valid()) { + RD::get_singleton()->free(color_msaa); + color_msaa = RID(); + } + + if (depth_msaa.is_valid()) { + RD::get_singleton()->free(depth_msaa); + depth_msaa = RID(); + } + + color = RID(); + depth = RID(); + for (int i = 0; i < FB_CONFIG_MAX; i++) { + color_fbs[i] = RID(); + } +} + +void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { + clear(); + + bool is_half_resolution = false; // Set this once we support this feature. + + msaa = p_msaa; + + width = p_width; + height = p_height; + view_count = p_view_count; + + color = p_color_buffer; + depth = p_depth_buffer; + + // We are creating 4 configurations here for our framebuffers. + + if (p_msaa == RS::VIEWPORT_MSAA_DISABLED) { + Vector<RID> fb; + fb.push_back(p_color_buffer); // 0 - color buffer + fb.push_back(depth); // 1 - depth buffer + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + color_fbs[FB_CONFIG_ONE_PASS] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + // - add sky pass + passes.push_back(pass); + color_fbs[FB_CONFIG_TWO_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + // - add alpha pass + passes.push_back(pass); + color_fbs[FB_CONFIG_THREE_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + if (!is_half_resolution) { + // - add blit to 2D pass + fb.push_back(p_target_buffer); // 2 - target buffer + + RD::FramebufferPass blit_pass; + blit_pass.color_attachments.push_back(2); + blit_pass.input_attachments.push_back(0); + passes.push_back(blit_pass); + + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + } else { + // can't do our blit pass if resolutions don't match + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RID(); + } + } else { + RD::DataFormat color_format = RenderForwardMobile::singleton->_render_buffers_get_color_format(); + + RD::TextureFormat tf; + if (view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } else { + tf.texture_type = RD::TEXTURE_TYPE_2D; + } + tf.format = color_format; + tf.width = p_width; + tf.height = p_height; + tf.array_layers = view_count; // create a layer for every view + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + RD::TextureSamples ts[RS::VIEWPORT_MSAA_MAX] = { + RD::TEXTURE_SAMPLES_1, + RD::TEXTURE_SAMPLES_2, + RD::TEXTURE_SAMPLES_4, + RD::TEXTURE_SAMPLES_8, + RD::TEXTURE_SAMPLES_16 + }; + + texture_samples = ts[p_msaa]; + tf.samples = texture_samples; + + color_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT; + tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + depth_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + { + Vector<RID> fb; + fb.push_back(color_msaa); // 0 - msaa color buffer + fb.push_back(depth_msaa); // 1 - msaa depth buffer + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + + // - add sky pass + fb.push_back(color); // 2 - color buffer + passes.push_back(pass); // without resolve for our 3 + 4 subpass config + { + // but with resolve for our 2 subpass config + Vector<RD::FramebufferPass> two_passes; + two_passes.push_back(pass); // opaque subpass without resolve + pass.resolve_attachments.push_back(2); + two_passes.push_back(pass); // sky subpass with resolve + + color_fbs[FB_CONFIG_TWO_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, two_passes, RenderingDevice::INVALID_ID, view_count); + } + + // - add alpha pass (with resolve, we just added that above) + passes.push_back(pass); + color_fbs[FB_CONFIG_THREE_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + { + // we also need our one pass with resolve + Vector<RD::FramebufferPass> one_pass_with_resolve; + one_pass_with_resolve.push_back(pass); // note our pass configuration already has resolve.. + color_fbs[FB_CONFIG_ONE_PASS] = RD::get_singleton()->framebuffer_create_multipass(fb, one_pass_with_resolve, RenderingDevice::INVALID_ID, view_count); + } + + if (!is_half_resolution) { + // - add blit to 2D pass + fb.push_back(p_target_buffer); // 3 - target buffer + RD::FramebufferPass blit_pass; + blit_pass.color_attachments.push_back(3); + blit_pass.input_attachments.push_back(2); + passes.push_back(blit_pass); + + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + } else { + // can't do our blit pass if resolutions don't match + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RID(); + } + } + } +} + +RID RenderForwardMobile::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + // Our attachments + Vector<RID> fb; + fb.push_back(p_color); // 0 + fb.push_back(p_depth); // 1 + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + + // - sky pass + passes.push_back(pass); + + // - alpha pass + passes.push_back(pass); + + return RD::get_singleton()->framebuffer_create_multipass(fb, passes); +} + +RenderForwardMobile::RenderBufferDataForwardMobile::~RenderBufferDataForwardMobile() { + clear(); +} + +RendererSceneRenderRD::RenderBufferData *RenderForwardMobile::_create_render_buffer_data() { + return memnew(RenderBufferDataForwardMobile); +} + +bool RenderForwardMobile::free(RID p_rid) { + if (RendererSceneRenderRD::free(p_rid)) { + return true; + } + return false; +} + +/* Render functions */ + +RD::DataFormat RenderForwardMobile::_render_buffers_get_color_format() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + return RD::DATA_FORMAT_A2B10G10R10_UNORM_PACK32; +} + +bool RenderForwardMobile::_render_buffers_can_be_storage() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + // Doesn't support storage + return false; +} + +RID RenderForwardMobile::_setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas, int p_index) { + //there should always be enough uniform buffers for render passes, otherwise bugs + ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); + + RenderBufferDataForwardMobile *rb = nullptr; + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = (RenderBufferDataForwardMobile *)render_buffers_get_data(p_render_data->render_buffers); + } + + // default render buffer and scene state uniform set + // loaded into set 1 + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + + { + RID radiance_texture; + if (p_radiance_texture.is_valid()) { + radiance_texture = p_radiance_texture; + } else { + radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + } + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(radiance_texture); + uniforms.push_back(u); + } + + { + RID ref_texture = (p_render_data && p_render_data->reflection_atlas.is_valid()) ? reflection_atlas_get_texture(p_render_data->reflection_atlas) : RID(); + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (ref_texture.is_valid()) { + u.ids.push_back(ref_texture); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK)); + } + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture; + if (p_render_data && p_render_data->shadow_atlas.is_valid()) { + texture = shadow_atlas_get_texture(p_render_data->shadow_atlas); + } + if (!texture.is_valid()) { + texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + } + u.ids.push_back(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && directional_shadow_get_texture().is_valid()) { + u.ids.push_back(directional_shadow_get_texture()); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } + uniforms.push_back(u); + } + + /* we have limited ability to keep textures like this so we're moving this to a set we change before drawing geometry and just pushing the needed texture in */ + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.resize(scene_state.max_lightmaps); + RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + if (p_render_data && i < p_render_data->lightmaps->size()) { + RID base = lightmap_instance_get_lightmap((*p_render_data->lightmaps)[i]); + RID texture = storage->lightmap_get_texture(base); + RID rd_texture = storage->texture_get_rd_texture(texture); + u.ids.write[i] = rd_texture; + } else { + u.ids.write[i] = default_tex; + } + } + + uniforms.push_back(u); + } + + /* + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.resize(MAX_VOXEL_GI_INSTANCESS); + RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { + if (i < (int)p_voxel_gi_instances.size()) { + RID tex = gi.voxel_gi_instance_get_texture(p_voxel_gi_instances[i]); + if (!tex.is_valid()) { + tex = default_tex; + } + u.ids.write[i] = tex; + } else { + u.ids.write[i] = default_tex; + } + } + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : default_vec4_xform_buffer; + u.ids.push_back(cb); + uniforms.push_back(u); + } + */ + + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID dbt = rb ? render_buffers_get_back_depth_texture(p_render_data->render_buffers) : RID(); + RID texture = (dbt.is_valid()) ? dbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + u.ids.push_back(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_data->render_buffers) : RID(); + RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + u.ids.push_back(texture); + uniforms.push_back(u); + } + + if (p_index >= (int)render_pass_uniform_sets.size()) { + render_pass_uniform_sets.resize(p_index + 1); + } + + if (render_pass_uniform_sets[p_index].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[p_index])) { + RD::get_singleton()->free(render_pass_uniform_sets[p_index]); + } + + render_pass_uniform_sets[p_index] = RD::get_singleton()->uniform_set_create(uniforms, scene_shader.default_shader_rd, RENDER_PASS_UNIFORM_SET); + return render_pass_uniform_sets[p_index]; +} + +void RenderForwardMobile::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform) { + // This probably needs to change... + scene_state.lightmaps_used = 0; + for (int i = 0; i < (int)p_lightmaps.size(); i++) { + if (i >= (int)scene_state.max_lightmaps) { + break; + } + + RID lightmap = lightmap_instance_get_lightmap(p_lightmaps[i]); + + Basis to_lm = lightmap_instance_get_transform(p_lightmaps[i]).basis.inverse() * p_cam_transform.basis; + to_lm = to_lm.inverse().transposed(); //will transform normals + RendererStorageRD::store_transform_3x3(to_lm, scene_state.lightmaps[i].normal_xform); + scene_state.lightmap_ids[i] = p_lightmaps[i]; + scene_state.lightmap_has_sh[i] = storage->lightmap_uses_spherical_harmonics(lightmap); + + scene_state.lightmaps_used++; + } + if (scene_state.lightmaps_used > 0) { + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); + } +} + +void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { + RenderBufferDataForwardMobile *render_buffer = nullptr; + if (p_render_data->render_buffers.is_valid()) { + render_buffer = (RenderBufferDataForwardMobile *)render_buffers_get_data(p_render_data->render_buffers); + } + RendererSceneEnvironmentRD *env = get_environment(p_render_data->environment); + + RENDER_TIMESTAMP("Setup 3D Scene"); + + Vector2 vp_he = p_render_data->cam_projection.get_viewport_half_extents(); + scene_state.ubo.viewport_size[0] = vp_he.x; + scene_state.ubo.viewport_size[1] = vp_he.y; + scene_state.ubo.directional_light_count = 0; + + // We can only use our full subpass approach if we're: + // - not reading from SCREEN_TEXTURE/DEPTH_TEXTURE + // - not using ssr/sss (currently not supported) + // - not using glow or other post effects (can't do 4th subpass) + // - rendering to a half sized render buffer (can't do 4th subpass) + // We'll need to restrict how far we're going with subpasses based on this. + + Size2i screen_size; + RID framebuffer; + bool reverse_cull = false; + bool using_subpass_transparent = true; + bool using_subpass_post_process = true; + + bool is_half_resolution = false; // Set this once we support this feature. + bool using_ssr = false; // I don't think we support this in our mobile renderer so probably should phase it out + bool using_sss = false; // I don't think we support this in our mobile renderer so probably should phase it out + + // fill our render lists early so we can find out if we use various features + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_element_info(RENDER_LIST_OPAQUE); + _fill_element_info(RENDER_LIST_ALPHA); + + if (p_render_data->render_info) { + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_render_data->instances->size(); + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_render_data->instances->size(); + } + + if (render_buffer) { + // setup rendering to render buffer + screen_size.x = render_buffer->width; + screen_size.y = render_buffer->height; + + if (is_half_resolution) { + // can't do blit subpass + using_subpass_post_process = false; + } else if (env && (env->glow_enabled || env->auto_exposure || camera_effects_uses_dof(p_render_data->camera_effects))) { + // can't do blit subpass + using_subpass_post_process = false; + } + + if (using_ssr || using_sss || scene_state.used_screen_texture || scene_state.used_depth_texture) { + // can't use our last two subpasses + using_subpass_transparent = false; + using_subpass_post_process = false; + } + + if (using_subpass_post_process) { + // all as subpasses + framebuffer = render_buffer->color_fbs[FB_CONFIG_FOUR_SUBPASSES]; + } else if (using_subpass_transparent) { + // our tonemap pass is separate + framebuffer = render_buffer->color_fbs[FB_CONFIG_THREE_SUBPASSES]; + } else { + // only opaque and sky as subpasses + framebuffer = render_buffer->color_fbs[FB_CONFIG_TWO_SUBPASSES]; + } + } else if (p_render_data->reflection_probe.is_valid()) { + uint32_t resolution = reflection_probe_instance_get_resolution(p_render_data->reflection_probe); + screen_size.x = resolution; + screen_size.y = resolution; + + framebuffer = reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); + + if (storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + p_render_data->environment = RID(); //no environment on interiors + env = nullptr; + } + + reverse_cull = true; + using_subpass_transparent = true; // we ignore our screen/depth texture here + using_subpass_post_process = false; // not applicable at all for reflection probes. + } else { + ERR_FAIL(); //bug? + } + + RD::get_singleton()->draw_command_begin_label("Render Setup"); + + _setup_lightmaps(*p_render_data->lightmaps, p_render_data->cam_transform); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) + + RD::get_singleton()->draw_command_end_label(); // Render Setup + + // setup environment + RID radiance_texture; + bool draw_sky = false; + bool draw_sky_fog_only = false; + + Color clear_color = p_default_bg_color; + bool keep_color = false; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW) { + clear_color = Color(0, 0, 0, 1); //in overdraw mode, BG should always be black + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG bg_mode = environment_get_background(p_render_data->environment); + float bg_energy = environment_get_bg_energy(p_render_data->environment); + switch (bg_mode) { + case RS::ENV_BG_CLEAR_COLOR: { + clear_color = p_default_bg_color; + clear_color.r *= bg_energy; + clear_color.g *= bg_energy; + clear_color.b *= bg_energy; + /* + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers) || environment_is_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); + } + */ + } break; + case RS::ENV_BG_COLOR: { + clear_color = environment_get_bg_color(p_render_data->environment); + clear_color.r *= bg_energy; + clear_color.g *= bg_energy; + clear_color.b *= bg_energy; + /* + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers) || environment_is_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); + } + */ + } break; + case RS::ENV_BG_SKY: { + draw_sky = true; + } break; + case RS::ENV_BG_CANVAS: { + keep_color = true; + } break; + case RS::ENV_BG_KEEP: { + keep_color = true; + } break; + case RS::ENV_BG_CAMERA_FEED: { + } break; + default: { + } + } + // setup sky if used for ambient, reflections, or background + if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_render_data->environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_render_data->environment) == RS::ENV_AMBIENT_SOURCE_SKY) { + RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); + CameraMatrix projection = p_render_data->cam_projection; + if (p_render_data->reflection_probe.is_valid()) { + CameraMatrix correction; + correction.set_depth_correction(true); + projection = correction * p_render_data->cam_projection; + } + + sky.setup(env, p_render_data->render_buffers, projection, p_render_data->cam_transform, screen_size, this); + + RID sky_rid = env->sky; + if (sky_rid.is_valid()) { + sky.update(env, projection, p_render_data->cam_transform, time); + radiance_texture = sky.sky_get_radiance_texture_rd(sky_rid); + } else { + // do not try to draw sky if invalid + draw_sky = false; + } + RD::get_singleton()->draw_command_end_label(); // Setup Sky + } + } else { + clear_color = p_default_bg_color; + } + + // update sky buffers (if required) + if (draw_sky || draw_sky_fog_only) { + // !BAS! @TODO See if we can limit doing some things double and maybe even move this into _pre_opaque_render + // and change Forward Clustered in the same way as we have here (but without using subpasses) + RENDER_TIMESTAMP("Setup Sky resolution buffers"); + + RD::get_singleton()->draw_command_begin_label("Setup Sky resolution buffers"); + + if (p_render_data->reflection_probe.is_valid()) { + CameraMatrix correction; + correction.set_depth_correction(true); + CameraMatrix projection = correction * p_render_data->cam_projection; + sky.update_res_buffers(env, 1, &projection, p_render_data->cam_transform, time); + } else { + sky.update_res_buffers(env, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time); + } + + RD::get_singleton()->draw_command_end_label(); // Setup Sky resolution buffers + } + + _pre_opaque_render(p_render_data, false, false, RID(), RID()); + + uint32_t spec_constant_base_flags = 0; + + { + //figure out spec constants + + if (p_render_data->directional_light_count > 0) { + if (p_render_data->directional_light_soft_shadows) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS; + } + } else { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS; + } + + if (!is_environment(p_render_data->environment) || environment_is_fog_enabled(p_render_data->environment)) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_FOG; + } + } + { + if (render_buffer) { + RD::get_singleton()->draw_command_begin_label("Render 3D Pass"); + } else { + RD::get_singleton()->draw_command_begin_label("Render Reflection Probe Pass"); + } + + // opaque pass + + RD::get_singleton()->draw_command_begin_label("Render Opaque Subpass"); + + scene_state.ubo.directional_light_count = p_render_data->directional_light_count; + + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); + + if (using_subpass_transparent && using_subpass_post_process) { + RENDER_TIMESTAMP("Render Opaque + Transparent + Tonemap"); + } else if (using_subpass_transparent) { + RENDER_TIMESTAMP("Render Opaque + Transparent"); + } else { + RENDER_TIMESTAMP("Render Opaque"); + } + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + + bool can_continue_color = !using_subpass_transparent && !scene_state.used_screen_texture && !using_ssr && !using_sss; + bool can_continue_depth = !using_subpass_transparent && !scene_state.used_depth_texture && !using_ssr && !using_sss; + + { + // regular forward for now + Vector<Color> c; + c.push_back(clear_color.to_linear()); // our render buffer + if (render_buffer) { + if (render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { + c.push_back(clear_color.to_linear()); // our resolve buffer + } + if (using_subpass_post_process) { + c.push_back(Color()); // our 2D buffer we're copying into + } + } + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + } + + RD::get_singleton()->draw_command_end_label(); //Render Opaque Subpass + + if (draw_sky || draw_sky_fog_only) { + RD::get_singleton()->draw_command_begin_label("Draw Sky Subpass"); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + if (p_render_data->reflection_probe.is_valid()) { + CameraMatrix correction; + correction.set_depth_correction(true); + CameraMatrix projection = correction * p_render_data->cam_projection; + sky.draw(draw_list, env, framebuffer, 1, &projection, p_render_data->cam_transform, time); + } else { + sky.draw(draw_list, env, framebuffer, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time); + } + + RD::get_singleton()->draw_command_end_label(); // Draw Sky Subpass + + // note, if MSAA is used in 2-subpass approach we should get an automatic resolve here + } else { + // switch to subpass but we do nothing here so basically we skip (though this should trigger resolve with 2-subpass MSAA). + RD::get_singleton()->draw_list_switch_to_next_pass(); + } + + if (!using_subpass_transparent) { + // We're done with our subpasses so end our container pass + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass + } + + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } + + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } + + // transparent pass + + RD::get_singleton()->draw_command_begin_label("Render Transparent Subpass"); + + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); + + if (using_subpass_transparent) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_switch_to_next_pass_split(thread_draw_lists.size(), thread_draw_lists.ptr()); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + + // note if we are using MSAA we should get an automatic resolve through our subpass configuration. + + // blit to tonemap + if (render_buffer && using_subpass_post_process) { + _post_process_subpass(render_buffer->color, framebuffer, p_render_data); + } + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass + + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + RENDER_TIMESTAMP("Render Transparent"); + + framebuffer = render_buffer->color_fbs[FB_CONFIG_ONE_PASS]; + + // this may be needed if we re-introduced steps that change info, not sure which do so in the previous implementation + // _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + } + } + + if (render_buffer && !using_subpass_post_process) { + RD::get_singleton()->draw_command_begin_label("Post process pass"); + + // If we need extra effects we do this in its own pass + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(p_render_data); + + RD::get_singleton()->draw_command_end_label(); // Post process pass + } + + if (render_buffer) { + _disable_clear_request(p_render_data); + } +} + +/* these are being called from RendererSceneRenderRD::_pre_opaque_render */ + +void RenderForwardMobile::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); + _update_render_base_uniform_set(); + + render_list[RENDER_LIST_SECONDARY].clear(); +} + +void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RendererScene::RenderInfo *p_render_info) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; + + if (p_render_info) { + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_instances.size(); + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_instances.size(); + } + RenderDataRD render_data; + render_data.cam_projection = p_projection; + render_data.cam_transform = p_transform; + render_data.view_projection[0] = p_projection; + render_data.z_near = 0.0; + render_data.z_far = p_zfar; + render_data.instances = &p_instances; + render_data.render_info = p_render_info; + render_data.lod_camera_plane = p_camera_plane; + render_data.lod_distance_multiplier = p_lod_distance_multiplier; + + scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; + + _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { + render_data.screen_lod_threshold = 0.0; + } else { + render_data.screen_lod_threshold = p_screen_lod_threshold; + } + + PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; + + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, true); + uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; + render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); + _fill_element_info(RENDER_LIST_SECONDARY, render_list_from, render_list_size); + + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } + + shadow_pass.element_from = render_list_from; + shadow_pass.element_count = render_list_size; + shadow_pass.flip_cull = flip_cull; + shadow_pass.pass_mode = pass_mode; + + shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete + shadow_pass.camera_plane = p_camera_plane; + shadow_pass.screen_lod_threshold = render_data.screen_lod_threshold; + shadow_pass.lod_distance_multiplier = render_data.lod_distance_multiplier; + + shadow_pass.framebuffer = p_framebuffer; + shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); + shadow_pass.final_depth_action = p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE; + shadow_pass.rect = p_rect; + + scene_state.shadow_passes.push_back(shadow_pass); + } +} + +void RenderForwardMobile::_render_shadow_process() { + //render shadows one after the other, so this can be done un-barriered and the driver can optimize (as well as allow us to run compute at the same time) + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + //render passes need to be configured after instance buffer is done, since they need the latest version + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID(), false, i); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_shadow_end(uint32_t p_barrier) { + RD::get_singleton()->draw_command_begin_label("Shadow Render"); + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, shadow_pass.rp_uniform_set, 0, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); + } + + if (p_barrier != RD::BARRIER_MASK_NO_BARRIER) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, p_barrier); + } + RD::get_singleton()->draw_command_end_label(); +} + +/* */ + +void RenderForwardMobile::_render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { + RENDER_TIMESTAMP("Setup Rendering Material"); + + RD::get_singleton()->draw_command_begin_label("Render Material"); + + _update_render_base_uniform_set(); + + scene_state.ubo.dual_paraboloid_side = 0; + scene_state.ubo.material_uv2_mode = false; + + RenderDataRD render_data; + render_data.cam_projection = p_cam_projection; + render_data.cam_transform = p_cam_transform; + render_data.view_projection[0] = p_cam_projection; + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, 0); + //regular forward for now + Vector<Color> clear; + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { + RENDER_TIMESTAMP("Setup Rendering UV2"); + + RD::get_singleton()->draw_command_begin_label("Render UV2"); + + _update_render_base_uniform_set(); + + scene_state.ubo.dual_paraboloid_side = 0; + scene_state.ubo.material_uv2_mode = true; + + RenderDataRD render_data; + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, true, 0); + //regular forward for now + Vector<Color> clear; + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + clear.push_back(Color(0, 0, 0, 0)); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + + const int uv_offset_count = 9; + static const Vector2 uv_offsets[uv_offset_count] = { + Vector2(-1, 1), + Vector2(1, 1), + Vector2(1, -1), + Vector2(-1, -1), + Vector2(-1, 0), + Vector2(1, 0), + Vector2(0, -1), + Vector2(0, 1), + Vector2(0, 0), + + }; + + for (int i = 0; i < uv_offset_count; i++) { + Vector2 ofs = uv_offsets[i]; + ofs.x /= p_region.size.width; + ofs.y /= p_region.size.height; + render_list_params.uv_offset = ofs; + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //first wireframe, for pseudo conservative + } + render_list_params.uv_offset = Vector2(); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //second regular triangles + + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { + // we don't do GI in low end.. +} + +void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { + RENDER_TIMESTAMP("Setup Render Collider Heightfield"); + + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + + _update_render_base_uniform_set(); + scene_state.ubo.dual_paraboloid_side = 0; + + RenderDataRD render_data; + render_data.cam_projection = p_cam_projection; + render_data.cam_transform = p_cam_transform; + render_data.view_projection[0] = p_cam_projection; + render_data.z_near = 0.0; + render_data.z_far = p_cam_projection.get_z_far(); + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), true, Color(), false, false); + + PassMode pass_mode = PASS_MODE_SHADOW; + + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render Collider Heightfield"); + + { + //regular forward for now + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, rp_uniform_set, 0); + _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); + } + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_base_uniforms_changed() { + if (!render_base_uniform_set.is_null() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + render_base_uniform_set = RID(); +} + +void RenderForwardMobile::_update_render_base_uniform_set() { + if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != storage->lightmap_array_get_version())) { + if (render_base_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + + // This is all loaded into set 0 + + lightmap_texture_array_version = storage->lightmap_array_get_version(); + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 1; + u.ids.resize(12); + RID *ids_ptr = u.ids.ptrw(); + ids_ptr[0] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.ids.push_back(scene_shader.shadow_sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_omni_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_spot_light_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_reflection_probe_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(get_directional_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(scene_state.lightmap_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(scene_state.lightmap_capture_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = storage->decal_atlas_get_texture(); + u.ids.push_back(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 12; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = storage->decal_atlas_get_texture_srgb(); + u.ids.push_back(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 13; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_decal_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 14; + u.ids.push_back(storage->global_variables_get_storage_buffer()); + uniforms.push_back(u); + } + + render_base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_shader.default_shader_rd, SCENE_UNIFORM_SET); + } +} + +RID RenderForwardMobile::_render_buffers_get_normal_texture(RID p_render_buffers) { + // RenderBufferDataForwardMobile *rb = (RenderBufferDataForwardMobile *)render_buffers_get_data(p_render_buffers); + + // We don't have this. This is for debugging + // return rb->normal_roughness_buffer; + return RID(); +} + +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} + +void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append) { + if (p_render_list == RENDER_LIST_OPAQUE) { + scene_state.used_sss = false; + scene_state.used_screen_texture = false; + scene_state.used_normal_texture = false; + scene_state.used_depth_texture = false; + } + uint32_t lightmap_captures_used = 0; + + Plane near_plane(p_render_data->cam_transform.origin, -p_render_data->cam_transform.basis.get_axis(Vector3::AXIS_Z)); + near_plane.d += p_render_data->cam_projection.get_z_near(); + float z_max = p_render_data->cam_projection.get_z_far() - p_render_data->cam_projection.get_z_near(); + + RenderList *rl = &render_list[p_render_list]; + + // Parse any updates on our geometry, updates surface caches and such + _update_dirty_geometry_instances(); + + if (!p_append) { + rl->clear(); + if (p_render_list == RENDER_LIST_OPAQUE) { + render_list[RENDER_LIST_ALPHA].clear(); //opaque fills alpha too + } + } + + //fill list + + for (int i = 0; i < (int)p_render_data->instances->size(); i++) { + GeometryInstanceForwardMobile *inst = static_cast<GeometryInstanceForwardMobile *>((*p_render_data->instances)[i]); + + Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); + inst->depth = near_plane.distance_to(support_min); + uint32_t depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15); + + uint32_t flags = inst->base_flags; //fill flags if appropriate + + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } + + bool uses_lightmap = false; + // bool uses_gi = false; + + if (p_render_list == RENDER_LIST_OPAQUE) { + if (inst->lightmap_instance.is_valid()) { + int32_t lightmap_cull_index = -1; + for (uint32_t j = 0; j < scene_state.lightmaps_used; j++) { + if (scene_state.lightmap_ids[j] == inst->lightmap_instance) { + lightmap_cull_index = j; + break; + } + } + if (lightmap_cull_index >= 0) { + inst->gi_offset_cache = inst->lightmap_slice_index << 16; + inst->gi_offset_cache |= lightmap_cull_index; + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; + if (scene_state.lightmap_has_sh[lightmap_cull_index]) { + flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; + } + uses_lightmap = true; + } else { + inst->gi_offset_cache = 0xFFFFFFFF; + } + + } else if (inst->lightmap_sh) { + if (lightmap_captures_used < scene_state.max_lightmap_captures) { + const Color *src_capture = inst->lightmap_sh->sh; + LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used]; + for (int j = 0; j < 9; j++) { + lcd.sh[j * 4 + 0] = src_capture[j].r; + lcd.sh[j * 4 + 1] = src_capture[j].g; + lcd.sh[j * 4 + 2] = src_capture[j].b; + lcd.sh[j * 4 + 3] = src_capture[j].a; + } + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; + inst->gi_offset_cache = lightmap_captures_used; + lightmap_captures_used++; + uses_lightmap = true; + } + } + } + inst->flags_cache = flags; + + GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; + + while (surf) { + surf->sort.uses_lightmap = 0; + + // LOD + + if (p_render_data->screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { + //lod + Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_render_data->lod_camera_plane.normal); + Vector3 lod_support_max = inst->transformed_aabb.get_support(p_render_data->lod_camera_plane.normal); + + float distance_min = p_render_data->lod_camera_plane.distance_to(lod_support_min); + float distance_max = p_render_data->lod_camera_plane.distance_to(lod_support_max); + + float distance = 0.0; + + if (distance_min * distance_max < 0.0) { + //crossing plane + distance = 0.0; + } else if (distance_min >= 0.0) { + distance = distance_min; + } else if (distance_max <= 0.0) { + distance = -distance_max; + } + + uint32_t indices; + surf->lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, &indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } + } else { + surf->lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } + } + + // ADD Element + if (p_pass_mode == PASS_MODE_COLOR) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { + rl->add_element(surf); + } + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { + render_list[RENDER_LIST_ALPHA].add_element(surf); + } + + if (uses_lightmap) { + surf->sort.uses_lightmap = 1; // This needs to become our lightmap index but we'll do that in a separate PR. + } + + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING) { + scene_state.used_sss = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE) { + scene_state.used_screen_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE) { + scene_state.used_normal_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE) { + scene_state.used_depth_texture = true; + } + + } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { + rl->add_element(surf); + } + } else { + if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } + } + + surf->sort.depth_layer = depth_layer; + + surf = surf->next; + } + } +} + +void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { + //!BAS! need to go through this and find out what we don't need anymore + + // This populates our UBO with main scene data that is pushed into set 1 + + //CameraMatrix projection = p_render_data->cam_projection; + //projection.flip_y(); // Vulkan and modern APIs use Y-Down + CameraMatrix correction; + correction.set_depth_correction(p_flip_y); + CameraMatrix projection = correction * p_render_data->cam_projection; + + //store camera into ubo + RendererStorageRD::store_camera(projection, scene_state.ubo.projection_matrix); + RendererStorageRD::store_camera(projection.inverse(), scene_state.ubo.inv_projection_matrix); + RendererStorageRD::store_transform(p_render_data->cam_transform, scene_state.ubo.camera_matrix); + RendererStorageRD::store_transform(p_render_data->cam_transform.affine_inverse(), scene_state.ubo.inv_camera_matrix); + + for (uint32_t v = 0; v < p_render_data->view_count; v++) { + projection = correction * p_render_data->view_projection[v]; + RendererStorageRD::store_camera(projection, scene_state.ubo.projection_matrix_view[v]); + RendererStorageRD::store_camera(projection.inverse(), scene_state.ubo.inv_projection_matrix_view[v]); + } + + scene_state.ubo.z_far = p_render_data->z_far; + scene_state.ubo.z_near = p_render_data->z_near; + + scene_state.ubo.pancake_shadows = p_pancake_shadows; + + RendererStorageRD::store_soft_shadow_kernel(directional_penumbra_shadow_kernel_get(), scene_state.ubo.directional_penumbra_shadow_kernel); + RendererStorageRD::store_soft_shadow_kernel(directional_soft_shadow_kernel_get(), scene_state.ubo.directional_soft_shadow_kernel); + RendererStorageRD::store_soft_shadow_kernel(penumbra_shadow_kernel_get(), scene_state.ubo.penumbra_shadow_kernel); + RendererStorageRD::store_soft_shadow_kernel(soft_shadow_kernel_get(), scene_state.ubo.soft_shadow_kernel); + + Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); + scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; + scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; + + if (p_render_data->shadow_atlas.is_valid()) { + Vector2 sas = shadow_atlas_get_size(p_render_data->shadow_atlas); + scene_state.ubo.shadow_atlas_pixel_size[0] = 1.0 / sas.x; + scene_state.ubo.shadow_atlas_pixel_size[1] = 1.0 / sas.y; + } + { + Vector2 dss = directional_shadow_get_size(); + scene_state.ubo.directional_shadow_pixel_size[0] = 1.0 / dss.x; + scene_state.ubo.directional_shadow_pixel_size[1] = 1.0 / dss.y; + } + + //time global variables + scene_state.ubo.time = time; + + /* + scene_state.ubo.gi_upscale_for_msaa = false; + scene_state.ubo.volumetric_fog_enabled = false; + scene_state.ubo.fog_enabled = false; + + if (p_render_data->render_buffers.is_valid()) { + RenderBufferDataForwardMobile *render_buffers = (RenderBufferDataForwardMobile *)render_buffers_get_data(p_render_data->render_buffers); + if (render_buffers->msaa != RS::VIEWPORT_MSAA_DISABLED) { + scene_state.ubo.gi_upscale_for_msaa = true; + } + + if (render_buffers_has_volumetric_fog(p_render_data->render_buffers)) { + scene_state.ubo.volumetric_fog_enabled = true; + float fog_end = render_buffers_get_volumetric_fog_end(p_render_data->render_buffers); + if (fog_end > 0.0) { + scene_state.ubo.volumetric_fog_inv_length = 1.0 / fog_end; + } else { + scene_state.ubo.volumetric_fog_inv_length = 1.0; + } + + float fog_detail_spread = render_buffers_get_volumetric_fog_detail_spread(p_render_data->render_buffers); //reverse lookup + if (fog_detail_spread > 0.0) { + scene_state.ubo.volumetric_fog_detail_spread = 1.0 / fog_detail_spread; + } else { + scene_state.ubo.volumetric_fog_detail_spread = 1.0; + } + } + } + + */ + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + scene_state.ubo.use_ambient_light = true; + scene_state.ubo.ambient_light_color_energy[0] = 1; + scene_state.ubo.ambient_light_color_energy[1] = 1; + scene_state.ubo.ambient_light_color_energy[2] = 1; + scene_state.ubo.ambient_light_color_energy[3] = 1.0; + scene_state.ubo.use_ambient_cubemap = false; + scene_state.ubo.use_reflection_cubemap = false; + scene_state.ubo.ssao_enabled = false; + + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG env_bg = environment_get_background(p_render_data->environment); + RS::EnvironmentAmbientSource ambient_src = environment_get_ambient_source(p_render_data->environment); + + float bg_energy = environment_get_bg_energy(p_render_data->environment); + scene_state.ubo.ambient_light_color_energy[3] = bg_energy; + + scene_state.ubo.ambient_color_sky_mix = environment_get_ambient_sky_contribution(p_render_data->environment); + + //ambient + if (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && (env_bg == RS::ENV_BG_CLEAR_COLOR || env_bg == RS::ENV_BG_COLOR)) { + Color color = env_bg == RS::ENV_BG_CLEAR_COLOR ? p_default_bg_color : environment_get_bg_color(p_render_data->environment); + color = color.to_linear(); + + scene_state.ubo.ambient_light_color_energy[0] = color.r * bg_energy; + scene_state.ubo.ambient_light_color_energy[1] = color.g * bg_energy; + scene_state.ubo.ambient_light_color_energy[2] = color.b * bg_energy; + scene_state.ubo.use_ambient_light = true; + scene_state.ubo.use_ambient_cubemap = false; + } else { + float energy = environment_get_ambient_light_energy(p_render_data->environment); + Color color = environment_get_ambient_light_color(p_render_data->environment); + color = color.to_linear(); + scene_state.ubo.ambient_light_color_energy[0] = color.r * energy; + scene_state.ubo.ambient_light_color_energy[1] = color.g * energy; + scene_state.ubo.ambient_light_color_energy[2] = color.b * energy; + + Basis sky_transform = environment_get_sky_orientation(p_render_data->environment); + sky_transform = sky_transform.inverse() * p_render_data->cam_transform.basis; + RendererStorageRD::store_transform_3x3(sky_transform, scene_state.ubo.radiance_inverse_xform); + + scene_state.ubo.use_ambient_cubemap = (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ambient_src == RS::ENV_AMBIENT_SOURCE_SKY; + scene_state.ubo.use_ambient_light = scene_state.ubo.use_ambient_cubemap || ambient_src == RS::ENV_AMBIENT_SOURCE_COLOR; + } + + //specular + RS::EnvironmentReflectionSource ref_src = environment_get_reflection_source(p_render_data->environment); + if ((ref_src == RS::ENV_REFLECTION_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ref_src == RS::ENV_REFLECTION_SOURCE_SKY) { + scene_state.ubo.use_reflection_cubemap = true; + } else { + scene_state.ubo.use_reflection_cubemap = false; + } + + scene_state.ubo.ssao_enabled = p_opaque_render_buffers && environment_is_ssao_enabled(p_render_data->environment); + scene_state.ubo.ssao_ao_affect = environment_get_ssao_ao_affect(p_render_data->environment); + scene_state.ubo.ssao_light_affect = environment_get_ssao_light_affect(p_render_data->environment); + + Color ao_color = environment_get_ao_color(p_render_data->environment).to_linear(); + scene_state.ubo.ao_color[0] = ao_color.r; + scene_state.ubo.ao_color[1] = ao_color.g; + scene_state.ubo.ao_color[2] = ao_color.b; + scene_state.ubo.ao_color[3] = ao_color.a; + + scene_state.ubo.fog_enabled = environment_is_fog_enabled(p_render_data->environment); + scene_state.ubo.fog_density = environment_get_fog_density(p_render_data->environment); + scene_state.ubo.fog_height = environment_get_fog_height(p_render_data->environment); + scene_state.ubo.fog_height_density = environment_get_fog_height_density(p_render_data->environment); + if (scene_state.ubo.fog_height_density >= 0.0001) { + scene_state.ubo.fog_height_density = 1.0 / scene_state.ubo.fog_height_density; + } + scene_state.ubo.fog_aerial_perspective = environment_get_fog_aerial_perspective(p_render_data->environment); + + Color fog_color = environment_get_fog_light_color(p_render_data->environment).to_linear(); + float fog_energy = environment_get_fog_light_energy(p_render_data->environment); + + scene_state.ubo.fog_light_color[0] = fog_color.r * fog_energy; + scene_state.ubo.fog_light_color[1] = fog_color.g * fog_energy; + scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy; + + scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_render_data->environment); + + } else { + if (p_render_data->reflection_probe.is_valid() && storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + scene_state.ubo.use_ambient_light = false; + } else { + scene_state.ubo.use_ambient_light = true; + Color clear_color = p_default_bg_color; + clear_color = clear_color.to_linear(); + scene_state.ubo.ambient_light_color_energy[0] = clear_color.r; + scene_state.ubo.ambient_light_color_energy[1] = clear_color.g; + scene_state.ubo.ambient_light_color_energy[2] = clear_color.b; + scene_state.ubo.ambient_light_color_energy[3] = 1.0; + } + + scene_state.ubo.use_ambient_cubemap = false; + scene_state.ubo.use_reflection_cubemap = false; + scene_state.ubo.ssao_enabled = false; + } + + scene_state.ubo.roughness_limiter_enabled = p_opaque_render_buffers && screen_space_roughness_limiter_is_active(); + scene_state.ubo.roughness_limiter_amount = screen_space_roughness_limiter_get_amount(); + scene_state.ubo.roughness_limiter_limit = screen_space_roughness_limiter_get_limit(); + + if (p_index >= (int)scene_state.uniform_buffers.size()) { + uint32_t from = scene_state.uniform_buffers.size(); + scene_state.uniform_buffers.resize(p_index + 1); + render_pass_uniform_sets.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); + } + } + RD::get_singleton()->buffer_update(scene_state.uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); +} + +void RenderForwardMobile::_fill_element_info(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements) { + RenderList *rl = &render_list[p_render_list]; + uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); + + rl->element_info.resize(p_offset + element_total); + + for (uint32_t i = 0; i < element_total; i++) { + GeometryInstanceSurfaceDataCache *surface = rl->elements[i + p_offset]; + RenderElementInfo &element_info = rl->element_info[p_offset + i]; + + element_info.lod_index = surface->lod_index; + element_info.uses_lightmap = surface->sort.uses_lightmap; + } +} + +/// RENDERING /// + +void RenderForwardMobile::_render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + //use template for faster performance (pass mode comparisons are inlined) + + switch (p_params->pass_mode) { + case PASS_MODE_COLOR: { + _render_list_template<PASS_MODE_COLOR>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_COLOR_TRANSPARENT: { + _render_list_template<PASS_MODE_COLOR_TRANSPARENT>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW: { + _render_list_template<PASS_MODE_SHADOW>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW_DP: { + _render_list_template<PASS_MODE_SHADOW_DP>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_MATERIAL: { + _render_list_template<PASS_MODE_DEPTH_MATERIAL>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + } +} + +void RenderForwardMobile::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { + uint32_t render_total = p_params->element_count; + uint32_t total_threads = RendererThreadPool::singleton->thread_work_pool.get_thread_count(); + uint32_t render_from = p_thread * render_total / total_threads; + uint32_t render_to = (p_thread + 1 == total_threads) ? render_total : ((p_thread + 1) * render_total / total_threads); + _render_list(thread_draw_lists[p_thread], p_params->framebuffer_format, p_params, render_from, render_to); +} + +void RenderForwardMobile::_render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_framebuffer); + p_params->framebuffer_format = fb_format; + + if ((uint32_t)p_params->element_count > render_list_thread_threshold && false) { // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(p_framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, p_params); + RD::get_singleton()->draw_list_end(p_params->barrier); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + _render_list(draw_list, fb_format, p_params, 0, p_params->element_count); + RD::get_singleton()->draw_list_end(p_params->barrier); + } +} + +void RenderForwardMobile::_fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance) { + // first zero out our indices + + p_push_constant->omni_lights[0] = 0xFFFF; + p_push_constant->omni_lights[1] = 0xFFFF; + + p_push_constant->spot_lights[0] = 0xFFFF; + p_push_constant->spot_lights[1] = 0xFFFF; + + p_push_constant->decals[0] = 0xFFFF; + p_push_constant->decals[1] = 0xFFFF; + + p_push_constant->reflection_probes[0] = 0xFFFF; + p_push_constant->reflection_probes[1] = 0xFFFF; + + if (p_instance->omni_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_OMNI_LIGHTS; + } + if (p_instance->spot_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_SPOT_LIGHTS; + } + if (p_instance->reflection_probe_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_REFLECTION_PROBES; + } + if (p_instance->decals_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_DECALS; + } + + for (uint32_t i = 0; i < MAX_RDL_CULL; i++) { + uint32_t ofs = i < 4 ? 0 : 1; + uint32_t shift = (i & 0x3) << 3; + uint32_t mask = ~(0xFF << shift); + if (i < p_instance->omni_light_count) { + p_push_constant->omni_lights[ofs] &= mask; + p_push_constant->omni_lights[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_OMNI_LIGHT].map[p_instance->omni_lights[i]]) << shift; + } + if (i < p_instance->spot_light_count) { + p_push_constant->spot_lights[ofs] &= mask; + p_push_constant->spot_lights[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_SPOT_LIGHT].map[p_instance->spot_lights[i]]) << shift; + } + if (i < p_instance->decals_count) { + p_push_constant->decals[ofs] &= mask; + p_push_constant->decals[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_DECAL].map[p_instance->decals[i]]) << shift; + } + if (i < p_instance->reflection_probe_count) { + p_push_constant->reflection_probes[ofs] &= mask; + p_push_constant->reflection_probes[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_REFLECTION_PROBE].map[p_instance->reflection_probes[i]]) << shift; + } + } +} + +template <RenderForwardMobile::PassMode p_pass_mode> +void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + RD::DrawListID draw_list = p_draw_list; + RD::FramebufferFormatID framebuffer_format = p_framebuffer_Format; + + //global scope bindings + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_base_uniform_set, SCENE_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_params->render_pass_uniform_set, RENDER_PASS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, scene_shader.default_vec4_xform_uniform_set, TRANSFORMS_UNIFORM_SET); + + RID prev_material_uniform_set; + + RID prev_vertex_array_rd; + RID prev_index_array_rd; + RID prev_pipeline_rd; + RID prev_xforms_uniform_set; + + bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); + + for (uint32_t i = p_from_element; i < p_to_element; i++) { + const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; + const RenderElementInfo &element_info = p_params->element_info[i]; + const GeometryInstanceForwardMobile *inst = surf->owner; + + uint32_t base_spec_constants = p_params->spec_constant_base_flags; + + // GeometryInstanceForwardMobile::PushConstant push_constant = inst->push_constant; + GeometryInstanceForwardMobile::PushConstant push_constant; + + if (inst->store_transform_cache) { + RendererStorageRD::store_transform(inst->transform, push_constant.transform); + } else { + RendererStorageRD::store_transform(Transform3D(), push_constant.transform); + } + + push_constant.flags = inst->flags_cache; + push_constant.gi_offset = inst->gi_offset_cache; + push_constant.layer_mask = inst->layer_mask; + push_constant.instance_uniforms_ofs = uint32_t(inst->shader_parameters_offset); + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + // abuse lightmap_uv_scale[0] here, should not be needed here + push_constant.lightmap_uv_scale[0] = p_params->uv_offset.x; + push_constant.lightmap_uv_scale[1] = p_params->uv_offset.y; + } else { + push_constant.lightmap_uv_scale[0] = inst->lightmap_uv_scale.position.x; + push_constant.lightmap_uv_scale[1] = inst->lightmap_uv_scale.position.y; + push_constant.lightmap_uv_scale[2] = inst->lightmap_uv_scale.size.x; + push_constant.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; + }; + + RID material_uniform_set; + SceneShaderForwardMobile::ShaderData *shader; + void *mesh_surface; + + if (shadow_pass) { + material_uniform_set = surf->material_uniform_set_shadow; + shader = surf->shader_shadow; + mesh_surface = surf->surface_shadow; + + } else { + if (inst->use_projector) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_PROJECTOR; + } + if (inst->use_soft_shadow) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_SOFT_SHADOWS; + } + _fill_push_constant_instance_indices(&push_constant, base_spec_constants, inst); + +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif + mesh_surface = surf->surface; + } + + if (!mesh_surface) { + continue; + } + + //find cull variant + SceneShaderForwardMobile::ShaderData::CullVariant cull_variant; + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL || ((p_params->pass_mode == PASS_MODE_SHADOW || p_params->pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { + cull_variant = SceneShaderForwardMobile::ShaderData::CULL_VARIANT_DOUBLE_SIDED; + } else { + bool mirror = surf->owner->mirror; + if (p_params->reverse_cull) { + mirror = !mirror; + } + cull_variant = mirror ? SceneShaderForwardMobile::ShaderData::CULL_VARIANT_REVERSED : SceneShaderForwardMobile::ShaderData::CULL_VARIANT_NORMAL; + } + + RS::PrimitiveType primitive = surf->primitive; + RID xforms_uniform_set = surf->owner->transforms_uniform_set; + + SceneShaderForwardMobile::ShaderVersion shader_version = SceneShaderForwardMobile::SHADER_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. + + switch (p_params->pass_mode) { + case PASS_MODE_COLOR: + case PASS_MODE_COLOR_TRANSPARENT: { + if (element_info.uses_lightmap) { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_LIGHTMAP_COLOR_PASS; + } else { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_COLOR_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_COLOR_PASS; + } + } break; + case PASS_MODE_SHADOW: { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS; + } break; + case PASS_MODE_SHADOW_DP: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for shadow DP pass"); + shader_version = SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS_DP; + } break; + case PASS_MODE_DEPTH_MATERIAL: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for material pass"); + shader_version = SceneShaderForwardMobile::SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL; + } break; + } + + PipelineCacheRD *pipeline = nullptr; + + pipeline = &shader->pipelines[cull_variant][primitive][shader_version]; + + RD::VertexFormatID vertex_format = -1; + RID vertex_array_rd; + RID index_array_rd; + + //skeleton and blend shape + if (surf->owner->mesh_instance.is_valid()) { + storage->mesh_instance_surface_get_vertex_arrays_and_format(surf->owner->mesh_instance, surf->surface_index, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } else { + storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } + + index_array_rd = storage->mesh_surface_get_index_array(mesh_surface, element_info.lod_index); + + if (prev_vertex_array_rd != vertex_array_rd) { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, vertex_array_rd); + prev_vertex_array_rd = vertex_array_rd; + } + + if (prev_index_array_rd != index_array_rd) { + if (index_array_rd.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array_rd); + } + prev_index_array_rd = index_array_rd; + } + + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, p_params->subpass, base_spec_constants); + + if (pipeline_rd != prev_pipeline_rd) { + // checking with prev shader does not make so much sense, as + // the pipeline may still be different. + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, pipeline_rd); + prev_pipeline_rd = pipeline_rd; + } + + if (xforms_uniform_set.is_valid() && prev_xforms_uniform_set != xforms_uniform_set) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, xforms_uniform_set, TRANSFORMS_UNIFORM_SET); + prev_xforms_uniform_set = xforms_uniform_set; + } + + if (material_uniform_set != prev_material_uniform_set) { + // Update uniform set. + if (RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); + } + + prev_material_uniform_set = material_uniform_set; + } + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(GeometryInstanceForwardMobile::PushConstant)); + + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : 1; + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS) { + instance_count /= surf->owner->trail_steps; + } + + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + } +} + +/* Geometry instance */ + +RendererSceneRender::GeometryInstance *RenderForwardMobile::geometry_instance_create(RID p_base) { + RS::InstanceType type = storage->get_base_type(p_base); + ERR_FAIL_COND_V(!((1 << type) & RS::INSTANCE_GEOMETRY_MASK), nullptr); + + GeometryInstanceForwardMobile *ginstance = geometry_instance_alloc.alloc(); + ginstance->data = memnew(GeometryInstanceForwardMobile::Data); + + ginstance->data->base = p_base; + ginstance->data->base_type = type; + + _geometry_instance_mark_dirty(ginstance); + + return ginstance; +} + +void RenderForwardMobile::geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->skeleton = p_skeleton; + + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} + +void RenderForwardMobile::geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->material_override = p_override; + + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} + +void RenderForwardMobile::geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->surface_materials = p_materials; + + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} + +void RenderForwardMobile::geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->mesh_instance = p_mesh_instance; + + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->transform = p_transform; + ginstance->mirror = p_transform.basis.determinant() < 0; + ginstance->data->aabb = p_aabb; + ginstance->transformed_aabb = p_transformed_aabb; + + Vector3 model_scale_vec = p_transform.basis.get_scale_abs(); + // handle non uniform scale here + + float max_scale = MAX(model_scale_vec.x, MAX(model_scale_vec.y, model_scale_vec.z)); + float min_scale = MIN(model_scale_vec.x, MIN(model_scale_vec.y, model_scale_vec.z)); + ginstance->non_uniform_scale = max_scale >= 0.0 && (min_scale / max_scale) < 0.9; + + ginstance->lod_model_scale = max_scale; +} + +void RenderForwardMobile::geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->layer_mask = p_layer_mask; +} + +void RenderForwardMobile::geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->lod_bias = p_lod_bias; +} + +void RenderForwardMobile::geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->use_baked_light = p_enable; + + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) { + // !BAS! do we support this in mobile? + // GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + // ERR_FAIL_COND(!ginstance); + // ginstance->data->use_dynamic_gi = p_enable; + // _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->lightmap_instance = p_lightmap_instance; + ginstance->lightmap_uv_scale = p_lightmap_uv_scale; + ginstance->lightmap_slice_index = p_lightmap_slice_index; + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (p_sh9) { + if (ginstance->lightmap_sh == nullptr) { + ginstance->lightmap_sh = geometry_instance_lightmap_sh.alloc(); + } + + memcpy(ginstance->lightmap_sh->sh, p_sh9, sizeof(Color) * 9); + } else { + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + ginstance->lightmap_sh = nullptr; + } + } + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->shader_parameters_offset = p_offset; + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardMobile::geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->data->cast_double_sided_shadows = p_enable; + _geometry_instance_mark_dirty(ginstance); +} + +Transform3D RenderForwardMobile::geometry_instance_get_transform(GeometryInstance *p_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_instance); + ERR_FAIL_COND_V(!ginstance, Transform3D()); + return ginstance->transform; +} + +AABB RenderForwardMobile::geometry_instance_get_aabb(GeometryInstance *p_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_instance); + ERR_FAIL_COND_V(!ginstance, AABB()); + return ginstance->data->aabb; +} + +void RenderForwardMobile::geometry_instance_free(GeometryInstance *p_geometry_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + } + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + memdelete(ginstance->data); + geometry_instance_alloc.free(ginstance); +} + +uint32_t RenderForwardMobile::geometry_instance_get_pair_mask() { + return ((1 << RS::INSTANCE_LIGHT) + (1 << RS::INSTANCE_REFLECTION_PROBE) + (1 << RS::INSTANCE_DECAL)); +} + +void RenderForwardMobile::geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->omni_light_count = 0; + ginstance->spot_light_count = 0; + + for (uint32_t i = 0; i < p_light_instance_count; i++) { + RS::LightType type = light_instance_get_type(p_light_instances[i]); + switch (type) { + case RS::LIGHT_OMNI: { + if (ginstance->omni_light_count < (uint32_t)MAX_RDL_CULL) { + ginstance->omni_lights[ginstance->omni_light_count] = light_instance_get_forward_id(p_light_instances[i]); + ginstance->omni_light_count++; + } + } break; + case RS::LIGHT_SPOT: { + if (ginstance->spot_light_count < (uint32_t)MAX_RDL_CULL) { + ginstance->spot_lights[ginstance->spot_light_count] = light_instance_get_forward_id(p_light_instances[i]); + ginstance->spot_light_count++; + } + } break; + default: + break; + } + } +} + +void RenderForwardMobile::geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->reflection_probe_count = p_reflection_probe_instance_count < (uint32_t)MAX_RDL_CULL ? p_reflection_probe_instance_count : (uint32_t)MAX_RDL_CULL; + for (uint32_t i = 0; i < ginstance->reflection_probe_count; i++) { + ginstance->reflection_probes[i] = reflection_probe_instance_get_forward_id(p_reflection_probe_instances[i]); + } +} + +void RenderForwardMobile::geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->decals_count = p_decal_instance_count < (uint32_t)MAX_RDL_CULL ? p_decal_instance_count : (uint32_t)MAX_RDL_CULL; + for (uint32_t i = 0; i < ginstance->decals_count; i++) { + ginstance->decals[i] = decal_instance_get_forward_id(p_decal_instances[i]); + } +} + +void RenderForwardMobile::geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) { + // We do not have this here! +} + +void RenderForwardMobile::geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->use_projector = p_projector; + ginstance->use_soft_shadow = p_softshadow; +} + +void RenderForwardMobile::_geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + if (ginstance->dirty_list_element.in_list()) { + return; + } + + //clear surface caches + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + + ginstance->surface_caches = nullptr; + + geometry_instance_dirty_list.add(&ginstance->dirty_list_element); +} + +void RenderForwardMobile::_geometry_instance_add_surface_with_material(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { + bool has_read_screen_alpha = p_material->shader_data->uses_screen_texture || p_material->shader_data->uses_depth_texture || p_material->shader_data->uses_normal_texture; + bool has_base_alpha = (p_material->shader_data->uses_alpha || has_read_screen_alpha); + bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; + bool has_alpha = has_base_alpha || has_blend_alpha; + + uint32_t flags = 0; + + if (p_material->shader_data->uses_sss) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING; + } + + if (p_material->shader_data->uses_screen_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE; + } + + if (p_material->shader_data->uses_depth_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE; + } + + if (p_material->shader_data->uses_normal_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE; + } + + if (ginstance->data->cast_double_sided_shadows) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS; + } + + if (has_alpha || has_read_screen_alpha || p_material->shader_data->depth_draw == SceneShaderForwardMobile::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardMobile::ShaderData::DEPTH_TEST_DISABLED) { + //material is only meant for alpha pass + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA; + if (p_material->shader_data->uses_depth_pre_pass && !(p_material->shader_data->depth_draw == SceneShaderForwardMobile::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardMobile::ShaderData::DEPTH_TEST_DISABLED)) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + } else { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + + if (p_material->shader_data->uses_particle_trails) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS; + } + + SceneShaderForwardMobile::MaterialData *material_shadow = nullptr; + void *surface_shadow = nullptr; + if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; + material_shadow = (SceneShaderForwardMobile::MaterialData *)storage->material_get_data(scene_shader.default_material, RendererStorageRD::SHADER_TYPE_3D); + + RID shadow_mesh = storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = storage->mesh_get_surface(shadow_mesh, p_surface); + } + + } else { + material_shadow = p_material; + } + + GeometryInstanceSurfaceDataCache *sdcache = geometry_instance_surface_alloc.alloc(); + + sdcache->flags = flags; + + sdcache->shader = p_material->shader_data; + sdcache->material_uniform_set = p_material->uniform_set; + sdcache->surface = storage->mesh_get_surface(p_mesh, p_surface); + sdcache->primitive = storage->mesh_surface_get_primitive(sdcache->surface); + sdcache->surface_index = p_surface; + + if (ginstance->data->dirty_dependencies) { + storage->base_update_dependency(p_mesh, &ginstance->data->dependency_tracker); + } + + //shadow + sdcache->shader_shadow = material_shadow->shader_data; + sdcache->material_uniform_set_shadow = material_shadow->uniform_set; + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; + + sdcache->owner = ginstance; + + sdcache->next = ginstance->surface_caches; + ginstance->surface_caches = sdcache; + + //sortkey + + sdcache->sort.sort_key1 = 0; + sdcache->sort.sort_key2 = 0; + + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0x0000FFFF; + sdcache->sort.material_id_hi = p_material_id >> 16; + sdcache->sort.shader_id = p_shader_id; + sdcache->sort.geometry_id = p_mesh.get_local_index(); + // sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; + sdcache->sort.priority = p_material->priority; +} + +void RenderForwardMobile::_geometry_instance_add_surface(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { + RID m_src; + + m_src = ginstance->data->material_override.is_valid() ? ginstance->data->material_override : p_material; + + SceneShaderForwardMobile::MaterialData *material = nullptr; + + if (m_src.is_valid()) { + material = (SceneShaderForwardMobile::MaterialData *)storage->material_get_data(m_src, RendererStorageRD::SHADER_TYPE_3D); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (material) { + if (ginstance->data->dirty_dependencies) { + storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + } else { + material = (SceneShaderForwardMobile::MaterialData *)storage->material_get_data(scene_shader.default_material, RendererStorageRD::SHADER_TYPE_3D); + m_src = scene_shader.default_material; + } + + ERR_FAIL_COND(!material); + + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, m_src.get_local_index(), storage->material_get_shader_id(m_src), p_mesh); + + while (material->next_pass.is_valid()) { + RID next_pass = material->next_pass; + material = (SceneShaderForwardMobile::MaterialData *)storage->material_get_data(next_pass, RendererStorageRD::SHADER_TYPE_3D); + if (!material || !material->shader_data->valid) { + break; + } + if (ginstance->data->dirty_dependencies) { + storage->material_update_dependency(next_pass, &ginstance->data->dependency_tracker); + } + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, next_pass.get_local_index(), storage->material_get_shader_id(next_pass), p_mesh); + } +} + +void RenderForwardMobile::_geometry_instance_update(GeometryInstance *p_geometry_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_begin(); + } + + //add geometry for drawing + switch (ginstance->data->base_type) { + case RS::INSTANCE_MESH: { + const RID *materials = nullptr; + uint32_t surface_count; + RID mesh = ginstance->data->base; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + //if no materials, no surfaces. + const RID *inst_materials = ginstance->data->surface_materials.ptr(); + uint32_t surf_mat_count = ginstance->data->surface_materials.size(); + + for (uint32_t j = 0; j < surface_count; j++) { + RID material = (j < surf_mat_count && inst_materials[j].is_valid()) ? inst_materials[j] : materials[j]; + _geometry_instance_add_surface(ginstance, j, material, mesh); + } + } + + ginstance->instance_count = 1; + + } break; + + case RS::INSTANCE_MULTIMESH: { + RID mesh = storage->multimesh_get_mesh(ginstance->data->base); + if (mesh.is_valid()) { + const RID *materials = nullptr; + uint32_t surface_count; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t j = 0; j < surface_count; j++) { + _geometry_instance_add_surface(ginstance, j, materials[j], mesh); + } + } + + ginstance->instance_count = storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + + } break; +#if 0 + case RS::INSTANCE_IMMEDIATE: { + RasterizerStorageGLES3::Immediate *immediate = storage->immediate_owner.getornull(inst->base); + ERR_CONTINUE(!immediate); + + _add_geometry(immediate, inst, nullptr, -1, p_depth_pass, p_shadow_pass); + + } break; +#endif + case RS::INSTANCE_PARTICLES: { + int draw_passes = storage->particles_get_draw_passes(ginstance->data->base); + + for (int j = 0; j < draw_passes; j++) { + RID mesh = storage->particles_get_draw_pass_mesh(ginstance->data->base, j); + if (!mesh.is_valid()) { + continue; + } + + const RID *materials = nullptr; + uint32_t surface_count; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t k = 0; k < surface_count; k++) { + _geometry_instance_add_surface(ginstance, k, materials[k], mesh); + } + } + } + + ginstance->instance_count = storage->particles_get_amount(ginstance->data->base, ginstance->trail_steps); + + } break; + + default: { + } + } + + //Fill push constant + + bool store_transform = true; + ginstance->base_flags = 0; + + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + if (storage->multimesh_get_transform_format(ginstance->data->base) == RS::MULTIMESH_TRANSFORM_2D) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + } + if (storage->multimesh_uses_colors(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + } + if (storage->multimesh_uses_custom_data(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + } + + ginstance->transforms_uniform_set = storage->multimesh_get_3d_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_PARTICLES) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + if (false) { // 2D particles + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + } + + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + + //for particles, stride is the trail size + ginstance->base_flags |= (ginstance->trail_steps << INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT); + + if (!storage->particles_is_using_local_coords(ginstance->data->base)) { + store_transform = false; + } + ginstance->transforms_uniform_set = storage->particles_get_instance_buffer_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_MESH) { + if (storage->skeleton_is_valid(ginstance->data->skeleton)) { + ginstance->transforms_uniform_set = storage->skeleton_get_3d_uniform_set(ginstance->data->skeleton, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + if (ginstance->data->dirty_dependencies) { + storage->skeleton_update_dependency(ginstance->data->skeleton, &ginstance->data->dependency_tracker); + } + } + } + + ginstance->store_transform_cache = store_transform; + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_end(); + ginstance->data->dirty_dependencies = false; + } + + ginstance->dirty_list_element.remove_from_list(); +} + +void RenderForwardMobile::_update_dirty_geometry_instances() { + while (geometry_instance_dirty_list.first()) { + _geometry_instance_update(geometry_instance_dirty_list.first()->self()); + } +} + +void RenderForwardMobile::_geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker) { + switch (p_notification) { + case RendererStorage::DEPENDENCY_CHANGED_MATERIAL: + case RendererStorage::DEPENDENCY_CHANGED_MESH: + case RendererStorage::DEPENDENCY_CHANGED_PARTICLES: + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH: + case RendererStorage::DEPENDENCY_CHANGED_SKELETON_DATA: { + static_cast<RenderForwardMobile *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); + } break; + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_tracker->userdata); + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->instance_count = static_cast<RenderForwardMobile *>(singleton)->storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + } break; + default: { + //rest of notifications of no interest + } break; + } +} +void RenderForwardMobile::_geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker) { + static_cast<RenderForwardMobile *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); +} + +/* misc */ + +bool RenderForwardMobile::is_dynamic_gi_supported() const { + return false; +} + +bool RenderForwardMobile::is_clustered_enabled() const { + return false; +} + +bool RenderForwardMobile::is_volumetric_supported() const { + return false; +} + +uint32_t RenderForwardMobile::get_max_elements() const { + return 256; +} + +RenderForwardMobile *RenderForwardMobile::singleton = nullptr; + +void RenderForwardMobile::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_USE_MIPMAPS; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + _base_uniforms_changed(); //also need this +} + +RenderForwardMobile::RenderForwardMobile(RendererStorageRD *p_storage) : + RendererSceneRenderRD(p_storage) { + singleton = this; + + sky.set_texture_format(_render_buffers_get_color_format()); + + String defines; + + defines += "\n#define MAX_ROUGHNESS_LOD " + itos(get_roughness_layers() - 1) + ".0\n"; + if (is_using_radiance_cubemap_array()) { + defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; + } + // defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; + + { + //lightmaps + scene_state.max_lightmaps = 2; + defines += "\n#define MAX_LIGHTMAP_TEXTURES " + itos(scene_state.max_lightmaps) + "\n"; + defines += "\n#define MAX_LIGHTMAPS " + itos(scene_state.max_lightmaps) + "\n"; + + scene_state.lightmap_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapData) * scene_state.max_lightmaps); + } + { + //captures + scene_state.max_lightmap_captures = 2048; + scene_state.lightmap_captures = memnew_arr(LightmapCaptureData, scene_state.max_lightmap_captures); + scene_state.lightmap_capture_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapCaptureData) * scene_state.max_lightmap_captures); + } + { + defines += "\n#define MATERIAL_UNIFORM_SET " + itos(MATERIAL_UNIFORM_SET) + "\n"; + } + + scene_shader.init(p_storage, defines); + + // !BAS! maybe we need a mobile version of this setting? + render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); +} + +RenderForwardMobile::~RenderForwardMobile() { + directional_shadow_atlas_set_size(0); + + //clear base uniform set if still valid + for (uint32_t i = 0; i < render_pass_uniform_sets.size(); i++) { + if (render_pass_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[i])) { + RD::get_singleton()->free(render_pass_uniform_sets[i]); + } + } + + { + for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.uniform_buffers[i]); + } + RD::get_singleton()->free(scene_state.lightmap_buffer); + RD::get_singleton()->free(scene_state.lightmap_capture_buffer); + memdelete_arr(scene_state.lightmap_captures); + } +} diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h new file mode 100644 index 0000000000..764d8e80df --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h @@ -0,0 +1,667 @@ +/*************************************************************************/ +/* render_forward_mobile.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDERING_SERVER_SCENE_RENDER_FORWARD_MOBILE_H +#define RENDERING_SERVER_SCENE_RENDER_FORWARD_MOBILE_H + +#include "core/templates/paged_allocator.h" +#include "servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/renderer_storage_rd.h" + +namespace RendererSceneRenderImplementation { + +class RenderForwardMobile : public RendererSceneRenderRD { + friend SceneShaderForwardMobile; + + struct ForwardIDAllocator { + LocalVector<bool> allocations; + LocalVector<uint8_t> map; + }; + + ForwardIDAllocator forward_id_allocators[FORWARD_ID_MAX]; + + virtual ForwardID _allocate_forward_id(ForwardIDType p_type) override; + virtual void _free_forward_id(ForwardIDType p_type, ForwardID p_id) override; + virtual void _map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) override; + virtual bool _uses_forward_ids() const override { return true; } + +protected: + /* Scene Shader */ + + enum { + SCENE_UNIFORM_SET = 0, + RENDER_PASS_UNIFORM_SET = 1, + TRANSFORMS_UNIFORM_SET = 2, + MATERIAL_UNIFORM_SET = 3 + }; + + enum { + + SPEC_CONSTANT_USING_PROJECTOR = 0, + SPEC_CONSTANT_USING_SOFT_SHADOWS = 1, + SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS = 2, + + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 3, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 4, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 5, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 6, + + SPEC_CONSTANT_DECAL_USE_MIPMAPS = 7, + SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS = 8, + + SPEC_CONSTANT_DISABLE_OMNI_LIGHTS = 9, + SPEC_CONSTANT_DISABLE_SPOT_LIGHTS = 10, + SPEC_CONSTANT_DISABLE_REFLECTION_PROBES = 11, + SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS = 12, + + SPEC_CONSTANT_DISABLE_DECALS = 13, + SPEC_CONSTANT_DISABLE_FOG = 14, + + }; + + enum { + MAX_LIGHTMAPS = 8, + MAX_RDL_CULL = 8, // maximum number of reflection probes, decals or lights we can cull per geometry instance + INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 + }; + + enum RenderListType { + RENDER_LIST_OPAQUE, //used for opaque objects + RENDER_LIST_ALPHA, //used for transparent objects + RENDER_LIST_SECONDARY, //used for shadows and other objects + RENDER_LIST_MAX + }; + + /* Scene Shader */ + + SceneShaderForwardMobile scene_shader; + + /* Render Buffer */ + + // We can have: + // - 4 subpasses combining the full render cycle + // - 3 subpasses + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + // - 2 subpasses + 1 normal pass for transparent + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + enum RenderBufferMobileFramebufferConfigType { + FB_CONFIG_ONE_PASS, // Single pass frame buffer for alpha pass + FB_CONFIG_TWO_SUBPASSES, // Opaque + Sky sub pass + FB_CONFIG_THREE_SUBPASSES, // Opaque + Sky + Alpha sub pass + FB_CONFIG_FOUR_SUBPASSES, // Opaque + Sky + Alpha sub pass + Tonemap pass + FB_CONFIG_MAX + }; + + struct RenderBufferDataForwardMobile : public RenderBufferData { + RID color; + RID depth; + // RID normal_roughness_buffer; + + RS::ViewportMSAA msaa; + RD::TextureSamples texture_samples; + + RID color_msaa; + RID depth_msaa; + // RID normal_roughness_buffer_msaa; + + RID color_fbs[FB_CONFIG_MAX]; + int width, height; + uint32_t view_count; + + void clear(); + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); + + ~RenderBufferDataForwardMobile(); + }; + + virtual RenderBufferData *_create_render_buffer_data() override; + + /* Rendering */ + + enum PassMode { + PASS_MODE_COLOR, + // PASS_MODE_COLOR_SPECULAR, + PASS_MODE_COLOR_TRANSPARENT, + PASS_MODE_SHADOW, + PASS_MODE_SHADOW_DP, + // PASS_MODE_DEPTH, + // PASS_MODE_DEPTH_NORMAL_ROUGHNESS, + // PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI, + PASS_MODE_DEPTH_MATERIAL, + // PASS_MODE_SDF, + }; + + struct GeometryInstanceForwardMobile; + struct GeometryInstanceSurfaceDataCache; + struct RenderElementInfo; + + struct RenderListParameters { + GeometryInstanceSurfaceDataCache **elements = nullptr; + RenderElementInfo *element_info = nullptr; + int element_count = 0; + bool reverse_cull = false; + PassMode pass_mode = PASS_MODE_COLOR; + // bool no_gi = false; + uint32_t view_count = 1; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + Plane lod_plane; + uint32_t spec_constant_base_flags = 0; + float lod_distance_multiplier = 0.0; + float screen_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + uint32_t subpass = 0; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, RID p_render_pass_uniform_set, uint32_t p_spec_constant_base_flags = 0, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_view_count = 1, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + elements = p_elements; + element_info = p_element_info; + element_count = p_element_count; + reverse_cull = p_reverse_cull; + pass_mode = p_pass_mode; + // no_gi = p_no_gi; + view_count = p_view_count; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_plane = p_lod_plane; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_lod_threshold = p_screen_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + spec_constant_base_flags = p_spec_constant_base_flags; + } + }; + + virtual RD::DataFormat _render_buffers_get_color_format() override; + virtual bool _render_buffers_can_be_storage() override; + + RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; + + virtual void _render_shadow_begin() override; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) override; + virtual void _render_shadow_process() override; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) override; + + virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) override; + + uint64_t lightmap_texture_array_version = 0xFFFFFFFF; + + virtual void _base_uniforms_changed() override; + void _update_render_base_uniform_set(); + virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) override; + + void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append = false); + void _fill_element_info(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1); + // void _update_instance_data_buffer(RenderListType p_render_list); + + static RenderForwardMobile *singleton; + + void _setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); + void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform); + + RID render_base_uniform_set; + LocalVector<RID> render_pass_uniform_sets; + + /* Light map */ + + struct LightmapData { + float normal_xform[12]; + }; + + struct LightmapCaptureData { + float sh[9 * 4]; + }; + + /* Scene state */ + + struct SceneState { + // This struct is loaded into Set 1 - Binding 0, populated at start of rendering a frame, must match with shader code + struct UBO { + float projection_matrix[16]; + float inv_projection_matrix[16]; + float camera_matrix[16]; + float inv_camera_matrix[16]; + + float projection_matrix_view[RendererSceneRender::MAX_RENDER_VIEWS][16]; + float inv_projection_matrix_view[RendererSceneRender::MAX_RENDER_VIEWS][16]; + + float viewport_size[2]; + float screen_pixel_size[2]; + + float directional_penumbra_shadow_kernel[128]; //32 vec4s + float directional_soft_shadow_kernel[128]; + float penumbra_shadow_kernel[128]; + float soft_shadow_kernel[128]; + + float ambient_light_color_energy[4]; + + float ambient_color_sky_mix; + uint32_t use_ambient_light; + uint32_t use_ambient_cubemap; + uint32_t use_reflection_cubemap; + + float radiance_inverse_xform[12]; + + float shadow_atlas_pixel_size[2]; + float directional_shadow_pixel_size[2]; + + uint32_t directional_light_count; + float dual_paraboloid_side; + float z_far; + float z_near; + + uint32_t ssao_enabled; + float ssao_light_affect; + float ssao_ao_affect; + uint32_t roughness_limiter_enabled; + + float roughness_limiter_amount; + float roughness_limiter_limit; + uint32_t roughness_limiter_pad[2]; + + float ao_color[4]; + + // Fog + uint32_t fog_enabled; + float fog_density; + float fog_height; + float fog_height_density; + + float fog_light_color[3]; + float fog_sun_scatter; + + float fog_aerial_perspective; + uint32_t material_uv2_mode; + + float time; + float reflection_multiplier; + + uint32_t pancake_shadows; + uint32_t pad1; + uint32_t pad2; + uint32_t pad3; + }; + + UBO ubo; + + LocalVector<RID> uniform_buffers; + + // !BAS! We need to change lightmaps, we're not going to do this with a buffer but pushing the used lightmap in + LightmapData lightmaps[MAX_LIGHTMAPS]; + RID lightmap_ids[MAX_LIGHTMAPS]; + bool lightmap_has_sh[MAX_LIGHTMAPS]; + uint32_t lightmaps_used = 0; + uint32_t max_lightmaps; + RID lightmap_buffer; + + LightmapCaptureData *lightmap_captures; + uint32_t max_lightmap_captures; + RID lightmap_capture_buffer; + + bool used_screen_texture = false; + bool used_normal_texture = false; + bool used_depth_texture = false; + bool used_sss = false; + + struct ShadowPass { + uint32_t element_from; + uint32_t element_count; + bool flip_cull; + PassMode pass_mode; + + RID rp_uniform_set; + Plane camera_plane; + float lod_distance_multiplier; + float screen_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; + } scene_state; + + /* Render List */ + + // !BAS! Render list can probably be reused between clustered and mobile? + struct RenderList { + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; + + void clear() { + elements.clear(); + element_info.clear(); + } + + //should eventually be replaced by radix + + struct SortByKey { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.sort_key2 == B->sort.sort_key2) ? (A->sort.sort_key1 < B->sort.sort_key1) : (A->sort.sort_key2 < B->sort.sort_key2); + } + }; + + void sort_by_key() { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + void sort_by_key_range(uint32_t p_from, uint32_t p_size) { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr() + p_from, p_size); + } + + struct SortByDepth { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->owner->depth < B->owner->depth); + } + }; + + void sort_by_depth() { //used for shadows + + SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + struct SortByReverseDepthAndPriority { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.priority == B->sort.priority) ? (A->owner->depth > B->owner->depth) : (A->sort.priority < B->sort.priority); + } + }; + + void sort_by_reverse_depth_and_priority() { //used for alpha + + SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { + elements.push_back(p_element); + } + }; + + struct RenderElementInfo { + uint32_t uses_lightmap : 1; + uint32_t lod_index : 8; + uint32_t reserved : 23; + }; + + template <PassMode p_pass_mode> + _FORCE_INLINE_ void _render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + LocalVector<RD::DrawListID> thread_draw_lists; + void _render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params); + void _render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); + + uint32_t render_list_thread_threshold = 500; + + RenderList render_list[RENDER_LIST_MAX]; + + /* Geometry instance */ + + // check which ones of these apply, probably all except GI and SDFGI + enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 5, + INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 6, + INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 7, + INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 8, + INSTANCE_DATA_FLAG_USE_LIGHTMAP = 1 << 9, + INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP = 1 << 10, + INSTANCE_DATA_FLAG_USE_VOXEL_GI = 1 << 11, + INSTANCE_DATA_FLAG_MULTIMESH = 1 << 12, + INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D = 1 << 13, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR = 1 << 14, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, + }; + + struct GeometryInstanceLightmapSH { + Color sh[9]; + }; + + // Cached data for drawing surfaces + struct GeometryInstanceSurfaceDataCache { + enum { + FLAG_PASS_DEPTH = 1, + FLAG_PASS_OPAQUE = 2, + FLAG_PASS_ALPHA = 4, + FLAG_PASS_SHADOW = 8, + FLAG_USES_SHARED_SHADOW_MATERIAL = 128, + FLAG_USES_SUBSURFACE_SCATTERING = 2048, + FLAG_USES_SCREEN_TEXTURE = 4096, + FLAG_USES_DEPTH_TEXTURE = 8192, + FLAG_USES_NORMAL_TEXTURE = 16384, + FLAG_USES_DOUBLE_SIDED_SHADOWS = 32768, + FLAG_USES_PARTICLE_TRAILS = 65536, + }; + + union { + struct { + // !BAS! CHECK BITS!!! + + uint64_t surface_index : 10; + uint64_t geometry_id : 32; + uint64_t material_id_low : 16; + + uint64_t material_id_hi : 16; + uint64_t shader_id : 32; + uint64_t uses_lightmap : 4; // sort by lightmap id here, not whether its yes/no (is 4 bits enough?) + uint64_t depth_layer : 4; + uint64_t priority : 8; + + // uint64_t lod_index : 8; // no need to sort on LOD + // uint64_t uses_forward_gi : 1; // no GI here, remove + }; + struct { + uint64_t sort_key1; + uint64_t sort_key2; + }; + } sort; + + RS::PrimitiveType primitive = RS::PRIMITIVE_MAX; + uint32_t flags = 0; + uint32_t surface_index = 0; + uint32_t lod_index = 0; + + void *surface = nullptr; + RID material_uniform_set; + SceneShaderForwardMobile::ShaderData *shader = nullptr; + + void *surface_shadow = nullptr; + RID material_uniform_set_shadow; + SceneShaderForwardMobile::ShaderData *shader_shadow = nullptr; + + GeometryInstanceSurfaceDataCache *next = nullptr; + GeometryInstanceForwardMobile *owner = nullptr; + }; + + // !BAS! GeometryInstanceForwardClustered and GeometryInstanceForwardMobile will likely have a lot of overlap + // may need to think about making this its own class like GeometryInstanceRD? + + struct GeometryInstanceForwardMobile : public GeometryInstance { + // setup + uint32_t base_flags = 0; + uint32_t flags_cache = 0; + + // this structure maps to our push constant in our shader and is populated right before our draw call + struct PushConstant { + float transform[16]; + uint32_t flags; + uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables + uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint32_t layer_mask = 1; + float lightmap_uv_scale[4]; // doubles as uv_offset when needed + uint32_t reflection_probes[2]; // packed reflection probes + uint32_t omni_lights[2]; // packed omni lights + uint32_t spot_lights[2]; // packed spot lights + uint32_t decals[2]; // packed spot lights + }; + + // PushConstant push_constant; // we populate this from our instance data + + //used during rendering + uint32_t layer_mask = 1; + RID transforms_uniform_set; + float depth = 0; + bool mirror = false; + bool use_projector = false; + bool use_soft_shadow = false; + Transform3D transform; + bool store_transform_cache = true; // if true we copy our transform into our PushConstant, if false we use our transforms UBO and clear our PushConstants transform + bool non_uniform_scale = false; + AABB transformed_aabb; //needed for LOD + float lod_bias = 0.0; + float lod_model_scale = 1.0; + int32_t shader_parameters_offset = -1; + uint32_t instance_count = 0; + uint32_t trail_steps = 1; + RID mesh_instance; + + // lightmap + uint32_t gi_offset_cache = 0; // !BAS! Should rename this to lightmap_offset_cache, in forward clustered this was shared between gi and lightmap + uint32_t lightmap_slice_index; + Rect2 lightmap_uv_scale; + RID lightmap_instance; + GeometryInstanceLightmapSH *lightmap_sh = nullptr; + + // culled light info + uint32_t reflection_probe_count = 0; + ForwardID reflection_probes[MAX_RDL_CULL]; + uint32_t omni_light_count = 0; + ForwardID omni_lights[MAX_RDL_CULL]; + uint32_t spot_light_count = 0; + ForwardID spot_lights[MAX_RDL_CULL]; + uint32_t decals_count = 0; + ForwardID decals[MAX_RDL_CULL]; + + GeometryInstanceSurfaceDataCache *surface_caches = nullptr; + + // do we use this? + SelfList<GeometryInstanceForwardMobile> dirty_list_element; + + struct Data { + //data used less often goes into regular heap + RID base; + RS::InstanceType base_type; + + RID skeleton; + Vector<RID> surface_materials; + RID material_override; + AABB aabb; + + bool use_baked_light = false; + bool cast_double_sided_shadows = false; + // bool mirror = false; // !BAS! Does not seem used, we already have this in the main struct + + bool dirty_dependencies = false; + + RendererStorage::DependencyTracker dependency_tracker; + }; + + Data *data = nullptr; + + GeometryInstanceForwardMobile() : + dirty_list_element(this) {} + }; + + _FORCE_INLINE_ void _fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance); + + void _update_shader_quality_settings() override; + +public: + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth) override; + + static void _geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker); + static void _geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker); + + SelfList<GeometryInstanceForwardMobile>::List geometry_instance_dirty_list; + + PagedAllocator<GeometryInstanceForwardMobile> geometry_instance_alloc; + PagedAllocator<GeometryInstanceSurfaceDataCache> geometry_instance_surface_alloc; + PagedAllocator<GeometryInstanceLightmapSH> geometry_instance_lightmap_sh; + + void _geometry_instance_add_surface_with_material(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); + void _geometry_instance_add_surface(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); + void _geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance); + void _geometry_instance_update(GeometryInstance *p_geometry_instance); + void _update_dirty_geometry_instances(); + + virtual GeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) override; + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) override; + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) override; + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) override; + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) override; + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) override; + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) override; + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) override; + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) override; + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) override; + + virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) override; + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) override; + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) override; + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override; + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) override; + virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override; + + virtual void geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) override; + + virtual bool free(RID p_rid) override; + + virtual bool is_dynamic_gi_supported() const override; + virtual bool is_clustered_enabled() const override; + virtual bool is_volumetric_supported() const override; + virtual uint32_t get_max_elements() const override; + + RenderForwardMobile(RendererStorageRD *p_storage); + ~RenderForwardMobile(); +}; +} // namespace RendererSceneRenderImplementation +#endif // !RENDERING_SERVER_SCENE_RENDER_FORWARD_MOBILE_H diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp new file mode 100644 index 0000000000..735014a2ec --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -0,0 +1,765 @@ +/*************************************************************************/ +/* scene_shader_forward_mobile.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "scene_shader_forward_mobile.h" +#include "core/config/project_settings.h" +#include "core/math/math_defs.h" +#include "render_forward_mobile.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" + +using namespace RendererSceneRenderImplementation; + +/* ShaderData */ + +void SceneShaderForwardMobile::ShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_screen_texture = false; + + if (code == String()) { + return; //just invalid, but no error + } + + ShaderCompilerRD::GeneratedCode gen_code; + + int blend_mode = BLEND_MODE_MIX; + int depth_testi = DEPTH_TEST_ENABLED; + int alpha_antialiasing_mode = ALPHA_ANTIALIASING_OFF; + int cull = CULL_BACK; + + uses_point_size = false; + uses_alpha = false; + uses_blend_alpha = false; + uses_depth_pre_pass = false; + uses_discard = false; + uses_roughness = false; + uses_normal = false; + bool wireframe = false; + + unshaded = false; + uses_vertex = false; + uses_sss = false; + uses_transmittance = false; + uses_screen_texture = false; + uses_depth_texture = false; + uses_normal_texture = false; + uses_time = false; + writes_modelview_or_projection = false; + uses_world_coordinates = false; + uses_particle_trails = false; + + int depth_drawi = DEPTH_DRAW_OPAQUE; + + ShaderCompilerRD::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompilerRD::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompilerRD::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompilerRD::STAGE_FRAGMENT; + + actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); + actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); + actions.render_mode_values["blend_sub"] = Pair<int *, int>(&blend_mode, BLEND_MODE_SUB); + actions.render_mode_values["blend_mul"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MUL); + + actions.render_mode_values["alpha_to_coverage"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE); + actions.render_mode_values["alpha_to_coverage_and_one"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE); + + actions.render_mode_values["depth_draw_never"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_DISABLED); + actions.render_mode_values["depth_draw_opaque"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_OPAQUE); + actions.render_mode_values["depth_draw_always"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_ALWAYS); + + actions.render_mode_values["depth_test_disabled"] = Pair<int *, int>(&depth_testi, DEPTH_TEST_DISABLED); + + actions.render_mode_values["cull_disabled"] = Pair<int *, int>(&cull, CULL_DISABLED); + actions.render_mode_values["cull_front"] = Pair<int *, int>(&cull, CULL_FRONT); + actions.render_mode_values["cull_back"] = Pair<int *, int>(&cull, CULL_BACK); + + actions.render_mode_flags["unshaded"] = &unshaded; + actions.render_mode_flags["wireframe"] = &wireframe; + actions.render_mode_flags["particle_trails"] = &uses_particle_trails; + + actions.usage_flag_pointers["ALPHA"] = &uses_alpha; + actions.render_mode_flags["depth_prepass_alpha"] = &uses_depth_pre_pass; + + // actions.usage_flag_pointers["SSS_STRENGTH"] = &uses_sss; + // actions.usage_flag_pointers["SSS_TRANSMITTANCE_DEPTH"] = &uses_transmittance; + + actions.usage_flag_pointers["SCREEN_TEXTURE"] = &uses_screen_texture; + actions.usage_flag_pointers["DEPTH_TEXTURE"] = &uses_depth_texture; + actions.usage_flag_pointers["NORMAL_TEXTURE"] = &uses_normal_texture; + actions.usage_flag_pointers["DISCARD"] = &uses_discard; + actions.usage_flag_pointers["TIME"] = &uses_time; + actions.usage_flag_pointers["ROUGHNESS"] = &uses_roughness; + actions.usage_flag_pointers["NORMAL"] = &uses_normal; + actions.usage_flag_pointers["NORMAL_MAP"] = &uses_normal; + + actions.usage_flag_pointers["POINT_SIZE"] = &uses_point_size; + actions.usage_flag_pointers["POINT_COORD"] = &uses_point_size; + + actions.write_flag_pointers["MODELVIEW_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["PROJECTION_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["VERTEX"] = &uses_vertex; + + actions.uniforms = &uniforms; + + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = shader_singleton->shader.version_create(); + } + + depth_draw = DepthDraw(depth_drawi); + depth_test = DepthTest(depth_testi); + +#if 0 + print_line("**compiling shader:"); + print_line("**defines:\n"); + for (int i = 0; i < gen_code.defines.size(); i++) { + print_line(gen_code.defines[i]); + } + + Map<String, String>::Element * el = gen_code.code.front(); + while (el) { + print_line("\n**code " + el->key() + ":\n" + el->value()); + + el = el->next(); + } + + print_line("\n**uniforms:\n" + gen_code.uniforms); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT]); +#endif + + shader_singleton->shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX], gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT], gen_code.defines); + ERR_FAIL_COND(!shader_singleton->shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //blend modes + + // if any form of Alpha Antialiasing is enabled, set the blend mode to alpha to coverage + if (alpha_antialiasing_mode != ALPHA_ANTIALIASING_OFF) { + blend_mode = BLEND_MODE_ALPHA_TO_COVERAGE; + } + + RD::PipelineColorBlendState::Attachment blend_attachment; + + switch (blend_mode) { + case BLEND_MODE_MIX: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + } break; + case BLEND_MODE_ADD: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_SUB: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.color_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_MUL: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_DST_COLOR; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ZERO; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_DST_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + uses_blend_alpha = true; //force alpha used because of blend + } break; + case BLEND_MODE_ALPHA_TO_COVERAGE: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + } + } + + RD::PipelineColorBlendState blend_state_blend; + blend_state_blend.attachments.push_back(blend_attachment); + RD::PipelineColorBlendState blend_state_opaque = RD::PipelineColorBlendState::create_disabled(1); + RD::PipelineColorBlendState blend_state_opaque_specular = RD::PipelineColorBlendState::create_disabled(2); + RD::PipelineColorBlendState blend_state_depth_normal_roughness = RD::PipelineColorBlendState::create_disabled(1); + RD::PipelineColorBlendState blend_state_depth_normal_roughness_giprobe = RD::PipelineColorBlendState::create_disabled(2); + + //update pipelines + + RD::PipelineDepthStencilState depth_stencil_state; + + if (depth_test != DEPTH_TEST_DISABLED) { + depth_stencil_state.enable_depth_test = true; + depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + depth_stencil_state.enable_depth_write = depth_draw != DEPTH_DRAW_DISABLED ? true : false; + } + + for (int i = 0; i < CULL_VARIANT_MAX; i++) { + RD::PolygonCullMode cull_mode_rd_table[CULL_VARIANT_MAX][3] = { + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_FRONT, RD::POLYGON_CULL_BACK }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_BACK, RD::POLYGON_CULL_FRONT }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED } + }; + + RD::PolygonCullMode cull_mode_rd = cull_mode_rd_table[i][cull]; + + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + RD::RenderPrimitive primitive_rd_table[RS::PRIMITIVE_MAX] = { + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_LINESTRIPS, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, + }; + + RD::RenderPrimitive primitive_rd = uses_point_size ? RD::RENDER_PRIMITIVE_POINTS : primitive_rd_table[j]; + + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + if (!static_cast<SceneShaderForwardMobile *>(singleton)->shader.is_variant_enabled(k)) { + continue; + } + RD::PipelineRasterizationState raster_state; + raster_state.cull_mode = cull_mode_rd; + raster_state.wireframe = wireframe; + + RD::PipelineColorBlendState blend_state; + RD::PipelineDepthStencilState depth_stencil = depth_stencil_state; + RD::PipelineMultisampleState multisample_state; + + if (uses_alpha || uses_blend_alpha) { + // only allow these flags to go through if we have some form of msaa + if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE) { + multisample_state.enable_alpha_to_coverage = true; + } else if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE) { + multisample_state.enable_alpha_to_coverage = true; + multisample_state.enable_alpha_to_one = true; + } + + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_MULTIVIEW || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW) { + blend_state = blend_state_blend; + if (depth_draw == DEPTH_DRAW_OPAQUE) { + depth_stencil.enable_depth_write = false; //alpha does not draw depth + } + } else if (k == SHADER_VERSION_SHADOW_PASS || k == SHADER_VERSION_SHADOW_PASS_MULTIVIEW || k == SHADER_VERSION_SHADOW_PASS_DP) { + //none, blend state contains nothing + } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) { + blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way + } else { + pipelines[i][j][k].clear(); + continue; // do not use this version (will error if using it is attempted) + } + } else { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_MULTIVIEW || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW) { + blend_state = blend_state_opaque; + } else if (k == SHADER_VERSION_SHADOW_PASS || k == SHADER_VERSION_SHADOW_PASS_MULTIVIEW || k == SHADER_VERSION_SHADOW_PASS_DP) { + //none, leave empty + } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) { + blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way + } else { + // ??? + } + } + + RID shader_variant = shader_singleton->shader.version_get_shader(version, k); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); + } + } + } + + valid = true; +} + +void SceneShaderForwardMobile::ShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) { + if (!p_texture.is_valid()) { + default_texture_params.erase(p_name); + } else { + default_texture_params[p_name] = p_texture; + } +} + +void SceneShaderForwardMobile::ShaderData::get_param_list(List<PropertyInfo> *p_param_list) const { + Map<int, StringName> order; + + for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { + if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_LOCAL) { + continue; + } + + if (E->get().texture_order >= 0) { + order[E->get().texture_order + 100000] = E->key(); + } else { + order[E->get().order] = E->key(); + } + } + + for (Map<int, StringName>::Element *E = order.front(); E; E = E->next()) { + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E->get()]); + pi.name = E->get(); + p_param_list->push_back(pi); + } +} + +void SceneShaderForwardMobile::ShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const { + for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { + if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E->get()); + p.info.name = E->key(); //supply name + p.index = E->get().instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E->get().default_value, E->get().type, E->get().hint); + p_param_list->push_back(p); + } +} + +bool SceneShaderForwardMobile::ShaderData::is_param_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool SceneShaderForwardMobile::ShaderData::is_animated() const { + return false; +} + +bool SceneShaderForwardMobile::ShaderData::casts_shadows() const { + return false; +} + +Variant SceneShaderForwardMobile::ShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode SceneShaderForwardMobile::ShaderData::get_native_source_code() const { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + return shader_singleton->shader.version_get_native_source_code(version); +} + +SceneShaderForwardMobile::ShaderData::ShaderData() : + shader_list_element(this) { + valid = false; + uses_screen_texture = false; +} + +SceneShaderForwardMobile::ShaderData::~ShaderData() { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + ERR_FAIL_COND(!shader_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + shader_singleton->shader.version_free(version); + } +} + +RendererStorageRD::ShaderData *SceneShaderForwardMobile::_create_shader_func() { + ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); + return shader_data; +} + +void SceneShaderForwardMobile::MaterialData::set_render_priority(int p_priority) { + priority = p_priority - RS::MATERIAL_RENDER_PRIORITY_MIN; //8 bits +} + +void SceneShaderForwardMobile::MaterialData::set_next_pass(RID p_pass) { + next_pass = p_pass; +} + +bool SceneShaderForwardMobile::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardMobile::MATERIAL_UNIFORM_SET, RD::BARRIER_MASK_RASTER); +} + +SceneShaderForwardMobile::MaterialData::~MaterialData() { + free_parameters_uniform_set(uniform_set); +} + +RendererStorageRD::MaterialData *SceneShaderForwardMobile::_create_material_func(ShaderData *p_shader) { + MaterialData *material_data = memnew(MaterialData); + material_data->shader_data = p_shader; + material_data->last_frame = false; + //update will happen later anyway so do nothing. + return material_data; +} + +/* Scene Shader */ + +SceneShaderForwardMobile *SceneShaderForwardMobile::singleton = nullptr; + +SceneShaderForwardMobile::SceneShaderForwardMobile() { + // there should be only one of these, contained within our RenderForwardMobile singleton. + singleton = this; +} + +void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p_defines) { + storage = p_storage; + + /* SCENE SHADER */ + + { + Vector<String> shader_versions; + shader_versions.push_back(""); // SHADER_VERSION_COLOR_PASS + shader_versions.push_back("\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_SHADOW_PASS, should probably change this to MODE_RENDER_SHADOW because we don't have a depth pass here... + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n"); // SHADER_VERSION_SHADOW_PASS_DP + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); // SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL + + // multiview versions of our shaders + shader_versions.push_back("\n#define USE_MULTIVIEW\n"); // SHADER_VERSION_COLOR_PASS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_SHADOW_PASS_MULTIVIEW + + shader.initialize(shader_versions, p_defines); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + shader.set_variant_enabled(SHADER_VERSION_COLOR_PASS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_SHADOW_PASS_MULTIVIEW, false); + } + } + + storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_3D, _create_shader_funcs); + storage->material_set_data_request_function(RendererStorageRD::SHADER_TYPE_3D, _create_material_funcs); + + { + //shader compiler + ShaderCompilerRD::DefaultIdentifierActions actions; + + actions.renames["WORLD_MATRIX"] = "world_matrix"; + actions.renames["WORLD_NORMAL_MATRIX"] = "world_normal_matrix"; + actions.renames["INV_CAMERA_MATRIX"] = "scene_data.inv_camera_matrix"; + actions.renames["CAMERA_MATRIX"] = "scene_data.camera_matrix"; + actions.renames["PROJECTION_MATRIX"] = "projection_matrix"; + actions.renames["INV_PROJECTION_MATRIX"] = "inv_projection_matrix"; + actions.renames["MODELVIEW_MATRIX"] = "modelview"; + actions.renames["MODELVIEW_NORMAL_MATRIX"] = "modelview_normal"; + + actions.renames["VERTEX"] = "vertex"; + actions.renames["NORMAL"] = "normal"; + actions.renames["TANGENT"] = "tangent"; + actions.renames["BINORMAL"] = "binormal"; + actions.renames["POSITION"] = "position"; + actions.renames["UV"] = "uv_interp"; + actions.renames["UV2"] = "uv2_interp"; + actions.renames["COLOR"] = "color_interp"; + actions.renames["POINT_SIZE"] = "gl_PointSize"; + actions.renames["INSTANCE_ID"] = "gl_InstanceIndex"; + + actions.renames["ALPHA_SCISSOR_THRESHOLD"] = "alpha_scissor_threshold"; + actions.renames["ALPHA_HASH_SCALE"] = "alpha_hash_scale"; + actions.renames["ALPHA_ANTIALIASING_EDGE"] = "alpha_antialiasing_edge"; + actions.renames["ALPHA_TEXTURE_COORDINATE"] = "alpha_texture_coordinate"; + + //builtins + + actions.renames["TIME"] = "scene_data.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["VIEWPORT_SIZE"] = "scene_data.viewport_size"; + + actions.renames["FRAGCOORD"] = "gl_FragCoord"; + actions.renames["FRONT_FACING"] = "gl_FrontFacing"; + actions.renames["NORMAL_MAP"] = "normal_map"; + actions.renames["NORMAL_MAP_DEPTH"] = "normal_map_depth"; + actions.renames["ALBEDO"] = "albedo"; + actions.renames["ALPHA"] = "alpha"; + actions.renames["METALLIC"] = "metallic"; + actions.renames["SPECULAR"] = "specular"; + actions.renames["ROUGHNESS"] = "roughness"; + actions.renames["RIM"] = "rim"; + actions.renames["RIM_TINT"] = "rim_tint"; + actions.renames["CLEARCOAT"] = "clearcoat"; + actions.renames["CLEARCOAT_GLOSS"] = "clearcoat_gloss"; + actions.renames["ANISOTROPY"] = "anisotropy"; + actions.renames["ANISOTROPY_FLOW"] = "anisotropy_flow"; + actions.renames["SSS_STRENGTH"] = "sss_strength"; + actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; + actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; + actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; + actions.renames["BACKLIGHT"] = "backlight"; + actions.renames["AO"] = "ao"; + actions.renames["AO_LIGHT_AFFECT"] = "ao_light_affect"; + actions.renames["EMISSION"] = "emission"; + actions.renames["POINT_COORD"] = "gl_PointCoord"; + actions.renames["INSTANCE_CUSTOM"] = "instance_custom"; + actions.renames["SCREEN_UV"] = "screen_uv"; + actions.renames["SCREEN_TEXTURE"] = "color_buffer"; + actions.renames["DEPTH_TEXTURE"] = "depth_buffer"; + actions.renames["NORMAL_ROUGHNESS_TEXTURE"] = "normal_roughness_buffer"; + actions.renames["DEPTH"] = "gl_FragDepth"; + actions.renames["OUTPUT_IS_SRGB"] = "true"; + actions.renames["FOG"] = "custom_fog"; + actions.renames["RADIANCE"] = "custom_radiance"; + actions.renames["IRRADIANCE"] = "custom_irradiance"; + actions.renames["BONE_INDICES"] = "bone_attrib"; + actions.renames["BONE_WEIGHTS"] = "weight_attrib"; + actions.renames["CUSTOM0"] = "custom0_attrib"; + actions.renames["CUSTOM1"] = "custom1_attrib"; + actions.renames["CUSTOM2"] = "custom2_attrib"; + actions.renames["CUSTOM3"] = "custom3_attrib"; + + actions.renames["VIEW_INDEX"] = "ViewIndex"; + actions.renames["VIEW_MONO_LEFT"] = "0"; + actions.renames["VIEW_RIGHT"] = "1"; + + //for light + actions.renames["VIEW"] = "view"; + actions.renames["LIGHT_COLOR"] = "light_color"; + actions.renames["LIGHT"] = "light"; + actions.renames["ATTENUATION"] = "attenuation"; + actions.renames["SHADOW_ATTENUATION"] = "shadow_attenuation"; + actions.renames["DIFFUSE_LIGHT"] = "diffuse_light"; + actions.renames["SPECULAR_LIGHT"] = "specular_light"; + + actions.usage_defines["NORMAL"] = "#define NORMAL_USED\n"; + actions.usage_defines["TANGENT"] = "#define TANGENT_USED\n"; + actions.usage_defines["BINORMAL"] = "@TANGENT"; + actions.usage_defines["RIM"] = "#define LIGHT_RIM_USED\n"; + actions.usage_defines["RIM_TINT"] = "@RIM"; + actions.usage_defines["CLEARCOAT"] = "#define LIGHT_CLEARCOAT_USED\n"; + actions.usage_defines["CLEARCOAT_GLOSS"] = "@CLEARCOAT"; + actions.usage_defines["ANISOTROPY"] = "#define LIGHT_ANISOTROPY_USED\n"; + actions.usage_defines["ANISOTROPY_FLOW"] = "@ANISOTROPY"; + actions.usage_defines["AO"] = "#define AO_USED\n"; + actions.usage_defines["AO_LIGHT_AFFECT"] = "#define AO_USED\n"; + actions.usage_defines["UV"] = "#define UV_USED\n"; + actions.usage_defines["UV2"] = "#define UV2_USED\n"; + actions.usage_defines["BONE_INDICES"] = "#define BONES_USED\n"; + actions.usage_defines["BONE_WEIGHTS"] = "#define WEIGHTS_USED\n"; + actions.usage_defines["CUSTOM0"] = "#define CUSTOM0\n"; + actions.usage_defines["CUSTOM1"] = "#define CUSTOM1\n"; + actions.usage_defines["CUSTOM2"] = "#define CUSTOM2\n"; + actions.usage_defines["CUSTOM3"] = "#define CUSTOM3\n"; + actions.usage_defines["NORMAL_MAP"] = "#define NORMAL_MAP_USED\n"; + actions.usage_defines["NORMAL_MAP_DEPTH"] = "@NORMAL_MAP"; + actions.usage_defines["COLOR"] = "#define COLOR_USED\n"; + actions.usage_defines["INSTANCE_CUSTOM"] = "#define ENABLE_INSTANCE_CUSTOM\n"; + actions.usage_defines["POSITION"] = "#define OVERRIDE_POSITION\n"; + + actions.usage_defines["ALPHA_SCISSOR_THRESHOLD"] = "#define ALPHA_SCISSOR_USED\n"; + actions.usage_defines["ALPHA_HASH_SCALE"] = "#define ALPHA_HASH_USED\n"; + actions.usage_defines["ALPHA_ANTIALIASING_EDGE"] = "#define ALPHA_ANTIALIASING_EDGE_USED\n"; + actions.usage_defines["ALPHA_TEXTURE_COORDINATE"] = "@ALPHA_ANTIALIASING_EDGE"; + + actions.usage_defines["SSS_STRENGTH"] = "#define ENABLE_SSS\n"; + actions.usage_defines["SSS_TRANSMITTANCE_DEPTH"] = "#define ENABLE_TRANSMITTANCE\n"; + actions.usage_defines["BACKLIGHT"] = "#define LIGHT_BACKLIGHT_USED\n"; + actions.usage_defines["SCREEN_TEXTURE"] = "#define SCREEN_TEXTURE_USED\n"; + actions.usage_defines["SCREEN_UV"] = "#define SCREEN_UV_USED\n"; + + actions.usage_defines["DIFFUSE_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + actions.usage_defines["SPECULAR_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + + actions.usage_defines["FOG"] = "#define CUSTOM_FOG_USED\n"; + actions.usage_defines["RADIANCE"] = "#define CUSTOM_RADIANCE_USED\n"; + actions.usage_defines["IRRADIANCE"] = "#define CUSTOM_IRRADIANCE_USED\n"; + + actions.render_mode_defines["skip_vertex_transform"] = "#define SKIP_TRANSFORM_USED\n"; + actions.render_mode_defines["world_vertex_coords"] = "#define VERTEX_WORLD_COORDS_USED\n"; + actions.render_mode_defines["ensure_correct_normals"] = "#define ENSURE_CORRECT_NORMALS\n"; + actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; + + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); + if (!force_lambert) { + actions.render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; + } + + actions.render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; + actions.render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; + + actions.render_mode_defines["sss_mode_skin"] = "#define SSS_MODE_SKIN\n"; + + bool force_blinn = GLOBAL_GET("rendering/shading/overrides/force_blinn_over_ggx"); + if (!force_blinn) { + actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; + } else { + actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_BLINN\n"; + } + + actions.render_mode_defines["specular_blinn"] = "#define SPECULAR_BLINN\n"; + actions.render_mode_defines["specular_phong"] = "#define SPECULAR_PHONG\n"; + actions.render_mode_defines["specular_toon"] = "#define SPECULAR_TOON\n"; + actions.render_mode_defines["specular_disabled"] = "#define SPECULAR_DISABLED\n"; + actions.render_mode_defines["shadows_disabled"] = "#define SHADOWS_DISABLED\n"; + actions.render_mode_defines["ambient_light_disabled"] = "#define AMBIENT_LIGHT_DISABLED\n"; + actions.render_mode_defines["shadow_to_opacity"] = "#define USE_SHADOW_TO_OPACITY\n"; + actions.render_mode_defines["unshaded"] = "#define MODE_UNSHADED\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = RenderForwardMobile::MATERIAL_UNIFORM_SET; + actions.base_uniform_string = "material."; + actions.base_varying_index = 10; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; + actions.global_buffer_array_variable = "global_variables.data"; + actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; + + compiler.initialize(actions); + } + + { + //default material and shader + default_shader = storage->shader_allocate(); + storage->shader_initialize(default_shader); + storage->shader_set_code(default_shader, R"( +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); + default_material = storage->material_allocate(); + storage->material_initialize(default_material); + storage->material_set_shader(default_material, default_shader); + + MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; + } + + { + overdraw_material_shader = storage->shader_allocate(); + storage->shader_initialize(overdraw_material_shader); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + storage->shader_set_code(overdraw_material_shader, R"( +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); + overdraw_material = storage->material_allocate(); + storage->material_initialize(overdraw_material); + storage->material_set_shader(overdraw_material, overdraw_material_shader); + + MaterialData *md = (MaterialData *)storage->material_get_data(overdraw_material, RendererStorageRD::SHADER_TYPE_3D); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; + } + + { + default_vec4_xform_buffer = RD::get_singleton()->storage_buffer_create(256); + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(default_vec4_xform_buffer); + u.binding = 0; + uniforms.push_back(u); + + default_vec4_xform_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RenderForwardMobile::TRANSFORMS_UNIFORM_SET); + } + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.enable_compare = true; + sampler.compare_op = RD::COMPARE_OP_LESS; + shadow_sampler = RD::get_singleton()->sampler_create(sampler); + } +} + +void SceneShaderForwardMobile::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} + +SceneShaderForwardMobile::~SceneShaderForwardMobile() { + RD::get_singleton()->free(default_vec4_xform_buffer); + RD::get_singleton()->free(shadow_sampler); + + storage->free(overdraw_material_shader); + storage->free(default_shader); + + storage->free(overdraw_material); + storage->free(default_material); +} diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h new file mode 100644 index 0000000000..e1c10f0206 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h @@ -0,0 +1,216 @@ +/*************************************************************************/ +/* scene_shader_forward_mobile.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RSSR_SCENE_SHADER_FM_H +#define RSSR_SCENE_SHADER_FM_H + +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/renderer_storage_rd.h" +#include "servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl.gen.h" + +namespace RendererSceneRenderImplementation { + +class SceneShaderForwardMobile { +private: + static SceneShaderForwardMobile *singleton; + RendererStorageRD *storage; + +public: + enum ShaderVersion { + SHADER_VERSION_COLOR_PASS, + SHADER_VERSION_LIGHTMAP_COLOR_PASS, + SHADER_VERSION_SHADOW_PASS, + SHADER_VERSION_SHADOW_PASS_DP, + SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, + + SHADER_VERSION_COLOR_PASS_MULTIVIEW, + SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW, + SHADER_VERSION_SHADOW_PASS_MULTIVIEW, + + SHADER_VERSION_MAX + }; + + struct ShaderData : public RendererStorageRD::ShaderData { + enum BlendMode { //used internally + BLEND_MODE_MIX, + BLEND_MODE_ADD, + BLEND_MODE_SUB, + BLEND_MODE_MUL, + BLEND_MODE_ALPHA_TO_COVERAGE + }; + + enum DepthDraw { + DEPTH_DRAW_DISABLED, + DEPTH_DRAW_OPAQUE, + DEPTH_DRAW_ALWAYS + }; + + enum DepthTest { + DEPTH_TEST_DISABLED, + DEPTH_TEST_ENABLED + }; + + enum Cull { + CULL_DISABLED, + CULL_FRONT, + CULL_BACK + }; + + enum CullVariant { + CULL_VARIANT_NORMAL, + CULL_VARIANT_REVERSED, + CULL_VARIANT_DOUBLE_SIDED, + CULL_VARIANT_MAX + + }; + + enum AlphaAntiAliasing { + ALPHA_ANTIALIASING_OFF, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE + }; + + bool valid; + RID version; + uint32_t vertex_input_mask; + PipelineCacheRD pipelines[CULL_VARIANT_MAX][RS::PRIMITIVE_MAX][SHADER_VERSION_MAX]; + + String path; + + Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size; + + String code; + Map<StringName, RID> default_texture_params; + + DepthDraw depth_draw; + DepthTest depth_test; + + bool uses_point_size; + bool uses_alpha; + bool uses_blend_alpha; + bool uses_alpha_clip; + bool uses_depth_pre_pass; + bool uses_discard; + bool uses_roughness; + bool uses_normal; + bool uses_particle_trails; + + bool unshaded; + bool uses_vertex; + bool uses_sss; + bool uses_transmittance; + bool uses_screen_texture; + bool uses_depth_texture; + bool uses_normal_texture; + bool uses_time; + bool writes_modelview_or_projection; + bool uses_world_coordinates; + + uint64_t last_pass = 0; + uint32_t index = 0; + + virtual void set_code(const String &p_Code); + virtual void set_default_texture_param(const StringName &p_name, RID p_texture); + virtual void get_param_list(List<PropertyInfo> *p_param_list) const; + void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; + + virtual bool is_param_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + SelfList<ShaderData> shader_list_element; + + ShaderData(); + virtual ~ShaderData(); + }; + + RendererStorageRD::ShaderData *_create_shader_func(); + static RendererStorageRD::ShaderData *_create_shader_funcs() { + return static_cast<SceneShaderForwardMobile *>(singleton)->_create_shader_func(); + } + + struct MaterialData : public RendererStorageRD::MaterialData { + uint64_t last_frame; + ShaderData *shader_data; + RID uniform_set; + uint64_t last_pass = 0; + uint32_t index = 0; + RID next_pass; + uint8_t priority; + virtual void set_render_priority(int p_priority); + virtual void set_next_pass(RID p_pass); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~MaterialData(); + }; + + SelfList<ShaderData>::List shader_list; + + RendererStorageRD::MaterialData *_create_material_func(ShaderData *p_shader); + static RendererStorageRD::MaterialData *_create_material_funcs(RendererStorageRD::ShaderData *p_shader) { + return static_cast<SceneShaderForwardMobile *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); + } + + SceneForwardMobileShaderRD shader; + ShaderCompilerRD compiler; + + RID default_shader; + RID default_material; + RID overdraw_material_shader; + RID overdraw_material; + RID default_shader_rd; + + RID default_vec4_xform_buffer; + RID default_vec4_xform_uniform_set; + + RID shadow_sampler; + + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + + SceneShaderForwardMobile(); + ~SceneShaderForwardMobile(); + + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; + + void init(RendererStorageRD *p_storage, const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); +}; + +} // namespace RendererSceneRenderImplementation +#endif // !RSSR_SCENE_SHADER_FM_H diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp index b2b919c40e..aefe926cb0 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp @@ -31,25 +31,46 @@ #include "pipeline_cache_rd.h" #include "core/os/memory.h" -RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe) { +RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations) { RD::PipelineMultisampleState multisample_state_version = multisample_state; - multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id); + multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id, p_render_pass); RD::PipelineRasterizationState raster_state_version = rasterization_state; raster_state_version.wireframe = p_wireframe; - RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, raster_state_version, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags); + Vector<RD::PipelineSpecializationConstant> specialization_constants = base_specialization_constants; + + uint32_t bool_index = 0; + uint32_t bool_specializations = p_bool_specializations; + while (bool_specializations) { + if (bool_specializations & (1 << bool_index)) { + RD::PipelineSpecializationConstant sc; + sc.bool_value = true; + sc.constant_id = bool_index; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + specialization_constants.push_back(sc); + bool_specializations &= ~(1 << bool_index); + } + bool_index++; + } + + RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, raster_state_version, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags, p_render_pass, specialization_constants); ERR_FAIL_COND_V(pipeline.is_null(), RID()); versions = (Version *)memrealloc(versions, sizeof(Version) * (version_count + 1)); versions[version_count].framebuffer_id = p_framebuffer_format_id; versions[version_count].vertex_id = p_vertex_format_id; versions[version_count].wireframe = p_wireframe; versions[version_count].pipeline = pipeline; + versions[version_count].render_pass = p_render_pass; + versions[version_count].bool_specializations = p_bool_specializations; version_count++; return pipeline; } void PipelineCacheRD::_clear() { +#ifndef _MSC_VER +#warning Clear should probably recompile all the variants already compiled instead to avoid stalls? needs discussion +#endif if (versions) { for (uint32_t i = 0; i < version_count; i++) { //shader may be gone, so this may not be valid @@ -63,7 +84,7 @@ void PipelineCacheRD::_clear() { } } -void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags) { +void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { ERR_FAIL_COND(p_shader.is_null()); _clear(); shader = p_shader; @@ -74,6 +95,11 @@ void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const depth_stencil_state = p_depth_stencil_state; blend_state = p_blend_state; dynamic_state_flags = p_dynamic_state_flags; + base_specialization_constants = p_base_specialization_constants; +} +void PipelineCacheRD::update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { + base_specialization_constants = p_base_specialization_constants; + _clear(); } void PipelineCacheRD::update_shader(RID p_shader) { diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.h b/servers/rendering/renderer_rd/pipeline_cache_rd.h index b1c8f21ecc..e52f47fa47 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.h +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.h @@ -46,26 +46,30 @@ class PipelineCacheRD { RD::PipelineDepthStencilState depth_stencil_state; RD::PipelineColorBlendState blend_state; int dynamic_state_flags; + Vector<RD::PipelineSpecializationConstant> base_specialization_constants; struct Version { RD::VertexFormatID vertex_id; RD::FramebufferFormatID framebuffer_id; + uint32_t render_pass; bool wireframe; + uint32_t bool_specializations; RID pipeline; }; Version *versions; uint32_t version_count; - RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe); + RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations = 0); void _clear(); public: - void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0); + void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants = Vector<RD::PipelineSpecializationConstant>()); + void update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants); void update_shader(RID p_shader); - _FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe = false) { + _FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe = false, uint32_t p_render_pass = 0, uint32_t p_bool_specializations = 0) { #ifdef DEBUG_ENABLED ERR_FAIL_COND_V_MSG(shader.is_null(), RID(), "Attempted to use an unused shader variant (shader is null),"); @@ -74,13 +78,13 @@ public: spin_lock.lock(); RID result; for (uint32_t i = 0; i < version_count; i++) { - if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id && versions[i].wireframe == p_wireframe) { + if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id && versions[i].wireframe == p_wireframe && versions[i].render_pass == p_render_pass && versions[i].bool_specializations == p_bool_specializations) { result = versions[i].pipeline; spin_lock.unlock(); return result; } } - result = _generate_version(p_vertex_format_id, p_framebuffer_format_id, p_wireframe); + result = _generate_version(p_vertex_format_id, p_framebuffer_format_id, p_wireframe, p_render_pass, p_bool_specializations); spin_lock.unlock(); return result; } diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 7d6e2fa8e4..f8aefdb29c 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -31,8 +31,10 @@ #include "renderer_canvas_render_rd.h" #include "core/config/project_settings.h" #include "core/math/geometry_2d.h" +#include "core/math/math_defs.h" #include "core/math/math_funcs.h" #include "renderer_compositor_rd.h" +#include "servers/rendering/rendering_server_default.h" void RendererCanvasRenderRD::_update_transform_2d_to_mat4(const Transform2D &p_transform, float *p_mat4) { p_mat4[0] = p_transform.elements[0][0]; @@ -74,7 +76,7 @@ void RendererCanvasRenderRD::_update_transform_2d_to_mat2x3(const Transform2D &p p_mat2x3[5] = p_transform.elements[2][1]; } -void RendererCanvasRenderRD::_update_transform_to_mat4(const Transform &p_transform, float *p_mat4) { +void RendererCanvasRenderRD::_update_transform_to_mat4(const Transform3D &p_transform, float *p_mat4) { p_mat4[0] = p_transform.basis.elements[0][0]; p_mat4[1] = p_transform.basis.elements[1][0]; p_mat4[2] = p_transform.basis.elements[2][0]; @@ -304,7 +306,7 @@ RendererCanvasRender::PolygonID RendererCanvasRenderRD::request_polygon(const Ve index_buffer.resize(p_indices.size() * sizeof(int32_t)); { uint8_t *w = index_buffer.ptrw(); - copymem(w, p_indices.ptr(), sizeof(int32_t) * p_indices.size()); + memcpy(w, p_indices.ptr(), sizeof(int32_t) * p_indices.size()); } pb.index_buffer = RD::get_singleton()->index_buffer_create(p_indices.size(), RD::INDEX_BUFFER_FORMAT_UINT32, index_buffer); pb.indices = RD::get_singleton()->index_array_create(pb.index_buffer, 0, p_indices.size()); @@ -390,7 +392,7 @@ void RendererCanvasRenderRD::_bind_canvas_texture(RD::DrawListID p_draw_list, RI r_last_texture = p_texture; } -void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item *p_item, RD::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants) { +void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_render_target, const Item *p_item, RD::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants) { //create an empty push constant RS::CanvasItemTextureFilter current_filter = default_filter; @@ -406,6 +408,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item PushConstant push_constant; Transform2D base_transform = p_canvas_transform_inverse * p_item->final_transform; + Transform2D draw_transform; _update_transform_2d_to_mat2x3(base_transform, push_constant.world); Color base_color = p_item->final_modulate; @@ -462,14 +465,25 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item RID last_texture; Size2 texpixel_size; + bool skipping = false; + const Item::Command *c = p_item->commands; while (c) { + if (skipping && c->type != Item::Command::TYPE_ANIMATION_SLICE) { + c = c->next; + continue; + } + push_constant.flags = base_flags | (push_constant.flags & (FLAGS_DEFAULT_NORMAL_MAP_USED | FLAGS_DEFAULT_SPECULAR_MAP_USED)); //reset on each command for sanity, keep canvastexture binding config switch (c->type) { case Item::Command::TYPE_RECT: { const Item::CommandRect *rect = static_cast<const Item::CommandRect *>(c); + if (rect->flags & CANVAS_RECT_TILE) { + current_repeat = RenderingServer::CanvasItemTextureRepeat::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED; + } + //bind pipeline { RID pipeline = pipeline_variants->variants[light_mode][PIPELINE_VARIANT_QUAD].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); @@ -610,7 +624,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item RD::get_singleton()->draw_list_bind_index_array(p_draw_list, shader.quad_index_array); RD::get_singleton()->draw_list_draw(p_draw_list, true); - //restore if overrided + // Restore if overridden. push_constant.color_texture_pixel_size[0] = texpixel_size.x; push_constant.color_texture_pixel_size[1] = texpixel_size.y; @@ -705,288 +719,160 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item case Item::Command::TYPE_MESH: case Item::Command::TYPE_MULTIMESH: case Item::Command::TYPE_PARTICLES: { - ERR_PRINT("FIXME: Mesh, MultiMesh and Particles render commands are unimplemented currently, they need to be ported to the 4.0 rendering architecture."); -#ifndef _MSC_VER -#warning Item::Command types for Mesh, MultiMesh and Particles need to be implemented. -#endif - // See #if 0'ed code below to port from GLES3. - } break; - -#if 0 - case Item::Command::TYPE_MESH: { - Item::CommandMesh *mesh = static_cast<Item::CommandMesh *>(c); - _set_texture_rect_mode(false); - - RasterizerStorageGLES3::Texture *texture = _bind_canvas_texture(mesh->texture, mesh->normal_map); - - if (texture) { - Size2 texpixel_size(1.0 / texture->width, 1.0 / texture->height); - state.canvas_shader.set_uniform(CanvasShaderGLES3::COLOR_TEXPIXEL_SIZE, texpixel_size); + RID mesh; + RID mesh_instance; + RID texture; + Color modulate(1, 1, 1, 1); + float world_backup[6]; + int instance_count = 1; + + for (int j = 0; j < 6; j++) { + world_backup[j] = push_constant.world[j]; } - state.canvas_shader.set_uniform(CanvasShaderGLES3::MODELVIEW_MATRIX, state.final_transform * mesh->transform); - - RasterizerStorageGLES3::Mesh *mesh_data = storage->mesh_owner.getornull(mesh->mesh); - if (mesh_data) { - for (int j = 0; j < mesh_data->surfaces.size(); j++) { - RasterizerStorageGLES3::Surface *s = mesh_data->surfaces[j]; - // materials are ignored in 2D meshes, could be added but many things (ie, lighting mode, reading from screen, etc) would break as they are not meant be set up at this point of drawing - glBindVertexArray(s->array_id); - - glVertexAttrib4f(RS::ARRAY_COLOR, mesh->modulate.r, mesh->modulate.g, mesh->modulate.b, mesh->modulate.a); - - if (s->index_array_len) { - glDrawElements(gl_primitive[s->primitive], s->index_array_len, (s->array_len >= (1 << 16)) ? GL_UNSIGNED_INT : GL_UNSIGNED_SHORT, 0); - } else { - glDrawArrays(gl_primitive[s->primitive], 0, s->array_len); - } - - glBindVertexArray(0); + if (c->type == Item::Command::TYPE_MESH) { + const Item::CommandMesh *m = static_cast<const Item::CommandMesh *>(c); + mesh = m->mesh; + mesh_instance = m->mesh_instance; + texture = m->texture; + modulate = m->modulate; + _update_transform_2d_to_mat2x3(base_transform * draw_transform * m->transform, push_constant.world); + } else if (c->type == Item::Command::TYPE_MULTIMESH) { + const Item::CommandMultiMesh *mm = static_cast<const Item::CommandMultiMesh *>(c); + RID multimesh = mm->multimesh; + mesh = storage->multimesh_get_mesh(multimesh); + texture = mm->texture; + + if (storage->multimesh_get_transform_format(multimesh) != RS::MULTIMESH_TRANSFORM_2D) { + break; } - } - state.canvas_shader.set_uniform(CanvasShaderGLES3::MODELVIEW_MATRIX, state.final_transform); - - } break; - case Item::Command::TYPE_MULTIMESH: { - Item::CommandMultiMesh *mmesh = static_cast<Item::CommandMultiMesh *>(c); - - RasterizerStorageGLES3::MultiMesh *multi_mesh = storage->multimesh_owner.getornull(mmesh->multimesh); - - if (!multi_mesh) - break; - - RasterizerStorageGLES3::Mesh *mesh_data = storage->mesh_owner.getornull(multi_mesh->mesh); - if (!mesh_data) - break; + instance_count = storage->multimesh_get_instances_to_draw(multimesh); - RasterizerStorageGLES3::Texture *texture = _bind_canvas_texture(mmesh->texture, mmesh->normal_map); + RID uniform_set = storage->multimesh_get_2d_uniform_set(multimesh, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, TRANSFORMS_UNIFORM_SET); + push_constant.flags |= 1; //multimesh, trails disabled + if (storage->multimesh_uses_colors(multimesh)) { + push_constant.flags |= FLAGS_INSTANCING_HAS_COLORS; + } + if (storage->multimesh_uses_custom_data(multimesh)) { + push_constant.flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; + } + } else if (c->type == Item::Command::TYPE_PARTICLES) { + const Item::CommandParticles *pt = static_cast<const Item::CommandParticles *>(c); + ERR_BREAK(storage->particles_get_mode(pt->particles) != RS::PARTICLES_MODE_2D); + storage->particles_request_process(pt->particles); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCE_CUSTOM, multi_mesh->custom_data_format != RS::MULTIMESH_CUSTOM_DATA_NONE); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCING, true); - //reset shader and force rebind - state.using_texture_rect = true; - _set_texture_rect_mode(false); + if (storage->particles_is_inactive(pt->particles)) { + break; + } - if (texture) { - Size2 texpixel_size(1.0 / texture->width, 1.0 / texture->height); - state.canvas_shader.set_uniform(CanvasShaderGLES3::COLOR_TEXPIXEL_SIZE, texpixel_size); - } + RenderingServerDefault::redraw_request(); // active particles means redraw request - int amount = MIN(multi_mesh->size, multi_mesh->visible_instances); + bool local_coords = true; + int dpc = storage->particles_get_draw_passes(pt->particles); + if (dpc == 0) { + break; //nothing to draw + } + uint32_t divisor = 1; + instance_count = storage->particles_get_amount(pt->particles, divisor); - if (amount == -1) { - amount = multi_mesh->size; - } + RID uniform_set = storage->particles_get_instance_buffer_uniform_set(pt->particles, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, TRANSFORMS_UNIFORM_SET); - for (int j = 0; j < mesh_data->surfaces.size(); j++) { - RasterizerStorageGLES3::Surface *s = mesh_data->surfaces[j]; - // materials are ignored in 2D meshes, could be added but many things (ie, lighting mode, reading from screen, etc) would break as they are not meant be set up at this point of drawing - glBindVertexArray(s->instancing_array_id); + push_constant.flags |= divisor; + instance_count /= divisor; - glBindBuffer(GL_ARRAY_BUFFER, multi_mesh->buffer); //modify the buffer + push_constant.flags |= FLAGS_INSTANCING_HAS_COLORS; + push_constant.flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; - int stride = (multi_mesh->xform_floats + multi_mesh->color_floats + multi_mesh->custom_data_floats) * 4; - glEnableVertexAttribArray(8); - glVertexAttribPointer(8, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(0)); - glVertexAttribDivisor(8, 1); - glEnableVertexAttribArray(9); - glVertexAttribPointer(9, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(4 * 4)); - glVertexAttribDivisor(9, 1); + mesh = storage->particles_get_draw_pass_mesh(pt->particles, 0); //higher ones are ignored + texture = pt->texture; - int color_ofs; + if (storage->particles_has_collision(pt->particles) && storage->render_target_is_sdf_enabled(p_render_target)) { + //pass collision information + Transform2D xform; + if (local_coords) { + xform = p_item->final_transform; + } else { + xform = p_canvas_transform_inverse; + } - if (multi_mesh->transform_format == RS::MULTIMESH_TRANSFORM_3D) { - glEnableVertexAttribArray(10); - glVertexAttribPointer(10, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(8 * 4)); - glVertexAttribDivisor(10, 1); - color_ofs = 12 * 4; - } else { - glDisableVertexAttribArray(10); - glVertexAttrib4f(10, 0, 0, 1, 0); - color_ofs = 8 * 4; - } + RID sdf_texture = storage->render_target_get_sdf_texture(p_render_target); - int custom_data_ofs = color_ofs; - - switch (multi_mesh->color_format) { - case RS::MULTIMESH_COLOR_NONE: { - glDisableVertexAttribArray(11); - glVertexAttrib4f(11, 1, 1, 1, 1); - } break; - case RS::MULTIMESH_COLOR_8BIT: { - glEnableVertexAttribArray(11); - glVertexAttribPointer(11, 4, GL_UNSIGNED_BYTE, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(color_ofs)); - glVertexAttribDivisor(11, 1); - custom_data_ofs += 4; - - } break; - case RS::MULTIMESH_COLOR_FLOAT: { - glEnableVertexAttribArray(11); - glVertexAttribPointer(11, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(color_ofs)); - glVertexAttribDivisor(11, 1); - custom_data_ofs += 4 * 4; - } break; - } + Rect2 to_screen; + { + Rect2 sdf_rect = storage->render_target_get_sdf_rect(p_render_target); - switch (multi_mesh->custom_data_format) { - case RS::MULTIMESH_CUSTOM_DATA_NONE: { - glDisableVertexAttribArray(12); - glVertexAttrib4f(12, 1, 1, 1, 1); - } break; - case RS::MULTIMESH_CUSTOM_DATA_8BIT: { - glEnableVertexAttribArray(12); - glVertexAttribPointer(12, 4, GL_UNSIGNED_BYTE, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(custom_data_ofs)); - glVertexAttribDivisor(12, 1); - - } break; - case RS::MULTIMESH_CUSTOM_DATA_FLOAT: { - glEnableVertexAttribArray(12); - glVertexAttribPointer(12, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(custom_data_ofs)); - glVertexAttribDivisor(12, 1); - } break; - } + to_screen.size = Vector2(1.0 / sdf_rect.size.width, 1.0 / sdf_rect.size.height); + to_screen.position = -sdf_rect.position * to_screen.size; + } - if (s->index_array_len) { - glDrawElementsInstanced(gl_primitive[s->primitive], s->index_array_len, (s->array_len >= (1 << 16)) ? GL_UNSIGNED_INT : GL_UNSIGNED_SHORT, 0, amount); + storage->particles_set_canvas_sdf_collision(pt->particles, true, xform, to_screen, sdf_texture); } else { - glDrawArraysInstanced(gl_primitive[s->primitive], 0, s->array_len, amount); + storage->particles_set_canvas_sdf_collision(pt->particles, false, Transform2D(), Rect2(), RID()); } - - glBindVertexArray(0); } - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCE_CUSTOM, false); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCING, false); - state.using_texture_rect = true; - _set_texture_rect_mode(false); - - } break; - case Item::Command::TYPE_PARTICLES: { - Item::CommandParticles *particles_cmd = static_cast<Item::CommandParticles *>(c); - - RasterizerStorageGLES3::Particles *particles = storage->particles_owner.getornull(particles_cmd->particles); - if (!particles) - break; - - if (particles->inactive && !particles->emitting) + if (mesh.is_null()) { break; + } - glVertexAttrib4f(RS::ARRAY_COLOR, 1, 1, 1, 1); //not used, so keep white + _bind_canvas_texture(p_draw_list, texture, current_filter, current_repeat, last_texture, push_constant, texpixel_size); - RenderingServerDefault::redraw_request(); + uint32_t surf_count = storage->mesh_get_surface_count(mesh); + static const PipelineVariant variant[RS::PRIMITIVE_MAX] = { PIPELINE_VARIANT_ATTRIBUTE_POINTS, PIPELINE_VARIANT_ATTRIBUTE_LINES, PIPELINE_VARIANT_ATTRIBUTE_LINES_STRIP, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLES, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLE_STRIP }; - storage->particles_request_process(particles_cmd->particles); - //enable instancing + push_constant.modulation[0] = base_color.r * modulate.r; + push_constant.modulation[1] = base_color.g * modulate.g; + push_constant.modulation[2] = base_color.b * modulate.b; + push_constant.modulation[3] = base_color.a * modulate.a; - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCE_CUSTOM, true); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_PARTICLES, true); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCING, true); - //reset shader and force rebind - state.using_texture_rect = true; - _set_texture_rect_mode(false); + for (int j = 0; j < 4; j++) { + push_constant.src_rect[j] = 0; + push_constant.dst_rect[j] = 0; + push_constant.ninepatch_margins[j] = 0; + } - RasterizerStorageGLES3::Texture *texture = _bind_canvas_texture(particles_cmd->texture, particles_cmd->normal_map); + for (uint32_t j = 0; j < surf_count; j++) { + void *surface = storage->mesh_get_surface(mesh, j); - if (texture) { - Size2 texpixel_size(1.0 / texture->width, 1.0 / texture->height); - state.canvas_shader.set_uniform(CanvasShaderGLES3::COLOR_TEXPIXEL_SIZE, texpixel_size); - } else { - state.canvas_shader.set_uniform(CanvasShaderGLES3::COLOR_TEXPIXEL_SIZE, Vector2(1.0, 1.0)); - } + RS::PrimitiveType primitive = storage->mesh_surface_get_primitive(surface); + ERR_CONTINUE(primitive < 0 || primitive >= RS::PRIMITIVE_MAX); - if (!particles->use_local_coords) { - Transform2D inv_xf; - inv_xf.set_axis(0, Vector2(particles->emission_transform.basis.get_axis(0).x, particles->emission_transform.basis.get_axis(0).y)); - inv_xf.set_axis(1, Vector2(particles->emission_transform.basis.get_axis(1).x, particles->emission_transform.basis.get_axis(1).y)); - inv_xf.set_origin(Vector2(particles->emission_transform.get_origin().x, particles->emission_transform.get_origin().y)); - inv_xf.affine_invert(); + uint32_t input_mask = pipeline_variants->variants[light_mode][variant[primitive]].get_vertex_input_mask(); - state.canvas_shader.set_uniform(CanvasShaderGLES3::MODELVIEW_MATRIX, state.final_transform * inv_xf); - } + RID vertex_array; + RD::VertexFormatID vertex_format = RD::INVALID_FORMAT_ID; - glBindVertexArray(data.particle_quad_array); //use particle quad array - glBindBuffer(GL_ARRAY_BUFFER, particles->particle_buffers[0]); //bind particle buffer - - int stride = sizeof(float) * 4 * 6; - - int amount = particles->amount; - - if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_LIFETIME) { - glEnableVertexAttribArray(8); //xform x - glVertexAttribPointer(8, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 3)); - glVertexAttribDivisor(8, 1); - glEnableVertexAttribArray(9); //xform y - glVertexAttribPointer(9, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 4)); - glVertexAttribDivisor(9, 1); - glEnableVertexAttribArray(10); //xform z - glVertexAttribPointer(10, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 5)); - glVertexAttribDivisor(10, 1); - glEnableVertexAttribArray(11); //color - glVertexAttribPointer(11, 4, GL_FLOAT, GL_FALSE, stride, nullptr); - glVertexAttribDivisor(11, 1); - glEnableVertexAttribArray(12); //custom - glVertexAttribPointer(12, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 2)); - glVertexAttribDivisor(12, 1); - - glDrawArraysInstanced(GL_TRIANGLE_FAN, 0, 4, amount); - } else { - //split - int split = int(Math::ceil(particles->phase * particles->amount)); - - if (amount - split > 0) { - glEnableVertexAttribArray(8); //xform x - glVertexAttribPointer(8, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * split + sizeof(float) * 4 * 3)); - glVertexAttribDivisor(8, 1); - glEnableVertexAttribArray(9); //xform y - glVertexAttribPointer(9, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * split + sizeof(float) * 4 * 4)); - glVertexAttribDivisor(9, 1); - glEnableVertexAttribArray(10); //xform z - glVertexAttribPointer(10, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * split + sizeof(float) * 4 * 5)); - glVertexAttribDivisor(10, 1); - glEnableVertexAttribArray(11); //color - glVertexAttribPointer(11, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * split + 0)); - glVertexAttribDivisor(11, 1); - glEnableVertexAttribArray(12); //custom - glVertexAttribPointer(12, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * split + sizeof(float) * 4 * 2)); - glVertexAttribDivisor(12, 1); - - glDrawArraysInstanced(GL_TRIANGLE_FAN, 0, 4, amount - split); + if (mesh_instance.is_valid()) { + storage->mesh_instance_surface_get_vertex_arrays_and_format(mesh_instance, j, input_mask, vertex_array, vertex_format); + } else { + storage->mesh_surface_get_vertex_arrays_and_format(surface, input_mask, vertex_array, vertex_format); } - if (split > 0) { - glEnableVertexAttribArray(8); //xform x - glVertexAttribPointer(8, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 3)); - glVertexAttribDivisor(8, 1); - glEnableVertexAttribArray(9); //xform y - glVertexAttribPointer(9, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 4)); - glVertexAttribDivisor(9, 1); - glEnableVertexAttribArray(10); //xform z - glVertexAttribPointer(10, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 5)); - glVertexAttribDivisor(10, 1); - glEnableVertexAttribArray(11); //color - glVertexAttribPointer(11, 4, GL_FLOAT, GL_FALSE, stride, nullptr); - glVertexAttribDivisor(11, 1); - glEnableVertexAttribArray(12); //custom - glVertexAttribPointer(12, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * 2)); - glVertexAttribDivisor(12, 1); - - glDrawArraysInstanced(GL_TRIANGLE_FAN, 0, 4, split); + RID pipeline = pipeline_variants->variants[light_mode][variant[primitive]].get_render_pipeline(vertex_format, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + + RID index_array = storage->mesh_surface_get_index_array(surface, 0); + + if (index_array.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, index_array); } - } - glBindVertexArray(0); + RD::get_singleton()->draw_list_bind_vertex_array(p_draw_list, vertex_array); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCE_CUSTOM, false); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_PARTICLES, false); - state.canvas_shader.set_conditional(CanvasShaderGLES3::USE_INSTANCING, false); - state.using_texture_rect = true; - _set_texture_rect_mode(false); + RD::get_singleton()->draw_list_draw(p_draw_list, index_array.is_valid(), instance_count); + } + for (int j = 0; j < 6; j++) { + push_constant.world[j] = world_backup[j]; + } } break; -#endif case Item::Command::TYPE_TRANSFORM: { const Item::CommandTransform *transform = static_cast<const Item::CommandTransform *>(c); + draw_transform = transform->xform; _update_transform_2d_to_mat2x3(base_transform * transform->xform, push_constant.world); } break; @@ -1005,6 +891,14 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item } } break; + case Item::Command::TYPE_ANIMATION_SLICE: { + const Item::CommandAnimationSlice *as = static_cast<const Item::CommandAnimationSlice *>(c); + double current_time = RendererCompositorRD::singleton->get_total_time(); + double local_time = Math::fposmod(current_time - as->offset, as->animation_length); + skipping = !(local_time >= as->slice_begin && local_time < as->slice_end); + + RenderingServerDefault::redraw_request(); // animation visible means redraw request + } break; } c = c->next; @@ -1199,7 +1093,8 @@ void RendererCanvasRenderRD::_render_items(RID p_to_render_target, int p_item_co if (material_data) { if (material_data->shader_data->version.is_valid() && material_data->shader_data->valid) { pipeline_variants = &material_data->shader_data->pipeline_variants; - if (material_data->uniform_set.is_valid()) { + // Update uniform set. + if (RD::get_singleton()->uniform_set_is_valid(material_data->uniform_set)) { // Material may not have a uniform set. RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_data->uniform_set, MATERIAL_UNIFORM_SET); } } else { @@ -1210,7 +1105,7 @@ void RendererCanvasRenderRD::_render_items(RID p_to_render_target, int p_item_co } } - _render_item(draw_list, ci, fb_format, canvas_transform_inverse, current_clip, p_lights, pipeline_variants); + _render_item(draw_list, p_to_render_target, ci, fb_format, canvas_transform_inverse, current_clip, p_lights, pipeline_variants); prev_material = material; } @@ -1376,7 +1271,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p Size2i ssize = storage->render_target_get_size(p_to_render_target); - Transform screen_transform; + Transform3D screen_transform; screen_transform.translate(-(ssize.width / 2.0f), -(ssize.height / 2.0f), 0.0f); screen_transform.scale(Vector3(2.0f / ssize.width, 2.0f / ssize.height, 1.0f)); _update_transform_to_mat4(screen_transform, state_buffer.screen_transform); @@ -1437,6 +1332,9 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p Item *canvas_group_owner = nullptr; + bool update_skeletons = false; + bool time_used = false; + while (ci) { if (ci->copy_back_buffer && canvas_group_owner == nullptr) { backbuffer_copy = true; @@ -1461,6 +1359,9 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p if (md->shader_data->uses_sdf) { r_sdf_used = true; } + if (md->shader_data->uses_time) { + time_used = true; + } if (md->last_frame != RendererCompositorRD::singleton->get_frame_number()) { md->last_frame = RendererCompositorRD::singleton->get_frame_number(); if (!RD::get_singleton()->uniform_set_is_valid(md->uniform_set)) { @@ -1472,9 +1373,27 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p } } + if (ci->skeleton.is_valid()) { + const Item::Command *c = ci->commands; + + while (c) { + if (c->type == Item::Command::TYPE_MESH) { + const Item::CommandMesh *cm = static_cast<const Item::CommandMesh *>(c); + if (cm->mesh_instance.is_valid()) { + storage->mesh_instance_check_for_update(cm->mesh_instance); + update_skeletons = true; + } + } + } + } + if (ci->canvas_group_owner != nullptr) { if (canvas_group_owner == nullptr) { //Canvas group begins here, render until before this item + if (update_skeletons) { + storage->update_mesh_instances(); + update_skeletons = false; + } _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); item_count = 0; @@ -1494,6 +1413,11 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p } if (ci == canvas_group_owner) { + if (update_skeletons) { + storage->update_mesh_instances(); + update_skeletons = false; + } + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list, true); item_count = 0; @@ -1506,6 +1430,10 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p if (backbuffer_copy) { //render anything pending, including clearing if no items + if (update_skeletons) { + storage->update_mesh_instances(); + update_skeletons = false; + } _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); item_count = 0; @@ -1518,6 +1446,11 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p items[item_count++] = ci; if (!ci->next || item_count == MAX_RENDER_ITEMS - 1) { + if (update_skeletons) { + storage->update_mesh_instances(); + update_skeletons = false; + } + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); //then reset item_count = 0; @@ -1525,6 +1458,10 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p ci = ci->next; } + + if (time_used) { + RenderingServerDefault::redraw_request(); + } } RID RendererCanvasRenderRD::light_create() { @@ -1623,7 +1560,7 @@ void RendererCanvasRenderRD::light_update_shadow(RID p_rid, int p_shadow_index, } Vector3 cam_target = Basis(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); - projection = projection * CameraMatrix(Transform().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); + projection = projection * CameraMatrix(Transform3D().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); ShadowRenderPushConstant push_constant; for (int y = 0; y < 4; y++) { @@ -1701,7 +1638,7 @@ void RendererCanvasRenderRD::light_update_directional_shadow(RID p_rid, int p_sh CameraMatrix projection; projection.set_orthogonal(-half_size, half_size, -0.5, 0.5, 0.0, distance); - projection = projection * CameraMatrix(Transform().looking_at(Vector3(0, 1, 0), Vector3(0, 0, -1)).affine_inverse()); + projection = projection * CameraMatrix(Transform3D().looking_at(Vector3(0, 1, 0), Vector3(0, 0, -1)).affine_inverse()); ShadowRenderPushConstant push_constant; for (int y = 0; y < 4; y++) { @@ -2001,6 +1938,7 @@ void RendererCanvasRenderRD::ShaderData::set_code(const String &p_code) { uniforms.clear(); uses_screen_texture = false; uses_sdf = false; + uses_time = false; if (code == String()) { return; //just invalid, but no error @@ -2012,6 +1950,9 @@ void RendererCanvasRenderRD::ShaderData::set_code(const String &p_code) { uses_screen_texture = false; ShaderCompilerRD::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompilerRD::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompilerRD::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompilerRD::STAGE_FRAGMENT; actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); @@ -2022,14 +1963,14 @@ void RendererCanvasRenderRD::ShaderData::set_code(const String &p_code) { actions.usage_flag_pointers["SCREEN_TEXTURE"] = &uses_screen_texture; actions.usage_flag_pointers["texture_sdf"] = &uses_sdf; + actions.usage_flag_pointers["TIME"] = &uses_time; actions.uniforms = &uniforms; RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; Error err = canvas_singleton->shader.compiler.compile(RS::SHADER_CANVAS_ITEM, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = canvas_singleton->shader.canvas_shader.version_create(); @@ -2048,7 +1989,7 @@ void RendererCanvasRenderRD::ShaderData::set_code(const String &p_code) { print_line("\n**fragment_code:\n" + gen_code.fragment); print_line("\n**light_code:\n" + gen_code.light); #endif - canvas_singleton->shader.canvas_shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.fragment, gen_code.defines); + canvas_singleton->shader.canvas_shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX], gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT], gen_code.defines); ERR_FAIL_COND(!canvas_singleton->shader.canvas_shader.version_is_valid(version)); ubo_size = gen_code.uniform_total_size; @@ -2264,94 +2205,14 @@ RendererStorageRD::ShaderData *RendererCanvasRenderRD::_create_shader_func() { return shader_data; } -void RendererCanvasRenderRD::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererCanvasRenderRD::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), false); - } - - if (shader_data->ubo_size == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, canvas_singleton->shader.canvas_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, canvas_singleton->shader.canvas_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET); } RendererCanvasRenderRD::MaterialData::~MaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *RendererCanvasRenderRD::_create_material_func(ShaderData *p_shader) { @@ -2488,6 +2349,9 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { actions.renames["CANVAS_MATRIX"] = "canvas_data.canvas_transform"; actions.renames["SCREEN_MATRIX"] = "canvas_data.screen_transform"; actions.renames["TIME"] = "canvas_data.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); actions.renames["AT_LIGHT_PASS"] = "false"; actions.renames["INSTANCE_CUSTOM"] = "instance_custom"; @@ -2710,8 +2574,19 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { default_canvas_group_shader = storage->shader_allocate(); storage->shader_initialize(default_canvas_group_shader); - storage->shader_set_code(default_canvas_group_shader, "shader_type canvas_item; \nvoid fragment() {\n\tvec4 c = textureLod(SCREEN_TEXTURE,SCREEN_UV,0.0); if (c.a > 0.0001) c.rgb/=c.a; COLOR *= c; \n}\n"); + storage->shader_set_code(default_canvas_group_shader, R"( +shader_type canvas_item; + +void fragment() { + vec4 c = textureLod(SCREEN_TEXTURE, SCREEN_UV, 0.0); + if (c.a > 0.0001) { + c.rgb /= c.a; + } + + COLOR *= c; +} +)"); default_canvas_group_material = storage->material_allocate(); storage->material_initialize(default_canvas_group_material); diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h index cb947d7180..7c4f62832c 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h @@ -67,12 +67,10 @@ class RendererCanvasRenderRD : public RendererCanvasRender { }; enum { - FLAGS_INSTANCING_STRIDE_MASK = 0xF, - FLAGS_INSTANCING_ENABLED = (1 << 4), - FLAGS_INSTANCING_HAS_COLORS = (1 << 5), - FLAGS_INSTANCING_COLOR_8BIT = (1 << 6), - FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 7), - FLAGS_INSTANCING_CUSTOM_DATA_8_BIT = (1 << 8), + + FLAGS_INSTANCING_MASK = 0x7F, + FLAGS_INSTANCING_HAS_COLORS = (1 << 7), + FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 8), FLAGS_CLIP_RECT_UV = (1 << 9), FLAGS_TRANSPOSE_RECT = (1 << 10), @@ -178,6 +176,7 @@ class RendererCanvasRenderRD : public RendererCanvasRender { bool uses_screen_texture = false; bool uses_sdf = false; + bool uses_time = false; virtual void set_code(const String &p_Code); virtual void set_default_texture_param(const StringName &p_name, RID p_texture); @@ -202,14 +201,11 @@ class RendererCanvasRenderRD : public RendererCanvasRender { struct MaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; ShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~MaterialData(); }; @@ -427,14 +423,14 @@ class RendererCanvasRenderRD : public RendererCanvasRender { RID _create_base_uniform_set(RID p_to_render_target, bool p_backbuffer); inline void _bind_canvas_texture(RD::DrawListID p_draw_list, RID p_texture, RS::CanvasItemTextureFilter p_base_filter, RS::CanvasItemTextureRepeat p_base_repeat, RID &r_last_texture, PushConstant &push_constant, Size2 &r_texpixel_size); //recursive, so regular inline used instead. - void _render_item(RenderingDevice::DrawListID p_draw_list, const Item *p_item, RenderingDevice::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants); + void _render_item(RenderingDevice::DrawListID p_draw_list, RID p_render_target, const Item *p_item, RenderingDevice::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants); void _render_items(RID p_to_render_target, int p_item_count, const Transform2D &p_canvas_transform_inverse, Light *p_lights, bool p_to_backbuffer = false); _FORCE_INLINE_ void _update_transform_2d_to_mat2x4(const Transform2D &p_transform, float *p_mat2x4); _FORCE_INLINE_ void _update_transform_2d_to_mat2x3(const Transform2D &p_transform, float *p_mat2x3); _FORCE_INLINE_ void _update_transform_2d_to_mat4(const Transform2D &p_transform, float *p_mat4); - _FORCE_INLINE_ void _update_transform_to_mat4(const Transform &p_transform, float *p_mat4); + _FORCE_INLINE_ void _update_transform_to_mat4(const Transform3D &p_transform, float *p_mat4); void _update_shadow_atlas(); @@ -458,8 +454,6 @@ public: void canvas_debug_viewport_shadows(Light *p_lights_with_shadow) {} - void draw_window_margins(int *p_margins, RID *p_margin_textures) {} - virtual void set_shadow_texture_size(int p_size); void set_time(double p_time); diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index 2247b841c9..62e9386f95 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -31,6 +31,7 @@ #include "renderer_compositor_rd.h" #include "core/config/project_settings.h" +#include "core/io/dir_access.h" void RendererCompositorRD::prepare_for_blitting_render_targets() { RD::get_singleton()->prepare_screen_for_drawing(); @@ -44,32 +45,39 @@ void RendererCompositorRD::blit_render_targets_to_screen(DisplayServer::WindowID ERR_CONTINUE(texture.is_null()); RID rd_texture = storage->texture_get_rd_texture(texture); ERR_CONTINUE(rd_texture.is_null()); + if (!render_target_descriptors.has(rd_texture) || !RD::get_singleton()->uniform_set_is_valid(render_target_descriptors[rd_texture])) { Vector<RD::Uniform> uniforms; RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; u.binding = 0; - u.ids.push_back(copy_viewports_sampler); + u.ids.push_back(blit.sampler); u.ids.push_back(rd_texture); uniforms.push_back(u); - RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, copy_viewports_rd_shader, 0); + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, blit.shader.version_get_shader(blit.shader_version, BLIT_MODE_NORMAL), 0); render_target_descriptors[rd_texture] = uniform_set; } Size2 screen_size(RD::get_singleton()->screen_get_width(p_screen), RD::get_singleton()->screen_get_height(p_screen)); - - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_viewports_rd_pipeline); - RD::get_singleton()->draw_list_bind_index_array(draw_list, copy_viewports_rd_array); + BlitMode mode = p_render_targets[i].lens_distortion.apply ? BLIT_MODE_LENS : (p_render_targets[i].multi_view.use_layer ? BLIT_MODE_USE_LAYER : BLIT_MODE_NORMAL); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[mode]); + RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_target_descriptors[rd_texture], 0); - float push_constant[4] = { - p_render_targets[i].rect.position.x / screen_size.width, - p_render_targets[i].rect.position.y / screen_size.height, - p_render_targets[i].rect.size.width / screen_size.width, - p_render_targets[i].rect.size.height / screen_size.height, - }; - RD::get_singleton()->draw_list_set_push_constant(draw_list, push_constant, 4 * sizeof(float)); + blit.push_constant.rect[0] = p_render_targets[i].rect.position.x / screen_size.width; + blit.push_constant.rect[1] = p_render_targets[i].rect.position.y / screen_size.height; + blit.push_constant.rect[2] = p_render_targets[i].rect.size.width / screen_size.width; + blit.push_constant.rect[3] = p_render_targets[i].rect.size.height / screen_size.height; + blit.push_constant.layer = p_render_targets[i].multi_view.layer; + blit.push_constant.eye_center[0] = p_render_targets[i].lens_distortion.eye_center.x; + blit.push_constant.eye_center[1] = p_render_targets[i].lens_distortion.eye_center.y; + blit.push_constant.k1 = p_render_targets[i].lens_distortion.k1; + blit.push_constant.k2 = p_render_targets[i].lens_distortion.k2; + blit.push_constant.upscale = p_render_targets[i].lens_distortion.upscale; + blit.push_constant.aspect_ratio = p_render_targets[i].lens_distortion.aspect_ratio; + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); RD::get_singleton()->draw_list_draw(draw_list, true); } @@ -96,40 +104,23 @@ void RendererCompositorRD::end_frame(bool p_swap_buffers) { } void RendererCompositorRD::initialize() { - { //create framebuffer copy shader - RenderingDevice::ShaderStageData vert; - vert.shader_stage = RenderingDevice::SHADER_STAGE_VERTEX; - vert.spir_v = RenderingDevice::get_singleton()->shader_compile_from_source(RenderingDevice::SHADER_STAGE_VERTEX, - "#version 450\n" - "layout(push_constant, binding = 0, std140) uniform Pos { vec4 dst_rect; } pos;\n" - "layout(location =0) out vec2 uv;\n" - "void main() { \n" - " vec2 base_arr[4] = vec2[](vec2(0.0,0.0),vec2(0.0,1.0),vec2(1.0,1.0),vec2(1.0,0.0));\n" - " uv = base_arr[gl_VertexIndex];\n" - " vec2 vtx = pos.dst_rect.xy+uv*pos.dst_rect.zw;\n" - " gl_Position = vec4(vtx * 2.0 - 1.0,0.0,1.0);\n" - "}\n"); - - RenderingDevice::ShaderStageData frag; - frag.shader_stage = RenderingDevice::SHADER_STAGE_FRAGMENT; - frag.spir_v = RenderingDevice::get_singleton()->shader_compile_from_source(RenderingDevice::SHADER_STAGE_FRAGMENT, - "#version 450\n" - "layout (location = 0) in vec2 uv;\n" - "layout (location = 0) out vec4 color;\n" - "layout (binding = 0) uniform sampler2D src_rt;\n" - "void main() { color=texture(src_rt,uv); }\n"); - - Vector<RenderingDevice::ShaderStageData> source; - source.push_back(vert); - source.push_back(frag); - String error; - copy_viewports_rd_shader = RD::get_singleton()->shader_create(source); - if (!copy_viewports_rd_shader.is_valid()) { - print_line("Failed compilation: " + error); + { + // Initialize blit + Vector<String> blit_modes; + blit_modes.push_back("\n"); + blit_modes.push_back("\n#define USE_LAYER\n"); + blit_modes.push_back("\n#define USE_LAYER\n#define APPLY_LENS_DISTORTION\n"); + blit_modes.push_back("\n"); + + blit.shader.initialize(blit_modes); + + blit.shader_version = blit.shader.version_create(); + + for (int i = 0; i < BLIT_MODE_MAX; i++) { + blit.pipelines[i] = RD::get_singleton()->render_pipeline_create(blit.shader.version_get_shader(blit.shader_version, i), RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), i == BLIT_MODE_NORMAL_ALPHA ? RenderingDevice::PipelineColorBlendState::create_blend() : RenderingDevice::PipelineColorBlendState::create_disabled(), 0); } - } - { //create index array for copy shader + //create index array for copy shader Vector<uint8_t> pv; pv.resize(6 * 4); { @@ -142,15 +133,10 @@ void RendererCompositorRD::initialize() { p32[4] = 2; p32[5] = 3; } - copy_viewports_rd_index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); - copy_viewports_rd_array = RD::get_singleton()->index_array_create(copy_viewports_rd_index_buffer, 0, 6); - } + blit.index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + blit.array = RD::get_singleton()->index_array_create(blit.index_buffer, 0, 6); - { //pipeline - copy_viewports_rd_pipeline = RD::get_singleton()->render_pipeline_create(copy_viewports_rd_shader, RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RenderingDevice::PipelineColorBlendState::create_disabled(), 0); - } - { // sampler - copy_viewports_sampler = RD::get_singleton()->sampler_create(RD::SamplerState()); + blit.sampler = RD::get_singleton()->sampler_create(RD::SamplerState()); } } @@ -162,18 +148,145 @@ void RendererCompositorRD::finalize() { memdelete(storage); //only need to erase these, the rest are erased by cascade - RD::get_singleton()->free(copy_viewports_rd_index_buffer); - RD::get_singleton()->free(copy_viewports_rd_shader); - RD::get_singleton()->free(copy_viewports_sampler); + blit.shader.version_free(blit.shader_version); + RD::get_singleton()->free(blit.index_buffer); + RD::get_singleton()->free(blit.sampler); +} + +void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) { + RD::get_singleton()->prepare_screen_for_drawing(); + + RID texture = storage->texture_allocate(); + storage->texture_2d_initialize(texture, p_image); + RID rd_texture = storage->texture_get_rd_texture(texture); + + RID uset; + { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.ids.push_back(blit.sampler); + u.ids.push_back(rd_texture); + uniforms.push_back(u); + uset = RD::get_singleton()->uniform_set_create(uniforms, blit.shader.version_get_shader(blit.shader_version, BLIT_MODE_NORMAL), 0); + } + + Size2 window_size = DisplayServer::get_singleton()->window_get_size(); + + Rect2 imgrect(0, 0, p_image->get_width(), p_image->get_height()); + Rect2 screenrect; + if (p_scale) { + if (window_size.width > window_size.height) { + //scale horizontally + screenrect.size.y = window_size.height; + screenrect.size.x = imgrect.size.x * window_size.height / imgrect.size.y; + screenrect.position.x = (window_size.width - screenrect.size.x) / 2; + + } else { + //scale vertically + screenrect.size.x = window_size.width; + screenrect.size.y = imgrect.size.y * window_size.width / imgrect.size.x; + screenrect.position.y = (window_size.height - screenrect.size.y) / 2; + } + } else { + screenrect = imgrect; + screenrect.position += ((Size2(window_size.width, window_size.height) - screenrect.size) / 2.0).floor(); + } + + screenrect.position /= window_size; + screenrect.size /= window_size; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin_for_screen(DisplayServer::MAIN_WINDOW_ID, p_color); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[BLIT_MODE_NORMAL_ALPHA]); + RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uset, 0); + + blit.push_constant.rect[0] = screenrect.position.x; + blit.push_constant.rect[1] = screenrect.position.y; + blit.push_constant.rect[2] = screenrect.size.width; + blit.push_constant.rect[3] = screenrect.size.height; + blit.push_constant.layer = 0; + blit.push_constant.eye_center[0] = 0; + blit.push_constant.eye_center[1] = 0; + blit.push_constant.k1 = 0; + blit.push_constant.k2 = 0; + blit.push_constant.upscale = 1.0; + blit.push_constant.aspect_ratio = 1.0; + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + + RD::get_singleton()->draw_list_end(); + + RD::get_singleton()->swap_buffers(); + + storage->free(texture); } RendererCompositorRD *RendererCompositorRD::singleton = nullptr; RendererCompositorRD::RendererCompositorRD() { + { + String shader_cache_dir = Engine::get_singleton()->get_shader_cache_path(); + if (shader_cache_dir == String()) { + shader_cache_dir = "user://"; + } + DirAccessRef da = DirAccess::open(shader_cache_dir); + if (!da) { + ERR_PRINT("Can't create shader cache folder, no shader caching will happen: " + shader_cache_dir); + } else { + Error err = da->change_dir("shader_cache"); + if (err != OK) { + err = da->make_dir("shader_cache"); + } + if (err != OK) { + ERR_PRINT("Can't create shader cache folder, no shader caching will happen: " + shader_cache_dir); + } else { + shader_cache_dir = shader_cache_dir.plus_file("shader_cache"); + + bool shader_cache_enabled = GLOBAL_GET("rendering/shader_compiler/shader_cache/enabled"); + if (!Engine::get_singleton()->is_editor_hint() && !shader_cache_enabled) { + shader_cache_dir = String(); //disable only if not editor + } + + if (shader_cache_dir != String()) { + bool compress = GLOBAL_GET("rendering/shader_compiler/shader_cache/compress"); + bool use_zstd = GLOBAL_GET("rendering/shader_compiler/shader_cache/use_zstd_compression"); + bool strip_debug = GLOBAL_GET("rendering/shader_compiler/shader_cache/strip_debug"); + + ShaderRD::set_shader_cache_dir(shader_cache_dir); + ShaderRD::set_shader_cache_save_compressed(compress); + ShaderRD::set_shader_cache_save_compressed_zstd(use_zstd); + ShaderRD::set_shader_cache_save_debug(!strip_debug); + } + } + } + } + singleton = this; time = 0; storage = memnew(RendererStorageRD); canvas = memnew(RendererCanvasRenderRD(storage)); - scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered(storage)); + + uint32_t back_end = GLOBAL_GET("rendering/vulkan/rendering/back_end"); + uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); + + if (back_end == 1 || textures_per_stage < 48) { + scene = memnew(RendererSceneRenderImplementation::RenderForwardMobile(storage)); + } else { // back_end == 0 + // default to our high end renderer + scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered(storage)); + } + + scene->init(); + + // now we're ready to create our effects, + storage->init_effects(!scene->_render_buffers_can_be_storage()); +} + +RendererCompositorRD::~RendererCompositorRD() { + ShaderRD::set_shader_cache_dir(String()); } diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index 5b5f3ad0cb..8639362da9 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -35,8 +35,10 @@ #include "core/templates/thread_work_pool.h" #include "servers/rendering/renderer_compositor.h" #include "servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h" +#include "servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h" #include "servers/rendering/renderer_rd/renderer_canvas_render_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" +#include "servers/rendering/renderer_rd/shaders/blit.glsl.gen.h" class RendererCompositorRD : public RendererCompositor { protected: @@ -44,16 +46,41 @@ protected: RendererStorageRD *storage; RendererSceneRenderRD *scene; - RID copy_viewports_rd_shader; - RID copy_viewports_rd_pipeline; - RID copy_viewports_rd_index_buffer; - RID copy_viewports_rd_array; - RID copy_viewports_sampler; + enum BlitMode { + BLIT_MODE_NORMAL, + BLIT_MODE_USE_LAYER, + BLIT_MODE_LENS, + BLIT_MODE_NORMAL_ALPHA, + BLIT_MODE_MAX + }; + + struct BlitPushConstant { + float rect[4]; + + float eye_center[2]; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint32_t layer; + uint32_t pad1; + }; + + struct Blit { + BlitPushConstant push_constant; + BlitShaderRD shader; + RID shader_version; + RID pipelines[BLIT_MODE_MAX]; + RID index_buffer; + RID array; + RID sampler; + } blit; Map<RID, RID> render_target_descriptors; double time; - float delta; + double delta; static uint64_t frame; @@ -62,7 +89,7 @@ public: RendererCanvasRender *get_canvas() { return canvas; } RendererSceneRender *get_scene() { return scene; } - void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) {} + void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter); void initialize(); void begin_frame(double frame_step); @@ -73,7 +100,7 @@ public: void finalize(); _ALWAYS_INLINE_ uint64_t get_frame_number() const { return frame; } - _ALWAYS_INLINE_ float get_frame_delta_time() const { return delta; } + _ALWAYS_INLINE_ double get_frame_delta_time() const { return delta; } _ALWAYS_INLINE_ double get_total_time() const { return time; } static Error is_viable() { @@ -92,6 +119,6 @@ public: static RendererCompositorRD *singleton; RendererCompositorRD(); - ~RendererCompositorRD() {} + ~RendererCompositorRD(); }; #endif // RASTERIZER_RD_H diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp index 3856f38457..098e2a5c87 100644 --- a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp @@ -1099,7 +1099,7 @@ void RendererSceneGIRD::SDFGI::update_cascades() { RD::get_singleton()->buffer_update(cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, RD::BARRIER_MASK_COMPUTE); } -void RendererSceneGIRD::SDFGI::debug_draw(const CameraMatrix &p_projection, const Transform &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture) { +void RendererSceneGIRD::SDFGI::debug_draw(const CameraMatrix &p_projection, const Transform3D &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture) { if (!debug_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(debug_uniform_set)) { Vector<RD::Uniform> uniforms; { @@ -1367,7 +1367,7 @@ void RendererSceneGIRD::SDFGI::debug_probes(RD::DrawListID p_draw_list, RID p_fr } } -void RendererSceneGIRD::SDFGI::pre_process_gi(const Transform &p_transform, RendererSceneRenderRD *p_scene_render) { +void RendererSceneGIRD::SDFGI::pre_process_gi(const Transform3D &p_transform, RenderDataRD *p_render_data, RendererSceneRenderRD *p_scene_render) { /* Update general SDFGI Buffer */ SDFGIData sdfgi_data; @@ -1564,7 +1564,7 @@ void RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, //clear dispatch indirect data SDFGIShader::PreprocessPushConstant push_constant; - zeromem(&push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + memset(&push_constant, 0, sizeof(SDFGIShader::PreprocessPushConstant)); RENDER_TIMESTAMP("Scroll SDF"); @@ -1862,7 +1862,7 @@ void RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(cascades[cascade].sdf, 0); Ref<Image> img; - img.instance(); + img.instantiate(); for (uint32_t i = 0; i < cascade_size; i++) { Vector<uint8_t> subarr = data.subarray(128 * 128 * i, 128 * 128 * (i + 1) - 1); img->create(cascade_size, cascade_size, false, Image::FORMAT_L8, subarr); @@ -1875,7 +1875,7 @@ void RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(render_albedo, 0); Ref<Image> img; - img.instance(); + img.instantiate(); for (uint32_t i = 0; i < cascade_size; i++) { Vector<uint8_t> subarr = data.subarray(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2 - 1); img->createcascade_size, cascade_size, false, Image::FORMAT_RGB565, subarr); @@ -2012,10 +2012,10 @@ void RendererSceneGIRD::SDFGI::render_static_lights(RID p_render_buffers, uint32 } //////////////////////////////////////////////////////////////////////////////// -// GIProbeInstance +// VoxelGIInstance -void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { - uint32_t data_version = storage->gi_probe_get_data_version(probe); +void RendererSceneGIRD::VoxelGIInstance::update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { + uint32_t data_version = storage->voxel_gi_get_data_version(probe); // (RE)CREATE IF NEEDED @@ -2034,11 +2034,11 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c dynamic_maps.clear(); - Vector3i octree_size = storage->gi_probe_get_octree_size(probe); + Vector3i octree_size = storage->voxel_gi_get_octree_size(probe); if (octree_size != Vector3i()) { //can create a 3D texture - Vector<int> levels = storage->gi_probe_get_level_counts(probe); + Vector<int> levels = storage->voxel_gi_get_level_counts(probe); RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; @@ -2064,7 +2064,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c } for (int i = 0; i < levels.size(); i++) { - GIProbeInstance::Mipmap mipmap; + VoxelGIInstance::Mipmap mipmap; mipmap.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), texture, 0, i, RD::TEXTURE_SLICE_3D); mipmap.level = levels.size() - i - 1; mipmap.cell_offset = 0; @@ -2078,14 +2078,14 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 1; - u.ids.push_back(storage->gi_probe_get_octree_buffer(probe)); + u.ids.push_back(storage->voxel_gi_get_octree_buffer(probe)); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 2; - u.ids.push_back(storage->gi_probe_get_data_buffer(probe)); + u.ids.push_back(storage->voxel_gi_get_data_buffer(probe)); uniforms.push_back(u); } @@ -2100,7 +2100,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + u.ids.push_back(storage->voxel_gi_get_sdf_texture(probe)); uniforms.push_back(u); } { @@ -2118,11 +2118,11 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 3; - u.ids.push_back(gi->gi_probe_lights_uniform); + u.ids.push_back(gi->voxel_gi_lights_uniform); copy_uniforms.push_back(u); } - mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT], 0); + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT], 0); copy_uniforms = uniforms; //restore @@ -2133,9 +2133,9 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c u.ids.push_back(texture); copy_uniforms.push_back(u); } - mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0); + mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0); } else { - mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP], 0); + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP], 0); } } @@ -2147,7 +2147,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c uniforms.push_back(u); } - mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE], 0); + mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE], 0); mipmaps.push_back(mipmap); } @@ -2158,7 +2158,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c int mipmap_index = 0; while (mipmap_index < mipmaps.size()) { - GIProbeInstance::DynamicMap dmap; + VoxelGIInstance::DynamicMap dmap; if (oversample > 0) { dmap.size = dynamic_map_size * (1 << oversample); @@ -2217,7 +2217,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 3; - u.ids.push_back(gi->gi_probe_lights_uniform); + u.ids.push_back(gi->voxel_gi_lights_uniform); uniforms.push_back(u); } @@ -2253,7 +2253,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + u.ids.push_back(storage->voxel_gi_get_sdf_texture(probe)); uniforms.push_back(u); } { @@ -2278,7 +2278,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c uniforms.push_back(u); } - dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING], 0); + dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING], 0); } } else { bool plot = dmap.mipmap >= 0; @@ -2322,7 +2322,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + u.ids.push_back(storage->voxel_gi_get_sdf_texture(probe)); uniforms.push_back(u); } { @@ -2345,7 +2345,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c dmap.uniform_set = RD::get_singleton()->uniform_set_create( uniforms, - gi->giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT)], + gi->voxel_gi_lighting_shader_version_shaders[(write && plot) ? VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT)], 0); } @@ -2370,15 +2370,15 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c uint32_t light_count = 0; if (p_update_light_instances || p_dynamic_objects.size() > 0) { - light_count = MIN(gi->gi_probe_max_lights, (uint32_t)p_light_instances.size()); + light_count = MIN(gi->voxel_gi_max_lights, (uint32_t)p_light_instances.size()); { - Transform to_cell = storage->gi_probe_get_to_cell_xform(probe); - Transform to_probe_xform = (transform * to_cell.affine_inverse()).affine_inverse(); + Transform3D to_cell = storage->voxel_gi_get_to_cell_xform(probe); + Transform3D to_probe_xform = (transform * to_cell.affine_inverse()).affine_inverse(); //update lights for (uint32_t i = 0; i < light_count; i++) { - GIProbeLight &l = gi->gi_probe_lights[i]; + VoxelGILight &l = gi->voxel_gi_lights[i]; RID light_instance = p_light_instances[i]; RID light = p_scene_render->light_instance_get_base_light(light_instance); @@ -2399,7 +2399,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c l.cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); l.inv_spot_attenuation = 1.0f / storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); - Transform xform = p_scene_render->light_instance_get_base_transform(light_instance); + Transform3D xform = p_scene_render->light_instance_get_base_transform(light_instance); Vector3 pos = to_probe_xform.xform(xform.origin); Vector3 dir = to_probe_xform.basis.xform(-xform.basis.get_axis(2)).normalized(); @@ -2415,7 +2415,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c l.has_shadow = storage->light_has_shadow(light); } - RD::get_singleton()->buffer_update(gi->gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi->gi_probe_lights); + RD::get_singleton()->buffer_update(gi->voxel_gi_lights_uniform, 0, sizeof(VoxelGILight) * light_count, gi->voxel_gi_lights); } } @@ -2424,17 +2424,17 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c if (mipmaps.size()) { //can update mipmaps - Vector3i probe_size = storage->gi_probe_get_octree_size(probe); + Vector3i probe_size = storage->voxel_gi_get_octree_size(probe); - GIProbePushConstant push_constant; + VoxelGIPushConstant push_constant; push_constant.limits[0] = probe_size.x; push_constant.limits[1] = probe_size.y; push_constant.limits[2] = probe_size.z; push_constant.stack_size = mipmaps.size(); push_constant.emission_scale = 1.0; - push_constant.propagation = storage->gi_probe_get_propagation(probe); - push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(probe); + push_constant.propagation = storage->voxel_gi_get_propagation(probe); + push_constant.dynamic_range = storage->voxel_gi_get_dynamic_range(probe); push_constant.light_count = light_count; push_constant.aniso_strength = 0; @@ -2446,7 +2446,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c int passes; if (p_update_light_instances) { - passes = storage->gi_probe_is_using_two_bounces(probe) ? 2 : 1; + passes = storage->voxel_gi_is_using_two_bounces(probe) ? 2 : 1; } else { passes = 1; //only re-blitting is necessary } @@ -2457,9 +2457,9 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c if (p_update_light_instances) { for (int i = 0; i < mipmaps.size(); i++) { if (i == 0) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[pass == 0 ? GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT : GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[pass == 0 ? VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT : VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]); } else if (i == 1) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP]); } if (pass == 1 || i > 0) { @@ -2477,7 +2477,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c int wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; while (wg_todo) { int wg_count = MIN(wg_todo, wg_limit_x); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIPushConstant)); RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1); wg_todo -= wg_count; push_constant.cell_offset += wg_count * wg_size; @@ -2487,7 +2487,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::get_singleton()->compute_list_add_barrier(compute_list); //wait til previous step is done } - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE]); for (int i = 0; i < mipmaps.size(); i++) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mipmaps[i].write_uniform_set, 0); @@ -2498,7 +2498,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c int wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; while (wg_todo) { int wg_count = MIN(wg_todo, wg_limit_x); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIPushConstant)); RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1); wg_todo -= wg_count; push_constant.cell_offset += wg_count * wg_size; @@ -2513,15 +2513,15 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c has_dynamic_object_data = false; //clear until dynamic object data is used again if (p_dynamic_objects.size() && dynamic_maps.size()) { - Vector3i octree_size = storage->gi_probe_get_octree_size(probe); + Vector3i octree_size = storage->voxel_gi_get_octree_size(probe); int multiplier = dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z); - Transform oversample_scale; + Transform3D oversample_scale; oversample_scale.basis.scale(Vector3(multiplier, multiplier, multiplier)); - Transform to_cell = oversample_scale * storage->gi_probe_get_to_cell_xform(probe); - Transform to_world_xform = transform * to_cell.affine_inverse(); - Transform to_probe_xform = to_world_xform.affine_inverse(); + Transform3D to_cell = oversample_scale * storage->voxel_gi_get_to_cell_xform(probe); + Transform3D to_world_xform = transform * to_cell.affine_inverse(); + Transform3D to_probe_xform = to_world_xform.affine_inverse(); AABB probe_aabb(Vector3(), octree_size); @@ -2529,7 +2529,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c for (int i = 0; i < (int)p_dynamic_objects.size(); i++) { RendererSceneRender::GeometryInstance *instance = p_dynamic_objects[i]; - //transform aabb to giprobe + //transform aabb to voxel_gi AABB aabb = (to_probe_xform * p_scene_render->geometry_instance_get_transform(instance)).xform(p_scene_render->geometry_instance_get_aabb(instance)); //this needs to wrap to grid resolution to avoid jitter @@ -2576,7 +2576,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c Vector3 up_dir = render_up[j]; Vector3 center = aabb.position + aabb.size * 0.5; - Transform xform; + Transform3D xform; xform.set_look_at(center - aabb.size * 0.5 * render_dir, center, up_dir); Vector3 x_dir = xform.basis.get_axis(0).abs(); @@ -2601,8 +2601,8 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c p_scene_render->_render_material(to_world_xform * xform, cm, true, p_scene_render->cull_argument, dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size)); - GIProbeDynamicPushConstant push_constant; - zeromem(&push_constant, sizeof(GIProbeDynamicPushConstant)); + VoxelGIDynamicPushConstant push_constant; + memset(&push_constant, 0, sizeof(VoxelGIDynamicPushConstant)); push_constant.limits[0] = octree_size.x; push_constant.limits[1] = octree_size.y; push_constant.limits[2] = octree_size.z; @@ -2619,7 +2619,7 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c push_constant.z_base = xform.origin[z_axis]; push_constant.z_sign = (z_flip ? -1.0 : 1.0); push_constant.pos_multiplier = float(1.0) / multiplier; - push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(probe); + push_constant.dynamic_range = storage->voxel_gi_get_dynamic_range(probe); push_constant.flip_x = x_flip; push_constant.flip_y = y_flip; push_constant.rect_pos[0] = rect.position[0]; @@ -2631,16 +2631,16 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c push_constant.prev_rect_size[0] = 0; push_constant.prev_rect_size[1] = 0; push_constant.on_mipmap = false; - push_constant.propagation = storage->gi_probe_get_propagation(probe); + push_constant.propagation = storage->voxel_gi_get_propagation(probe); push_constant.pad[0] = 0; push_constant.pad[1] = 0; push_constant.pad[2] = 0; //process lighting RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, dynamic_maps[0].uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant)); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIDynamicPushConstant)); RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1); //print_line("rect: " + itos(i) + ": " + rect); @@ -2695,14 +2695,14 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c RD::get_singleton()->compute_list_add_barrier(compute_list); if (dynamic_maps[k].mipmap < 0) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]); } else if (k < dynamic_maps.size() - 1) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]); } else { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT]); } RD::get_singleton()->compute_list_bind_uniform_set(compute_list, dynamic_maps[k].uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant)); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIDynamicPushConstant)); RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1); } @@ -2713,22 +2713,22 @@ void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, c has_dynamic_object_data = true; //clear until dynamic object data is used again } - last_probe_version = storage->gi_probe_get_version(probe); + last_probe_version = storage->voxel_gi_get_version(probe); } -void RendererSceneGIRD::GIProbeInstance::debug(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { +void RendererSceneGIRD::VoxelGIInstance::debug(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { if (mipmaps.size() == 0) { return; } - CameraMatrix cam_transform = (p_camera_with_transform * CameraMatrix(transform)) * CameraMatrix(storage->gi_probe_get_to_cell_xform(probe).affine_inverse()); + CameraMatrix cam_transform = (p_camera_with_transform * CameraMatrix(transform)) * CameraMatrix(storage->voxel_gi_get_to_cell_xform(probe).affine_inverse()); int level = 0; - Vector3i octree_size = storage->gi_probe_get_octree_size(probe); + Vector3i octree_size = storage->voxel_gi_get_octree_size(probe); - GIProbeDebugPushConstant push_constant; + VoxelGIDebugPushConstant push_constant; push_constant.alpha = p_alpha; - push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(probe); + push_constant.dynamic_range = storage->voxel_gi_get_dynamic_range(probe); push_constant.cell_offset = mipmaps[level].cell_offset; push_constant.level = level; @@ -2743,15 +2743,15 @@ void RendererSceneGIRD::GIProbeInstance::debug(RD::DrawListID p_draw_list, RID p } } - if (gi->giprobe_debug_uniform_set.is_valid()) { - RD::get_singleton()->free(gi->giprobe_debug_uniform_set); + if (gi->voxel_gi_debug_uniform_set.is_valid()) { + RD::get_singleton()->free(gi->voxel_gi_debug_uniform_set); } Vector<RD::Uniform> uniforms; { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 1; - u.ids.push_back(storage->gi_probe_get_data_buffer(probe)); + u.ids.push_back(storage->voxel_gi_get_data_buffer(probe)); uniforms.push_back(u); } { @@ -2776,19 +2776,19 @@ void RendererSceneGIRD::GIProbeInstance::debug(RD::DrawListID p_draw_list, RID p cell_count = mipmaps[level].cell_count; } - gi->giprobe_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_debug_shader_version_shaders[0], 0); + gi->voxel_gi_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_debug_shader_version_shaders[0], 0); - int giprobe_debug_pipeline = GI_PROBE_DEBUG_COLOR; + int voxel_gi_debug_pipeline = VOXEL_GI_DEBUG_COLOR; if (p_emission) { - giprobe_debug_pipeline = GI_PROBE_DEBUG_EMISSION; + voxel_gi_debug_pipeline = VOXEL_GI_DEBUG_EMISSION; } else if (p_lighting) { - giprobe_debug_pipeline = has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT; + voxel_gi_debug_pipeline = has_dynamic_object_data ? VOXEL_GI_DEBUG_LIGHT_FULL : VOXEL_GI_DEBUG_LIGHT; } RD::get_singleton()->draw_list_bind_render_pipeline( p_draw_list, - gi->giprobe_debug_shader_version_pipelines[giprobe_debug_pipeline].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); - RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, gi->giprobe_debug_uniform_set, 0); - RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(GIProbeDebugPushConstant)); + gi->voxel_gi_debug_shader_version_pipelines[voxel_gi_debug_pipeline].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, gi->voxel_gi_debug_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(VoxelGIDebugPushConstant)); RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36); } @@ -2812,13 +2812,11 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p { //kinda complicated to compute the amount of slots, we try to use as many as we can - gi_probe_max_lights = 32; + voxel_gi_lights = memnew_arr(VoxelGILight, voxel_gi_max_lights); + voxel_gi_lights_uniform = RD::get_singleton()->uniform_buffer_create(voxel_gi_max_lights * sizeof(VoxelGILight)); + voxel_gi_quality = RS::VoxelGIQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/voxel_gi/quality")), 0, 1)); - gi_probe_lights = memnew_arr(GIProbeLight, gi_probe_max_lights); - gi_probe_lights_uniform = RD::get_singleton()->uniform_buffer_create(gi_probe_max_lights * sizeof(GIProbeLight)); - gi_probe_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/gi_probes/quality")), 0, 1)); - - String defines = "\n#define MAX_LIGHTS " + itos(gi_probe_max_lights) + "\n"; + String defines = "\n#define MAX_LIGHTS " + itos(voxel_gi_max_lights) + "\n"; Vector<String> versions; versions.push_back("\n#define MODE_COMPUTE_LIGHT\n"); @@ -2830,11 +2828,11 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n"); versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n#define MODE_DYNAMIC_SHRINK_WRITE\n"); - giprobe_shader.initialize(versions, defines); - giprobe_lighting_shader_version = giprobe_shader.version_create(); - for (int i = 0; i < GI_PROBE_SHADER_VERSION_MAX; i++) { - giprobe_lighting_shader_version_shaders[i] = giprobe_shader.version_get_shader(giprobe_lighting_shader_version, i); - giprobe_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(giprobe_lighting_shader_version_shaders[i]); + voxel_gi_shader.initialize(versions, defines); + voxel_gi_lighting_shader_version = voxel_gi_shader.version_create(); + for (int i = 0; i < VOXEL_GI_SHADER_VERSION_MAX; i++) { + voxel_gi_lighting_shader_version_shaders[i] = voxel_gi_shader.version_get_shader(voxel_gi_lighting_shader_version, i); + voxel_gi_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(voxel_gi_lighting_shader_version_shaders[i]); } } @@ -2846,10 +2844,10 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p versions.push_back("\n#define MODE_DEBUG_EMISSION\n"); versions.push_back("\n#define MODE_DEBUG_LIGHT\n#define MODE_DEBUG_LIGHT_FULL\n"); - giprobe_debug_shader.initialize(versions, defines); - giprobe_debug_shader_version = giprobe_debug_shader.version_create(); - for (int i = 0; i < GI_PROBE_DEBUG_MAX; i++) { - giprobe_debug_shader_version_shaders[i] = giprobe_debug_shader.version_get_shader(giprobe_debug_shader_version, i); + voxel_gi_debug_shader.initialize(versions, defines); + voxel_gi_debug_shader_version = voxel_gi_debug_shader.version_create(); + for (int i = 0; i < VOXEL_GI_DEBUG_MAX; i++) { + voxel_gi_debug_shader_version_shaders[i] = voxel_gi_debug_shader.version_get_shader(voxel_gi_debug_shader_version, i); RD::PipelineRasterizationState rs; rs.cull_mode = RD::POLYGON_CULL_FRONT; @@ -2858,7 +2856,7 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p ds.enable_depth_write = true; ds.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; - giprobe_debug_shader_version_pipelines[i].setup(giprobe_debug_shader_version_shaders[i], RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); + voxel_gi_debug_shader_version_pipelines[i].setup(voxel_gi_debug_shader_version_shaders[i], RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); } } @@ -2944,12 +2942,12 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p //calculate tables String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; Vector<String> gi_modes; - gi_modes.push_back("\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define USE_VOXEL_GI_INSTANCES\n"); gi_modes.push_back("\n#define USE_SDFGI\n"); - gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); - gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_VOXEL_GI_INSTANCES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_VOXEL_GI_INSTANCES\n"); gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n"); - gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_VOXEL_GI_INSTANCES\n"); shader.initialize(gi_modes, defines); shader_version = shader.version_create(); @@ -2991,17 +2989,17 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p } } } - default_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GIProbeData) * MAX_GIPROBES); + default_voxel_gi_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(VoxelGIData) * MAX_VOXEL_GI_INSTANCES); half_resolution = GLOBAL_GET("rendering/global_illumination/gi/use_half_resolution"); } void RendererSceneGIRD::free() { - RD::get_singleton()->free(default_giprobe_buffer); - RD::get_singleton()->free(gi_probe_lights_uniform); + RD::get_singleton()->free(default_voxel_gi_buffer); + RD::get_singleton()->free(voxel_gi_lights_uniform); RD::get_singleton()->free(sdfgi_ubo); - giprobe_debug_shader.version_free(giprobe_debug_shader_version); - giprobe_shader.version_free(giprobe_lighting_shader_version); + voxel_gi_debug_shader.version_free(voxel_gi_debug_shader_version); + voxel_gi_shader.version_free(voxel_gi_lighting_shader_version); shader.version_free(shader_version); sdfgi_shader.debug_probes.version_free(sdfgi_shader.debug_probes_shader); sdfgi_shader.debug.version_free(sdfgi_shader.debug_shader); @@ -3009,7 +3007,9 @@ void RendererSceneGIRD::free() { sdfgi_shader.integrate.version_free(sdfgi_shader.integrate_shader); sdfgi_shader.preprocess.version_free(sdfgi_shader.preprocess_shader); - memdelete_arr(gi_probe_lights); + if (voxel_gi_lights) { + memdelete_arr(voxel_gi_lights); + } } RendererSceneGIRD::SDFGI *RendererSceneGIRD::create_sdfgi(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size) { @@ -3020,36 +3020,34 @@ RendererSceneGIRD::SDFGI *RendererSceneGIRD::create_sdfgi(RendererSceneEnvironme return sdfgi; } -void RendererSceneGIRD::setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used, RendererSceneRenderRD *p_scene_render) { - r_gi_probes_used = 0; +void RendererSceneGIRD::setup_voxel_gi_instances(RID p_render_buffers, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, uint32_t &r_voxel_gi_instances_used, RendererSceneRenderRD *p_scene_render) { + r_voxel_gi_instances_used = 0; // feels a little dirty to use our container this way but.... RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); - RID gi_probe_buffer = p_scene_render->render_buffers_get_gi_probe_buffer(p_render_buffers); - - RD::get_singleton()->draw_command_begin_label("GIProbes Setup"); + RID voxel_gi_buffer = p_scene_render->render_buffers_get_voxel_gi_buffer(p_render_buffers); - GIProbeData gi_probe_data[MAX_GIPROBES]; + VoxelGIData voxel_gi_data[MAX_VOXEL_GI_INSTANCES]; - bool giprobes_changed = false; + bool voxel_gi_instances_changed = false; - Transform to_camera; + Transform3D to_camera; to_camera.origin = p_transform.origin; //only translation, make local - for (int i = 0; i < MAX_GIPROBES; i++) { + for (int i = 0; i < MAX_VOXEL_GI_INSTANCES; i++) { RID texture; - if (i < (int)p_gi_probes.size()) { - GIProbeInstance *gipi = get_probe_instance(p_gi_probes[i]); + if (i < (int)p_voxel_gi_instances.size()) { + VoxelGIInstance *gipi = get_probe_instance(p_voxel_gi_instances[i]); if (gipi) { texture = gipi->texture; - GIProbeData &gipd = gi_probe_data[i]; + VoxelGIData &gipd = voxel_gi_data[i]; RID base_probe = gipi->probe; - Transform to_cell = storage->gi_probe_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera; + Transform3D to_cell = storage->voxel_gi_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera; gipd.xform[0] = to_cell.basis.elements[0][0]; gipd.xform[1] = to_cell.basis.elements[1][0]; @@ -3068,36 +3066,33 @@ void RendererSceneGIRD::setup_giprobes(RID p_render_buffers, const Transform &p_ gipd.xform[14] = to_cell.origin.z; gipd.xform[15] = 1; - Vector3 bounds = storage->gi_probe_get_octree_size(base_probe); + Vector3 bounds = storage->voxel_gi_get_octree_size(base_probe); gipd.bounds[0] = bounds.x; gipd.bounds[1] = bounds.y; gipd.bounds[2] = bounds.z; - gipd.dynamic_range = storage->gi_probe_get_dynamic_range(base_probe) * storage->gi_probe_get_energy(base_probe); - gipd.bias = storage->gi_probe_get_bias(base_probe); - gipd.normal_bias = storage->gi_probe_get_normal_bias(base_probe); - gipd.blend_ambient = !storage->gi_probe_is_interior(base_probe); - gipd.anisotropy_strength = 0; - gipd.ao = storage->gi_probe_get_ao(base_probe); - gipd.ao_size = Math::pow(storage->gi_probe_get_ao_size(base_probe), 4.0f); + gipd.dynamic_range = storage->voxel_gi_get_dynamic_range(base_probe) * storage->voxel_gi_get_energy(base_probe); + gipd.bias = storage->voxel_gi_get_bias(base_probe); + gipd.normal_bias = storage->voxel_gi_get_normal_bias(base_probe); + gipd.blend_ambient = !storage->voxel_gi_is_interior(base_probe); gipd.mipmaps = gipi->mipmaps.size(); } - r_gi_probes_used++; + r_voxel_gi_instances_used++; } if (texture == RID()) { texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); } - if (texture != rb->gi.giprobe_textures[i]) { - giprobes_changed = true; - rb->gi.giprobe_textures[i] = texture; + if (texture != rb->gi.voxel_gi_textures[i]) { + voxel_gi_instances_changed = true; + rb->gi.voxel_gi_textures[i] = texture; } } - if (giprobes_changed) { + if (voxel_gi_instances_changed) { if (RD::get_singleton()->uniform_set_is_valid(rb->gi.uniform_set)) { RD::get_singleton()->free(rb->gi.uniform_set); } @@ -3112,14 +3107,16 @@ void RendererSceneGIRD::setup_giprobes(RID p_render_buffers, const Transform &p_ } } - if (p_gi_probes.size() > 0) { - RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GIProbeData) * MIN((uint64_t)MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, RD::BARRIER_MASK_COMPUTE); - } + if (p_voxel_gi_instances.size() > 0) { + RD::get_singleton()->draw_command_begin_label("VoxelGIs Setup"); - RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->buffer_update(voxel_gi_buffer, 0, sizeof(VoxelGIData) * MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()), voxel_gi_data, RD::BARRIER_MASK_COMPUTE); + + RD::get_singleton()->draw_command_end_label(); + } } -void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, RendererSceneRenderRD *p_scene_render) { +void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, RendererSceneRenderRD *p_scene_render) { RD::get_singleton()->draw_command_begin_label("GI Render"); RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); @@ -3144,8 +3141,6 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ rb->reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); rb->ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); rb->gi.using_half_size_gi = half_resolution; - - p_scene_render->_render_buffers_uniform_set_changed(p_render_buffers); } PushConstant push_constant; @@ -3159,11 +3154,11 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; - push_constant.max_giprobes = MIN((uint64_t)MAX_GIPROBES, p_gi_probes.size()); - push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; + push_constant.max_voxel_gi_instances = MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()); + push_constant.high_quality_vct = voxel_gi_quality == RS::VOXEL_GI_QUALITY_HIGH; bool use_sdfgi = rb->sdfgi != nullptr; - bool use_giprobes = push_constant.max_giprobes > 0; + bool use_voxel_gi_instances = push_constant.max_voxel_gi_instances > 0; if (env) { push_constant.ao_color[0] = env->ao_color.r; @@ -3313,7 +3308,7 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 14; - RID buffer = p_gi_probe_buffer.is_valid() ? p_gi_probe_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID buffer = p_voxel_gi_buffer.is_valid() ? p_voxel_gi_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(buffer); uniforms.push_back(u); } @@ -3328,15 +3323,15 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 16; - u.ids.push_back(rb->gi.giprobe_buffer); + u.ids.push_back(rb->gi.voxel_gi_buffer); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 17; - for (int i = 0; i < MAX_GIPROBES; i++) { - u.ids.push_back(rb->gi.giprobe_textures[i]); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCES; i++) { + u.ids.push_back(rb->gi.voxel_gi_textures[i]); } uniforms.push_back(u); } @@ -3347,10 +3342,11 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ Mode mode; if (rb->gi.using_half_size_gi) { - mode = (use_sdfgi && use_giprobes) ? MODE_HALF_RES_COMBINED : (use_sdfgi ? MODE_HALF_RES_SDFGI : MODE_HALF_RES_GIPROBE); + mode = (use_sdfgi && use_voxel_gi_instances) ? MODE_HALF_RES_COMBINED : (use_sdfgi ? MODE_HALF_RES_SDFGI : MODE_HALF_RES_VOXEL_GI); } else { - mode = (use_sdfgi && use_giprobes) ? MODE_COMBINED : (use_sdfgi ? MODE_SDFGI : MODE_GIPROBE); + mode = (use_sdfgi && use_voxel_gi_instances) ? MODE_COMBINED : (use_sdfgi ? MODE_SDFGI : MODE_VOXEL_GI); } + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipelines[mode]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi.uniform_set, 0); @@ -3366,39 +3362,39 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ RD::get_singleton()->draw_command_end_label(); } -RID RendererSceneGIRD::gi_probe_instance_create(RID p_base) { - GIProbeInstance gi_probe; - gi_probe.gi = this; - gi_probe.storage = storage; - gi_probe.probe = p_base; - RID rid = gi_probe_instance_owner.make_rid(gi_probe); +RID RendererSceneGIRD::voxel_gi_instance_create(RID p_base) { + VoxelGIInstance voxel_gi; + voxel_gi.gi = this; + voxel_gi.storage = storage; + voxel_gi.probe = p_base; + RID rid = voxel_gi_instance_owner.make_rid(voxel_gi); return rid; } -void RendererSceneGIRD::gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) { - GIProbeInstance *gi_probe = get_probe_instance(p_probe); - ERR_FAIL_COND(!gi_probe); +void RendererSceneGIRD::voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) { + VoxelGIInstance *voxel_gi = get_probe_instance(p_probe); + ERR_FAIL_COND(!voxel_gi); - gi_probe->transform = p_xform; + voxel_gi->transform = p_xform; } -bool RendererSceneGIRD::gi_probe_needs_update(RID p_probe) const { - GIProbeInstance *gi_probe = get_probe_instance(p_probe); - ERR_FAIL_COND_V(!gi_probe, false); +bool RendererSceneGIRD::voxel_gi_needs_update(RID p_probe) const { + VoxelGIInstance *voxel_gi = get_probe_instance(p_probe); + ERR_FAIL_COND_V(!voxel_gi, false); - return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe); + return voxel_gi->last_probe_version != storage->voxel_gi_get_version(voxel_gi->probe); } -void RendererSceneGIRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { - GIProbeInstance *gi_probe = get_probe_instance(p_probe); - ERR_FAIL_COND(!gi_probe); +void RendererSceneGIRD::voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { + VoxelGIInstance *voxel_gi = get_probe_instance(p_probe); + ERR_FAIL_COND(!voxel_gi); - gi_probe->update(p_update_light_instances, p_light_instances, p_dynamic_objects, p_scene_render); + voxel_gi->update(p_update_light_instances, p_light_instances, p_dynamic_objects, p_scene_render); } -void RendererSceneGIRD::debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererSceneGIRD::debug_voxel_gi(RID p_voxel_gi, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->debug(p_draw_list, p_framebuffer, p_camera_with_transform, p_lighting, p_emission, p_alpha); + voxel_gi->debug(p_draw_list, p_framebuffer, p_camera_with_transform, p_lighting, p_emission, p_alpha); } diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.h b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h index df20011b23..0b4622646f 100644 --- a/servers/rendering/renderer_rd/renderer_scene_gi_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h @@ -38,26 +38,27 @@ #include "servers/rendering/renderer_rd/renderer_scene_sky_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" #include "servers/rendering/renderer_rd/shaders/gi.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/giprobe.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/giprobe_debug.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/voxel_gi.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl.gen.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering/rendering_device.h" -// Forward declare RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +// Forward declare RenderDataRD and RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +struct RenderDataRD; class RendererSceneRenderRD; class RendererSceneGIRD { private: RendererStorageRD *storage; - /* GIPROBE INSTANCE */ + /* VOXEL_GI INSTANCE */ - struct GIProbeLight { + struct VoxelGILight { uint32_t type; float energy; float radius; @@ -73,7 +74,7 @@ private: uint32_t has_shadow; }; - struct GIProbePushConstant { + struct VoxelGIPushConstant { int32_t limits[3]; uint32_t stack_size; @@ -88,7 +89,7 @@ private: uint32_t pad; }; - struct GIProbeDynamicPushConstant { + struct VoxelGIDynamicPushConstant { int32_t limits[3]; uint32_t light_count; int32_t x_dir[3]; @@ -109,36 +110,36 @@ private: float pad[3]; }; - GIProbeLight *gi_probe_lights; - uint32_t gi_probe_max_lights; - RID gi_probe_lights_uniform; + VoxelGILight *voxel_gi_lights = nullptr; + uint32_t voxel_gi_max_lights = 32; + RID voxel_gi_lights_uniform; enum { - GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT, - GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE, - GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP, - GI_PROBE_SHADER_VERSION_WRITE_TEXTURE, - GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT, - GI_PROBE_SHADER_VERSION_MAX + VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT, + VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE, + VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP, + VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE, + VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT, + VOXEL_GI_SHADER_VERSION_MAX }; - GiprobeShaderRD giprobe_shader; - RID giprobe_lighting_shader_version; - RID giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_MAX]; - RID giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_MAX]; + VoxelGiShaderRD voxel_gi_shader; + RID voxel_gi_lighting_shader_version; + RID voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_MAX]; + RID voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_MAX]; enum { - GI_PROBE_DEBUG_COLOR, - GI_PROBE_DEBUG_LIGHT, - GI_PROBE_DEBUG_EMISSION, - GI_PROBE_DEBUG_LIGHT_FULL, - GI_PROBE_DEBUG_MAX + VOXEL_GI_DEBUG_COLOR, + VOXEL_GI_DEBUG_LIGHT, + VOXEL_GI_DEBUG_EMISSION, + VOXEL_GI_DEBUG_LIGHT_FULL, + VOXEL_GI_DEBUG_MAX }; - struct GIProbeDebugPushConstant { + struct VoxelGIDebugPushConstant { float projection[16]; uint32_t cell_offset; float dynamic_range; @@ -148,11 +149,11 @@ private: uint32_t pad; }; - GiprobeDebugShaderRD giprobe_debug_shader; - RID giprobe_debug_shader_version; - RID giprobe_debug_shader_version_shaders[GI_PROBE_DEBUG_MAX]; - PipelineCacheRD giprobe_debug_shader_version_pipelines[GI_PROBE_DEBUG_MAX]; - RID giprobe_debug_uniform_set; + VoxelGiDebugShaderRD voxel_gi_debug_shader; + RID voxel_gi_debug_shader_version; + RID voxel_gi_debug_shader_version_shaders[VOXEL_GI_DEBUG_MAX]; + PipelineCacheRD voxel_gi_debug_shader_version_pipelines[VOXEL_GI_DEBUG_MAX]; + RID voxel_gi_debug_uniform_set; /* SDFGI */ @@ -325,11 +326,11 @@ private: } sdfgi_shader; public: - /* GIPROBE INSTANCE */ + /* VOXEL_GI INSTANCE */ - //@TODO GIProbeInstance is still directly used in the render code, we'll address this when we refactor the render code itself. + //@TODO VoxelGIInstance is still directly used in the render code, we'll address this when we refactor the render code itself. - struct GIProbeInstance { + struct VoxelGIInstance { // access to our containers RendererStorageRD *storage; RendererSceneGIRD *gi; @@ -373,25 +374,25 @@ public: bool has_dynamic_object_data = false; - Transform transform; + Transform3D transform; void update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render); void debug(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); }; - mutable RID_Owner<GIProbeInstance> gi_probe_instance_owner; + mutable RID_Owner<VoxelGIInstance> voxel_gi_instance_owner; - _FORCE_INLINE_ GIProbeInstance *get_probe_instance(RID p_probe) const { - return gi_probe_instance_owner.getornull(p_probe); + _FORCE_INLINE_ VoxelGIInstance *get_probe_instance(RID p_probe) const { + return voxel_gi_instance_owner.getornull(p_probe); }; - _FORCE_INLINE_ RID gi_probe_instance_get_texture(RID p_probe) { - GIProbeInstance *gi_probe = get_probe_instance(p_probe); - ERR_FAIL_COND_V(!gi_probe, RID()); - return gi_probe->texture; + _FORCE_INLINE_ RID voxel_gi_instance_get_texture(RID p_probe) { + VoxelGIInstance *voxel_gi = get_probe_instance(p_probe); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->texture; }; - RS::GIProbeQuality gi_probe_quality = RS::GI_PROBE_QUALITY_HIGH; + RS::VoxelGIQuality voxel_gi_quality = RS::VOXEL_GI_QUALITY_HIGH; /* SDFGI */ @@ -526,10 +527,10 @@ public: int get_pending_region_data(int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const; void update_cascades(); - void debug_draw(const CameraMatrix &p_projection, const Transform &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture); + void debug_draw(const CameraMatrix &p_projection, const Transform3D &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture); void debug_probes(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); - void pre_process_gi(const Transform &p_transform, RendererSceneRenderRD *p_scene_render); + void pre_process_gi(const Transform3D &p_transform, RenderDataRD *p_render_data, RendererSceneRenderRD *p_scene_render); void render_region(RID p_render_buffers, int p_region, const PagedArray<RendererSceneRender::GeometryInstance *> &p_instances, RendererSceneRenderRD *p_scene_render); void render_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result, RendererSceneRenderRD *p_scene_render); }; @@ -550,13 +551,13 @@ public: /* GI */ enum { - MAX_GIPROBES = 8 + MAX_VOXEL_GI_INSTANCES = 8 }; // Struct for use in render buffer struct RenderBuffersGI { - RID giprobe_textures[MAX_GIPROBES]; - RID giprobe_buffer; + RID voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + RID voxel_gi_buffer; RID full_buffer; RID full_dispatch; @@ -600,7 +601,7 @@ public: ProbeCascadeData cascades[SDFGI::MAX_CASCADES]; }; - struct GIProbeData { + struct VoxelGIData { float xform[16]; float bounds[3]; float dynamic_range; @@ -610,9 +611,9 @@ public: uint32_t blend_ambient; uint32_t texture_slot; - float anisotropy_strength; - float ao; - float ao_size; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; uint32_t mipmaps; }; @@ -623,7 +624,7 @@ public: float proj_info[4]; float ao_color[3]; - uint32_t max_giprobes; + uint32_t max_voxel_gi_instances; uint32_t high_quality_vct; uint32_t orthogonal; @@ -634,16 +635,16 @@ public: RID sdfgi_ubo; enum Mode { - MODE_GIPROBE, + MODE_VOXEL_GI, MODE_SDFGI, MODE_COMBINED, - MODE_HALF_RES_GIPROBE, + MODE_HALF_RES_VOXEL_GI, MODE_HALF_RES_SDFGI, MODE_HALF_RES_COMBINED, MODE_MAX }; - RID default_giprobe_buffer; + RID default_voxel_gi_buffer; bool half_resolution = false; GiShaderRD shader; @@ -658,14 +659,14 @@ public: SDFGI *create_sdfgi(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size); - void setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used, RendererSceneRenderRD *p_scene_render); - void process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, RendererSceneRenderRD *p_scene_render); + void setup_voxel_gi_instances(RID p_render_buffers, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, uint32_t &r_voxel_gi_instances_used, RendererSceneRenderRD *p_scene_render); + void process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, RendererSceneRenderRD *p_scene_render); - RID gi_probe_instance_create(RID p_base); - void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform); - bool gi_probe_needs_update(RID p_probe) const; - void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render); - void debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); + RID voxel_gi_instance_create(RID p_base); + void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform); + bool voxel_gi_needs_update(RID p_probe) const; + void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render); + void debug_voxel_gi(RID p_voxel_gi, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); }; #endif /* !RENDERING_SERVER_SCENE_GI_RD_H */ diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 4c5bded2ff..8496ef631b 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -58,8 +58,6 @@ void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment rb->sdfgi->erase(); memdelete(rb->sdfgi); rb->sdfgi = nullptr; - - _render_buffers_uniform_set_changed(p_render_buffers); } return; } @@ -78,8 +76,6 @@ void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment if (sdfgi == nullptr) { // re-create rb->sdfgi = gi.create_sdfgi(env, p_world_position, requested_history_size); - - _render_buffers_uniform_set_changed(p_render_buffers); } else { //check for updates rb->sdfgi->update(env, p_world_position); @@ -317,7 +313,7 @@ void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS:: RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - if (low_end) { + if (!is_dynamic_gi_supported()) { return; } @@ -379,7 +375,7 @@ void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_ena RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - if (low_end) { + if (!is_volumetric_supported()) { return; } @@ -410,10 +406,6 @@ void RendererSceneRenderRD::environment_set_ssr(RID p_env, bool p_enable, int p_ RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - if (low_end) { - return; - } - env->set_ssr(p_enable, p_max_steps, p_fade_int, p_fade_out, p_depth_tolerance); } @@ -429,10 +421,6 @@ void RendererSceneRenderRD::environment_set_ssao(RID p_env, bool p_enable, float RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - if (low_end) { - return; - } - env->set_ssao(p_enable, p_radius, p_intensity, p_power, p_detail, p_horizon, p_sharpness, p_light_affect, p_ao_channel_affect); } @@ -498,7 +486,7 @@ Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_ba color.b *= env->bg_energy; Ref<Image> ret; - ret.instance(); + ret.instantiate(); ret->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF); for (int i = 0; i < p_size.width; i++) { for (int j = 0; j < p_size.height; j++) { @@ -579,10 +567,12 @@ int RendererSceneRenderRD::reflection_atlas_get_size(RID p_ref_atlas) const { RID RendererSceneRenderRD::reflection_probe_instance_create(RID p_probe) { ReflectionProbeInstance rpi; rpi.probe = p_probe; + rpi.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE); + return reflection_probe_instance_owner.make_rid(rpi); } -void RendererSceneRenderRD::reflection_probe_instance_set_transform(RID p_instance, const Transform &p_transform) { +void RendererSceneRenderRD::reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND(!rpi); @@ -639,6 +629,8 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, false); + RD::get_singleton()->draw_command_begin_label("Reflection probe render"); + if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->size != 256) { WARN_PRINT("ReflectionProbes set to UPDATE_ALWAYS must have an atlas size of 256. Please update the atlas size in the ProjectSettings."); reflection_atlas_set_size(p_reflection_atlas, 256, atlas->count); @@ -666,12 +658,12 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc //reflection atlas was unused, create: RD::TextureFormat tf; tf.array_layers = 6 * atlas->count; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; tf.mipmaps = mipmaps; tf.width = atlas->size; tf.height = atlas->size; - tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0); atlas->reflection = RD::get_singleton()->texture_create(tf, RD::TextureView()); } @@ -685,12 +677,9 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } atlas->reflections.resize(atlas->count); for (int i = 0; i < atlas->count; i++) { - atlas->reflections.write[i].data.update_reflection_data(atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers); + atlas->reflections.write[i].data.update_reflection_data(storage, atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers, _render_buffers_get_color_format()); for (int j = 0; j < 6; j++) { - Vector<RID> fb; - fb.push_back(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j]); - fb.push_back(atlas->depth_buffer); - atlas->reflections.write[i].fbs[j] = RD::get_singleton()->framebuffer_create(fb); + atlas->reflections.write[i].fbs[j] = reflection_probe_create_framebuffer(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j], atlas->depth_buffer); } } @@ -721,15 +710,28 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } } + if (rpi->atlas_index != -1) { // should we fail if this is still -1 ? + atlas->reflections.write[rpi->atlas_index].owner = p_instance; + } + rpi->atlas = p_reflection_atlas; rpi->rendering = true; rpi->dirty = false; rpi->processing_layer = 1; rpi->processing_side = 0; + RD::get_singleton()->draw_command_end_label(); + return true; } +RID RendererSceneRenderRD::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + Vector<RID> fb; + fb.push_back(p_color); + fb.push_back(p_depth); + return RD::get_singleton()->framebuffer_create(fb); +} + bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_instance) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, false); @@ -858,7 +860,7 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool shadow_atlas->shadow_owners.clear(); shadow_atlas->size = p_size; - shadow_atlas->use_16_bits = p_size; + shadow_atlas->use_16_bits = p_16_bits; } void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { @@ -933,7 +935,7 @@ bool RendererSceneRenderRD::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, //look for an empty space int sc = shadow_atlas->quadrants[qidx].shadows.size(); - ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptrw(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); int found_free_idx = -1; //found a free one int found_used_idx = -1; //found existing one, must steal it @@ -978,6 +980,78 @@ bool RendererSceneRenderRD::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, return false; } +bool RendererSceneRenderRD::_shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow) { + for (int i = p_quadrant_count - 1; i >= 0; i--) { + int qidx = p_in_quadrants[i]; + + if (shadow_atlas->quadrants[qidx].subdivision == (uint32_t)p_current_subdiv) { + return false; + } + + //look for an empty space + int sc = shadow_atlas->quadrants[qidx].shadows.size(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); + + int found_idx = -1; + uint64_t min_pass = 0; // sum of currently selected spots, try to get the least recently used pair + + for (int j = 0; j < sc - 1; j++) { + uint64_t pass = 0; + + if (sarr[j].owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(sarr[j].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == scene_pass) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (sarr[j + 1].owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(sarr[j + 1].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == scene_pass) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j + 1].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (found_idx == -1 || pass < min_pass) { + found_idx = j; + min_pass = pass; + + // we found two empty spots, no need to check the rest + if (pass == 0) { + break; + } + } + } + + if (found_idx == -1) { + continue; //nothing found + } + + r_quadrant = qidx; + r_shadow = found_idx; + + return true; + } + + return false; +} + bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND_V(!shadow_atlas, false); @@ -1023,94 +1097,104 @@ bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_i uint64_t tick = OS::get_singleton()->get_ticks_msec(); - //see if it already exists + uint32_t old_key = ShadowAtlas::SHADOW_INVALID; + uint32_t old_quadrant = ShadowAtlas::SHADOW_INVALID; + uint32_t old_shadow = ShadowAtlas::SHADOW_INVALID; + int old_subdivision = -1; + + bool should_realloc = false; + bool should_redraw = false; if (shadow_atlas->shadow_owners.has(p_light_intance)) { - //it does! - uint32_t key = shadow_atlas->shadow_owners[p_light_intance]; - uint32_t q = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; - uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK; + old_key = shadow_atlas->shadow_owners[p_light_intance]; + old_quadrant = (old_key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; + old_shadow = old_key & ShadowAtlas::SHADOW_INDEX_MASK; - bool should_realloc = shadow_atlas->quadrants[q].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[q].shadows[s].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec); - bool should_redraw = shadow_atlas->quadrants[q].shadows[s].version != p_light_version; + should_realloc = shadow_atlas->quadrants[old_quadrant].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec); + should_redraw = shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].version != p_light_version; if (!should_realloc) { - shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = p_light_version; //already existing, see if it should redraw or it's just OK return should_redraw; } - int new_quadrant, new_shadow; + old_subdivision = shadow_atlas->quadrants[old_quadrant].subdivision; + } - //find a better place - if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, shadow_atlas->quadrants[q].subdivision, tick, new_quadrant, new_shadow)) { - //found a better place! - ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; - if (sh->owner.is_valid()) { - //is taken, but is invalid, erasing it - shadow_atlas->shadow_owners.erase(sh->owner); - LightInstance *sli = light_instance_owner.getornull(sh->owner); - sli->shadow_atlases.erase(p_atlas); - } + bool is_omni = li->light_type == RS::LIGHT_OMNI; + bool found_shadow = false; + int new_quadrant = -1; + int new_shadow = -1; - //erase previous - shadow_atlas->quadrants[q].shadows.write[s].version = 0; - shadow_atlas->quadrants[q].shadows.write[s].owner = RID(); + if (is_omni) { + found_shadow = _shadow_atlas_find_omni_shadows(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } else { + found_shadow = _shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } - sh->owner = p_light_intance; - sh->alloc_tick = tick; - sh->version = p_light_version; - li->shadow_atlases.insert(p_atlas); - - //make new key - key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; - key |= new_shadow; - //update it in map - shadow_atlas->shadow_owners[p_light_intance] = key; - //make it dirty, as it should redraw anyway - return true; + if (found_shadow) { + if (old_quadrant != ShadowAtlas::SHADOW_INVALID) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].owner = RID(); + + if (old_key & ShadowAtlas::OMNI_LIGHT_FLAG) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].owner = RID(); + } } - //no better place for this shadow found, keep current + uint32_t new_key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; + new_key |= new_shadow; - //already existing, see if it should redraw or it's just OK + ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; + _shadow_atlas_invalidate_shadow(sh, p_atlas, shadow_atlas, new_quadrant, new_shadow); - shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version; + sh->owner = p_light_intance; + sh->alloc_tick = tick; + sh->version = p_light_version; - return should_redraw; - } + if (is_omni) { + new_key |= ShadowAtlas::OMNI_LIGHT_FLAG; - int new_quadrant, new_shadow; + int new_omni_shadow = new_shadow + 1; + ShadowAtlas::Quadrant::Shadow *extra_sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_omni_shadow]; + _shadow_atlas_invalidate_shadow(extra_sh, p_atlas, shadow_atlas, new_quadrant, new_omni_shadow); - //find a better place - if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, -1, tick, new_quadrant, new_shadow)) { - //found a better place! - ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; - if (sh->owner.is_valid()) { - //is taken, but is invalid, erasing it - shadow_atlas->shadow_owners.erase(sh->owner); - LightInstance *sli = light_instance_owner.getornull(sh->owner); - sli->shadow_atlases.erase(p_atlas); + extra_sh->owner = p_light_intance; + extra_sh->alloc_tick = tick; + extra_sh->version = p_light_version; } - sh->owner = p_light_intance; - sh->alloc_tick = tick; - sh->version = p_light_version; li->shadow_atlases.insert(p_atlas); - //make new key - uint32_t key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; - key |= new_shadow; //update it in map - shadow_atlas->shadow_owners[p_light_intance] = key; + shadow_atlas->shadow_owners[p_light_intance] = new_key; //make it dirty, as it should redraw anyway - return true; } - //no place to allocate this light, apologies + return should_redraw; +} - return false; +void RendererSceneRenderRD::_shadow_atlas_invalidate_shadow(RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, RendererSceneRenderRD::ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx) { + if (p_shadow->owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(p_shadow->owner); + uint32_t old_key = p_shadow_atlas->shadow_owners[p_shadow->owner]; + + if (old_key & ShadowAtlas::OMNI_LIGHT_FLAG) { + uint32_t s = old_key & ShadowAtlas::SHADOW_INDEX_MASK; + uint32_t omni_shadow_idx = p_shadow_idx + (s == (uint32_t)p_shadow_idx ? 1 : -1); + RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *omni_shadow = &p_shadow_atlas->quadrants[p_quadrant].shadows.write[omni_shadow_idx]; + omni_shadow->version = 0; + omni_shadow->owner = RID(); + } + + p_shadow->version = 0; + p_shadow->owner = RID(); + sli->shadow_atlases.erase(p_atlas); + p_shadow_atlas->shadow_owners.erase(p_shadow->owner); + } } void RendererSceneRenderRD::_update_directional_shadow_atlas() { @@ -1135,6 +1219,7 @@ void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size, bool p } directional_shadow.size = p_size; + directional_shadow.use_16_bits = p_16_bits; if (directional_shadow.depth.is_valid()) { RD::get_singleton()->free(directional_shadow.depth); @@ -1241,11 +1326,14 @@ RID RendererSceneRenderRD::light_instance_create(RID p_light) { light_instance->self = li; light_instance->light = p_light; light_instance->light_type = storage->light_get_type(p_light); + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + light_instance->forward_id = _allocate_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT); + } return li; } -void RendererSceneRenderRD::light_instance_set_transform(RID p_light_instance, const Transform &p_transform) { +void RendererSceneRenderRD::light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform) { LightInstance *light_instance = light_instance_owner.getornull(p_light_instance); ERR_FAIL_COND(!light_instance); @@ -1259,7 +1347,7 @@ void RendererSceneRenderRD::light_instance_set_aabb(RID p_light_instance, const light_instance->aabb = p_aabb; } -void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale, float p_range_begin, const Vector2 &p_uv_scale) { +void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale, float p_range_begin, const Vector2 &p_uv_scale) { LightInstance *light_instance = light_instance_owner.getornull(p_light_instance); ERR_FAIL_COND(!light_instance); @@ -1314,10 +1402,11 @@ RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap RID RendererSceneRenderRD::decal_instance_create(RID p_decal) { DecalInstance di; di.decal = p_decal; + di.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_DECAL); return decal_instance_owner.make_rid(di); } -void RendererSceneRenderRD::decal_instance_set_transform(RID p_decal, const Transform &p_transform) { +void RendererSceneRenderRD::decal_instance_set_transform(RID p_decal, const Transform3D &p_transform) { DecalInstance *di = decal_instance_owner.getornull(p_decal); ERR_FAIL_COND(!di); di->transform = p_transform; @@ -1330,7 +1419,7 @@ RID RendererSceneRenderRD::lightmap_instance_create(RID p_lightmap) { li.lightmap = p_lightmap; return lightmap_instance_owner.make_rid(li); } -void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform) { +void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) { LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap); ERR_FAIL_COND(!li); li->transform = p_transform; @@ -1338,28 +1427,28 @@ void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, cons ///////////////////////////////// -RID RendererSceneRenderRD::gi_probe_instance_create(RID p_base) { - return gi.gi_probe_instance_create(p_base); +RID RendererSceneRenderRD::voxel_gi_instance_create(RID p_base) { + return gi.voxel_gi_instance_create(p_base); } -void RendererSceneRenderRD::gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) { - gi.gi_probe_instance_set_transform_to_data(p_probe, p_xform); +void RendererSceneRenderRD::voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) { + gi.voxel_gi_instance_set_transform_to_data(p_probe, p_xform); } -bool RendererSceneRenderRD::gi_probe_needs_update(RID p_probe) const { - if (low_end) { +bool RendererSceneRenderRD::voxel_gi_needs_update(RID p_probe) const { + if (!is_dynamic_gi_supported()) { return false; } - return gi.gi_probe_needs_update(p_probe); + return gi.voxel_gi_needs_update(p_probe); } -void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) { - if (low_end) { +void RendererSceneRenderRD::voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) { + if (!is_dynamic_gi_supported()) { return; } - gi.gi_probe_update(p_probe, p_update_light_instances, p_light_instances, p_dynamic_objects, this); + gi.voxel_gi_update(p_probe, p_update_light_instances, p_light_instances, p_dynamic_objects, this); } void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform) { @@ -1385,12 +1474,20 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { uint32_t mipmaps_required = Image::get_image_required_mipmaps(rb->width, rb->height, Image::FORMAT_RGBAH); + // TODO make sure texture_create_shared_from_slice works for multiview + RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); // RD::DATA_FORMAT_R16G16B16A16_SFLOAT; tf.width = rb->width; tf.height = rb->height; - tf.texture_type = RD::TEXTURE_TYPE_2D; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; + tf.array_layers = rb->view_count; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + if (_render_buffers_can_be_storage()) { + tf.usage_bits += RD::TEXTURE_USAGE_STORAGE_BIT; + } else { + tf.usage_bits += RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } tf.mipmaps = mipmaps_required; rb->blur[0].texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); @@ -1410,17 +1507,125 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { mm.width = base_width; mm.height = base_height; + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(mm.texture); + + mm.fb = RD::get_singleton()->framebuffer_create(fb); + } + + if (!_render_buffers_can_be_storage()) { + // and half texture, this is an intermediate result so just allocate a texture, is this good enough? + tf.width = MAX(1, base_width >> 1); + tf.height = base_height; + tf.mipmaps = 1; // 1 or 0? + + mm.half_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + Vector<RID> half_fb; + half_fb.push_back(mm.half_texture); + mm.half_fb = RD::get_singleton()->framebuffer_create(half_fb); + } + rb->blur[0].mipmaps.push_back(mm); if (i > 0) { mm.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->blur[1].texture, 0, i - 1); + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(mm.texture); + + mm.fb = RD::get_singleton()->framebuffer_create(fb); + + // We can re-use the half texture here as it is an intermediate result + } + rb->blur[1].mipmaps.push_back(mm); } base_width = MAX(1, base_width >> 1); base_height = MAX(1, base_height >> 1); } + + if (!_render_buffers_can_be_storage()) { + // create 4 weight textures, 2 full size, 2 half size + + tf.format = RD::DATA_FORMAT_R16_SFLOAT; // We could probably use DATA_FORMAT_R8_SNORM if we don't pre-multiply by blur_size but that depends on whether we can remove DEPTH_GAP + tf.width = rb->width; + tf.height = rb->height; + tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; + tf.array_layers = rb->view_count; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.mipmaps = 1; + for (uint32_t i = 0; i < 4; i++) { + // associated blur texture + RID texture; + if (i == 0) { + texture = rb->texture; + } else if (i == 1) { + texture = rb->blur[0].mipmaps[0].texture; + } else if (i == 2) { + texture = rb->blur[1].mipmaps[0].texture; + } else if (i == 3) { + texture = rb->blur[0].mipmaps[1].texture; + } + + // create weight texture + rb->weight_buffers[i].weight = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + // create frame buffer + Vector<RID> fb; + fb.push_back(texture); + fb.push_back(rb->weight_buffers[i].weight); + rb->weight_buffers[i].fb = RD::get_singleton()->framebuffer_create(fb); + + if (i == 1) { + // next 2 are half size + tf.width = MAX(1, tf.width >> 1); + tf.height = MAX(1, tf.height >> 1); + } + } + + { + // and finally an FB for just our base weights + Vector<RID> fb; + fb.push_back(rb->weight_buffers[0].weight); + rb->base_weight_fb = RD::get_singleton()->framebuffer_create(fb); + } + } +} + +void RendererSceneRenderRD::_allocate_depth_backbuffer_textures(RenderBuffers *rb) { + ERR_FAIL_COND(!rb->depth_back_texture.is_null()); + + { + RD::TextureFormat tf; + if (rb->view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } + // We're not using this as a depth stencil, just copying our data into this. May need to look into using a different format on mobile, maybe R16? + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + + tf.width = rb->width; + tf.height = rb->height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.array_layers = rb->view_count; // create a layer for every view + + tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; // set this as color attachment because we're copying data into it, it's not actually used as a depth buffer + + rb->depth_back_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + + if (!_render_buffers_can_be_storage()) { + // create framebuffer so we can write into this... + + Vector<RID> fb; + fb.push_back(rb->depth_back_texture); + + rb->depth_back_fb = RD::get_singleton()->framebuffer_create(fb, RD::INVALID_ID, rb->view_count); + } } void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) { @@ -1437,26 +1642,48 @@ void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) { tf.format = RD::DATA_FORMAT_R32_SFLOAT; tf.width = w; tf.height = h; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; bool final = w == 1 && h == 1; - if (final) { - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + if (_render_buffers_can_be_storage()) { + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + if (final) { + tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + } + } else { + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; } RID texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); rb->luminance.reduce.push_back(texture); + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(texture); + + rb->luminance.fb.push_back(RD::get_singleton()->framebuffer_create(fb)); + } if (final) { rb->luminance.current = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(rb->luminance.current); + + rb->luminance.current_fb = RD::get_singleton()->framebuffer_create(fb); + } break; } } } void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { + if (rb->texture_fb.is_valid()) { + RD::get_singleton()->free(rb->texture_fb); + rb->texture_fb = RID(); + } + if (rb->texture.is_valid()) { RD::get_singleton()->free(rb->texture); rb->texture = RID(); @@ -1467,20 +1694,54 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { rb->depth_texture = RID(); } + if (rb->depth_back_fb.is_valid()) { + RD::get_singleton()->free(rb->depth_back_fb); + rb->depth_back_fb = RID(); + } + + if (rb->depth_back_texture.is_valid()) { + RD::get_singleton()->free(rb->depth_back_texture); + rb->depth_back_texture = RID(); + } + for (int i = 0; i < 2; i++) { + for (int m = 0; m < rb->blur[i].mipmaps.size(); m++) { + // do we free the texture slice here? or is it enough to free the main texture? + + // do free the mobile extra stuff + if (rb->blur[i].mipmaps[m].fb.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].fb); + } + if (rb->blur[i].mipmaps[m].half_fb.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_fb); + } + if (rb->blur[i].mipmaps[m].half_texture.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_texture); + } + } + rb->blur[i].mipmaps.clear(); + if (rb->blur[i].texture.is_valid()) { RD::get_singleton()->free(rb->blur[i].texture); rb->blur[i].texture = RID(); - rb->blur[i].mipmaps.clear(); } } + for (int i = 0; i < rb->luminance.fb.size(); i++) { + RD::get_singleton()->free(rb->luminance.fb[i]); + } + rb->luminance.fb.clear(); + for (int i = 0; i < rb->luminance.reduce.size(); i++) { RD::get_singleton()->free(rb->luminance.reduce[i]); } - rb->luminance.reduce.clear(); + if (rb->luminance.current_fb.is_valid()) { + RD::get_singleton()->free(rb->luminance.current_fb); + rb->luminance.current_fb = RID(); + } + if (rb->luminance.current.is_valid()) { RD::get_singleton()->free(rb->luminance.current); rb->luminance.current = RID(); @@ -1541,7 +1802,6 @@ void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatri if (rb->blur[0].texture.is_null()) { _allocate_blur_textures(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } storage->get_effects()->sub_surface_scattering(rb->texture, rb->blur[0].mipmaps[0].texture, rb->depth_texture, p_camera, Size2i(rb->width, rb->height), sss_scale, sss_depth_scale, sss_quality); @@ -1593,7 +1853,6 @@ void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_frameb if (rb->blur[0].texture.is_null()) { _allocate_blur_textures(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } storage->get_effects()->screen_space_reflection(rb->texture, p_normal_buffer, ssr_roughness_quality, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->width / 2, rb->height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection); @@ -1719,7 +1978,6 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); RD::get_singleton()->set_resource_name(rb->ssao.ao_final, "SSAO Final"); - _render_buffers_uniform_set_changed(p_render_buffers); } ssao_using_half_size = ssao_half_size; uniform_sets_are_invalid = true; @@ -1743,54 +2001,146 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen settings.half_screen_size = Size2i(buffer_width, buffer_height); settings.quarter_screen_size = Size2i(half_width, half_height); - storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid); + storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid, rb->ssao.downsample_uniform_set, rb->ssao.gather_uniform_set, rb->ssao.importance_map_uniform_set); } -void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_render_buffers, RID p_environment, RID p_camera_effects, const CameraMatrix &p_projection) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); +void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); ERR_FAIL_COND(!rb); - RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); + RD::get_singleton()->draw_command_begin_label("Copy screen texture"); + + if (rb->blur[0].texture.is_null()) { + _allocate_blur_textures(rb); + } + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye + + bool can_use_storage = _render_buffers_can_be_storage(); + + if (can_use_storage) { + storage->get_effects()->copy_to_rect(rb->texture, rb->blur[0].mipmaps[0].texture, Rect2i(0, 0, rb->width, rb->height)); + for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) { + storage->get_effects()->make_mipmap(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].texture, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height)); + } + } else { + storage->get_effects()->copy_to_fb_rect(rb->texture, rb->blur[0].mipmaps[0].fb, Rect2i(0, 0, rb->width, rb->height)); + for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) { + storage->get_effects()->make_mipmap_raster(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].fb, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height)); + } + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RD::get_singleton()->draw_command_begin_label("Copy depth texture"); + + if (rb->depth_back_texture.is_null()) { + _allocate_depth_backbuffer_textures(rb); + } + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye + + bool can_use_storage = _render_buffers_can_be_storage(); + + if (can_use_storage) { + storage->get_effects()->copy_to_rect(rb->depth_texture, rb->depth_back_texture, Rect2i(0, 0, rb->width, rb->height)); + } else { + storage->get_effects()->copy_to_fb_rect(rb->depth_texture, rb->depth_back_fb, Rect2i(0, 0, rb->width, rb->height)); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment); //glow (if enabled) - CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects); + CameraEffects *camfx = camera_effects_owner.getornull(p_render_data->camera_effects); bool can_use_effects = rb->width >= 8 && rb->height >= 8; + bool can_use_storage = _render_buffers_can_be_storage(); + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye if (can_use_effects && camfx && (camfx->dof_blur_near_enabled || camfx->dof_blur_far_enabled) && camfx->dof_blur_amount > 0.0) { + RD::get_singleton()->draw_command_begin_label("DOF"); if (rb->blur[0].texture.is_null()) { _allocate_blur_textures(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } + EffectsRD::BokehBuffers buffers; + + // textures we use + buffers.base_texture_size = Size2i(rb->width, rb->height); + buffers.base_texture = rb->texture; + buffers.depth_texture = rb->depth_texture; + buffers.secondary_texture = rb->blur[0].mipmaps[0].texture; + buffers.half_texture[0] = rb->blur[1].mipmaps[0].texture; + buffers.half_texture[1] = rb->blur[0].mipmaps[1].texture; + float bokeh_size = camfx->dof_blur_amount * 64.0; - storage->get_effects()->bokeh_dof(rb->texture, rb->depth_texture, Size2i(rb->width, rb->height), rb->blur[0].mipmaps[0].texture, rb->blur[1].mipmaps[0].texture, rb->blur[0].mipmaps[1].texture, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, dof_blur_use_jitter, p_projection.get_z_near(), p_projection.get_z_far(), p_projection.is_orthogonal()); + if (can_use_storage) { + storage->get_effects()->bokeh_dof(buffers, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, dof_blur_use_jitter, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + } else { + // set framebuffers + buffers.base_fb = rb->texture_fb; + buffers.secondary_fb = rb->weight_buffers[1].fb; + buffers.half_fb[0] = rb->weight_buffers[2].fb; + buffers.half_fb[1] = rb->weight_buffers[3].fb; + buffers.weight_texture[0] = rb->weight_buffers[0].weight; + buffers.weight_texture[1] = rb->weight_buffers[1].weight; + buffers.weight_texture[2] = rb->weight_buffers[2].weight; + buffers.weight_texture[3] = rb->weight_buffers[3].weight; + + // set weight buffers + buffers.base_weight_fb = rb->base_weight_fb; + + storage->get_effects()->bokeh_dof_raster(buffers, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + } + RD::get_singleton()->draw_command_end_label(); } if (can_use_effects && env && env->auto_exposure) { + RD::get_singleton()->draw_command_begin_label("Auto exposure"); if (rb->luminance.current.is_null()) { _allocate_luminance_textures(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } bool set_immediate = env->auto_exposure_version != rb->auto_exposure_version; rb->auto_exposure_version = env->auto_exposure_version; double step = env->auto_exp_speed * time_step; - storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); - + if (can_use_storage) { + storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + } else { + storage->get_effects()->luminance_reduction_raster(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + } //swap final reduce with prev luminance SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]); + if (!can_use_storage) { + SWAP(rb->luminance.current_fb, rb->luminance.fb.write[rb->luminance.fb.size() - 1]); + } + RenderingServerDefault::redraw_request(); //redraw all the time if auto exposure rendering is on + RD::get_singleton()->draw_command_end_label(); } int max_glow_level = -1; if (can_use_effects && env && env->glow_enabled) { + RD::get_singleton()->draw_command_begin_label("Gaussian Glow"); + /* see that blur textures are allocated */ if (rb->blur[1].texture.is_null()) { _allocate_blur_textures(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } for (int i = 0; i < RS::MAX_GLOW_LEVELS; i++) { @@ -1812,14 +2162,26 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_rende if (env->auto_exposure && rb->luminance.current.is_valid()) { luminance_texture = rb->luminance.current; } - storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + if (can_use_storage) { + storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + } else { + storage->get_effects()->gaussian_glow_raster(rb->texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + } } else { - storage->get_effects()->gaussian_glow(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality); + if (can_use_storage) { + storage->get_effects()->gaussian_glow(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality); + } else { + storage->get_effects()->gaussian_glow_raster(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Vector2(1.0 / vp_w, 1.0 / vp_h), env->glow_strength, glow_high_quality); + } } } + + RD::get_singleton()->draw_command_end_label(); } { + RD::get_singleton()->draw_command_begin_label("Tonemap"); + //tonemap EffectsRD::TonemapSettings tonemap; @@ -1875,13 +2237,85 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_rende } } + tonemap.view_count = p_render_data->view_count; + storage->get_effects()->tonemapper(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), tonemap); + + RD::get_singleton()->draw_command_end_label(); } storage->render_target_disable_clear_request(rb->render_target); } -void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas) { +void RendererSceneRenderRD::_post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data) { + RD::get_singleton()->draw_command_begin_label("Post Process Subpass"); + + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment); + + bool can_use_effects = rb->width >= 8 && rb->height >= 8; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + EffectsRD::TonemapSettings tonemap; + + if (env) { + tonemap.tonemap_mode = env->tone_mapper; + tonemap.exposure = env->exposure; + tonemap.white = env->white; + } + + // We don't support glow or auto exposure here, if they are needed, don't use subpasses! + // The problem is that we need to use the result so far and process them before we can + // apply this to our results. + if (can_use_effects && env && env->glow_enabled) { + ERR_FAIL_MSG("Glow is not supported when using subpasses."); + } + if (can_use_effects && env && env->auto_exposure) { + ERR_FAIL_MSG("Glow is not supported when using subpasses."); + } + + tonemap.use_glow = false; + tonemap.glow_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + tonemap.use_auto_exposure = false; + tonemap.exposure_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + + tonemap.use_color_correction = false; + tonemap.use_1d_color_correction = false; + tonemap.color_correction_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); + + if (can_use_effects && env) { + tonemap.use_bcs = env->adjustments_enabled; + tonemap.brightness = env->adjustments_brightness; + tonemap.contrast = env->adjustments_contrast; + tonemap.saturation = env->adjustments_saturation; + if (env->adjustments_enabled && env->color_correction.is_valid()) { + tonemap.use_color_correction = true; + tonemap.use_1d_color_correction = env->use_1d_color_correction; + tonemap.color_correction_texture = storage->texture_get_rd_texture(env->color_correction); + } + } + + tonemap.use_debanding = rb->use_debanding; + tonemap.texture_size = Vector2i(rb->width, rb->height); + + tonemap.view_count = p_render_data->view_count; + + storage->get_effects()->tonemapper(draw_list, p_source_texture, RD::get_singleton()->framebuffer_get_format(p_framebuffer), tonemap); + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_disable_clear_request(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + storage->render_target_disable_clear_request(rb->render_target); +} + +void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer) { EffectsRD *effects = storage->get_effects(); RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); @@ -1940,6 +2374,13 @@ void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID RID reflection_texture = rb->reflection_buffer; effects->copy_to_fb_rect(ambient_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture); } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_OCCLUDERS) { + if (p_occlusion_buffer.is_valid()) { + Size2 rtsize = storage->render_target_get_size(rb->render_target); + effects->copy_to_fb_rect(storage->texture_get_rd_texture(p_occlusion_buffer), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize), true, false); + } + } } void RendererSceneRenderRD::environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) { @@ -1963,6 +2404,15 @@ RID RendererSceneRenderRD::render_buffers_get_back_buffer_texture(RID p_render_b return rb->blur[0].texture; } +RID RendererSceneRenderRD::render_buffers_get_back_depth_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + if (!rb->depth_back_texture.is_valid()) { + return RID(); //not valid at the moment + } + return rb->depth_back_texture; +} + RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, RID()); @@ -1970,17 +2420,17 @@ RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) { return rb->ssao.ao_final; } -RID RendererSceneRenderRD::render_buffers_get_gi_probe_buffer(RID p_render_buffers) { +RID RendererSceneRenderRD::render_buffers_get_voxel_gi_buffer(RID p_render_buffers) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, RID()); - if (rb->gi.giprobe_buffer.is_null()) { - rb->gi.giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(RendererSceneGIRD::GIProbeData) * RendererSceneGIRD::MAX_GIPROBES); + if (rb->gi.voxel_gi_buffer.is_null()) { + rb->gi.voxel_gi_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(RendererSceneGIRD::VoxelGIData) * RendererSceneGIRD::MAX_VOXEL_GI_INSTANCES); } - return rb->gi.giprobe_buffer; + return rb->gi.voxel_gi_buffer; } -RID RendererSceneRenderRD::render_buffers_get_default_gi_probe_buffer() { - return gi.default_giprobe_buffer; +RID RendererSceneRenderRD::render_buffers_get_default_voxel_gi_buffer() { + return gi.default_voxel_gi_buffer; } RID RendererSceneRenderRD::render_buffers_get_gi_ambient_texture(RID p_render_buffers) { @@ -2123,7 +2573,17 @@ float RendererSceneRenderRD::render_buffers_get_volumetric_fog_detail_spread(RID return rb->volumetric_fog->spread; } -void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding) { +RD::DataFormat RendererSceneRenderRD::_render_buffers_get_color_format() { + return RD::DATA_FORMAT_R16G16B16A16_SFLOAT; +} + +bool RendererSceneRenderRD::_render_buffers_can_be_storage() { + return true; +} + +void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) { + ERR_FAIL_COND_MSG(p_view_count == 0, "Must have at least 1 view"); + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); rb->width = p_width; rb->height = p_height; @@ -2131,6 +2591,7 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->msaa = p_msaa; rb->screen_space_aa = p_screen_space_aa; rb->use_debanding = p_use_debanding; + rb->view_count = p_view_count; if (is_clustered_enabled()) { if (rb->cluster_builder == nullptr) { @@ -2143,21 +2604,27 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p { RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + if (rb->view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } + tf.format = _render_buffers_get_color_format(); tf.width = rb->width; tf.height = rb->height; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.array_layers = rb->view_count; // create a layer for every view + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0) | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) { - tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - } else { - tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; } + tf.usage_bits |= RD::TEXTURE_USAGE_INPUT_ATTACHMENT_BIT; // only needed when using subpasses in the mobile renderer rb->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); } { RD::TextureFormat tf; + if (rb->view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } if (rb->msaa == RS::VIEWPORT_MSAA_DISABLED) { tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT; } else { @@ -2167,6 +2634,7 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p tf.width = p_width; tf.height = p_height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.array_layers = rb->view_count; // create a layer for every view if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) { tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; @@ -2177,8 +2645,16 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->depth_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); } - rb->data->configure(rb->texture, rb->depth_texture, p_width, p_height, p_msaa); - _render_buffers_uniform_set_changed(p_render_buffers); + if (!_render_buffers_can_be_storage()) { + // ONLY USED ON MOBILE RENDERER, ONLY USED FOR POST EFFECTS! + Vector<RID> fb; + fb.push_back(rb->texture); + + rb->texture_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, rb->view_count); + } + + RID target_texture = storage->render_target_get_rd_texture(rb->render_target); + rb->data->configure(rb->texture, rb->depth_texture, target_texture, p_width, p_height, p_msaa, p_view_count); if (is_clustered_enabled()) { rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); @@ -2240,6 +2716,8 @@ void RendererSceneRenderRD::shadows_quality_set(RS::ShadowQuality p_quality) { get_vogel_disk(penumbra_shadow_kernel, penumbra_shadow_samples); get_vogel_disk(soft_shadow_kernel, soft_shadow_samples); } + + _update_shader_quality_settings(); } void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_quality) { @@ -2280,6 +2758,23 @@ void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_q get_vogel_disk(directional_penumbra_shadow_kernel, directional_penumbra_shadow_samples); get_vogel_disk(directional_soft_shadow_kernel, directional_soft_shadow_samples); } + + _update_shader_quality_settings(); +} + +void RendererSceneRenderRD::decals_set_filter(RenderingServer::DecalFilter p_filter) { + if (decals_filter == p_filter) { + return; + } + decals_filter = p_filter; + _update_shader_quality_settings(); +} +void RendererSceneRenderRD::light_projectors_set_filter(RenderingServer::LightProjectorFilter p_filter) { + if (light_projectors_filter == p_filter) { + return; + } + light_projectors_filter = p_filter; + _update_shader_quality_settings(); } int RendererSceneRenderRD::get_roughness_layers() const { @@ -2296,7 +2791,7 @@ RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_g return rb->data; } -void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment) { +void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform3D &p_camera_inverse_transform, RID p_environment) { cluster.reflection_count = 0; for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) { @@ -2319,10 +2814,13 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti sort_array.sort(cluster.reflection_sort, cluster.reflection_count); } + bool using_forward_ids = _uses_forward_ids(); for (uint32_t i = 0; i < cluster.reflection_count; i++) { ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance; - rpi->render_index = i; + if (using_forward_ids) { + _map_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id, i); + } RID base_probe = rpi->probe; @@ -2330,6 +2828,8 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti Vector3 extents = storage->reflection_probe_get_extents(base_probe); + rpi->cull_mask = storage->reflection_probe_get_cull_mask(base_probe); + reflection_ubo.box_extents[0] = extents.x; reflection_ubo.box_extents[1] = extents.y; reflection_ubo.box_extents[2] = extents.z; @@ -2354,22 +2854,24 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti reflection_ubo.ambient[1] = ambient_linear.g * interior_ambient_energy; reflection_ubo.ambient[2] = ambient_linear.b * interior_ambient_energy; - Transform transform = rpi->transform; - Transform proj = (p_camera_inverse_transform * transform).inverse(); + Transform3D transform = rpi->transform; + Transform3D proj = (p_camera_inverse_transform * transform).inverse(); RendererStorageRD::store_transform(proj, reflection_ubo.local_matrix); - current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, transform, extents); + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, transform, extents); + } rpi->last_pass = RSG::rasterizer->get_frame_number(); } if (cluster.reflection_count) { - RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(RendererSceneSkyRD::ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(Cluster::ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } -void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { - Transform inverse_transform = p_camera_transform.affine_inverse(); +void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows) { + Transform3D inverse_transform = p_camera_transform.affine_inverse(); r_directional_light_count = 0; r_positional_light_count = 0; @@ -2380,6 +2882,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const cluster.omni_light_count = 0; cluster.spot_light_count = 0; + r_directional_light_soft_shadows = false; + for (int i = 0; i < (int)p_lights.size(); i++) { LightInstance *li = light_instance_owner.getornull(p_lights[i]); if (!li) { @@ -2395,7 +2899,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const // Copy to SkyDirectionalLightData if (r_directional_light_count < sky.sky_scene_state.max_directional_lights) { RendererSceneSkyRD::SkyDirectionalLightData &sky_light_data = sky.sky_scene_state.directional_lights[r_directional_light_count]; - Transform light_transform = li->transform; + Transform3D light_transform = li->transform; Vector3 world_direction = light_transform.basis.xform(Vector3(0, 0, 1)).normalized(); sky_light_data.direction[0] = world_direction.x; @@ -2418,6 +2922,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const // technically this will keep expanding until reaching the sun, but all we care // is expand until we reach the radius of the near plane (there can't be more occluders than that) angular_diameter = Math::tan(Math::deg2rad(angular_diameter)); + if (storage->light_has_shadow(base)) { + r_directional_light_soft_shadows = true; + } } else { angular_diameter = 0.0; } @@ -2431,7 +2938,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Cluster::DirectionalLightData &light_data = cluster.directional_lights[r_directional_light_count]; - Transform light_transform = li->transform; + Transform3D light_transform = li->transform; Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized(); @@ -2521,12 +3028,12 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const CameraMatrix rectm; rectm.set_light_atlas_rect(atlas_rect); - Transform modelview = (inverse_transform * li->shadow_transform[j].transform).inverse(); + Transform3D modelview = (inverse_transform * li->shadow_transform[j].transform).inverse(); CameraMatrix shadow_mtx = rectm * bias * matrix * modelview; light_data.shadow_split_offsets[j] = split; float bias_scale = li->shadow_transform[j].bias_scale; - light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_scale; + light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0 * bias_scale; light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * li->shadow_transform[j].shadow_texel_size; light_data.shadow_transmittance_bias[j] = storage->light_get_transmittance_bias(base) * bias_scale; light_data.shadow_z_range[j] = li->shadow_transform[j].farplane; @@ -2562,6 +3069,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR); light_data.softshadow_angle = angular_diameter; + light_data.bake_mode = storage->light_get_bake_mode(base); if (angular_diameter <= 0.0) { light_data.soft_shadow_scale *= directional_shadow_quality_radius_get(); // Only use quality radius for PCF @@ -2609,6 +3117,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); } + bool using_forward_ids = _uses_forward_ids(); + for (uint32_t i = 0; i < (cluster.omni_light_count + cluster.spot_light_count); i++) { uint32_t index = (i < cluster.omni_light_count) ? i : i - (cluster.omni_light_count); Cluster::LightData &light_data = (i < cluster.omni_light_count) ? cluster.omni_lights[index] : cluster.spot_lights[index]; @@ -2616,7 +3126,11 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance; RID base = li->light; - Transform light_transform = li->transform; + if (using_forward_ids) { + _map_forward_id(type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, li->forward_id, index); + } + + Transform3D light_transform = li->transform; float sign = storage->light_is_negative(base) ? -1 : 1; Color linear_col = storage->light_get_color(base).to_linear(); @@ -2629,6 +3143,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.color[1] = linear_col.g * energy; light_data.color[2] = linear_col.b * energy; light_data.specular_amount = storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 2.0; + light_data.bake_mode = storage->light_get_bake_mode(base); float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE)); light_data.inv_radius = 1.0 / radius; @@ -2688,22 +3203,19 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_enabled = true; - if (type == RS::LIGHT_SPOT) { - light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); - float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; - shadow_texel_size *= light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; + float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 10.0; + if (type == RS::LIGHT_SPOT) { + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0; } else { //omni - light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; - float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS); } light_data.transmittance_bias = storage->light_get_transmittance_bias(base); - Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas); + Vector2i omni_offset; + Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas, omni_offset); light_data.atlas_rect[0] = rect.position.x; light_data.atlas_rect[1] = rect.position.y; @@ -2714,8 +3226,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); if (type == RS::LIGHT_OMNI) { - light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another - Transform proj = (inverse_transform * light_transform).inverse(); + Transform3D proj = (inverse_transform * light_transform).inverse(); RendererStorageRD::store_transform(proj, light_data.shadow_matrix); @@ -2726,8 +3237,10 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF } + light_data.direction[0] = omni_offset.x * float(rect.size.width); + light_data.direction[1] = omni_offset.y * float(rect.size.height); } else if (type == RS::LIGHT_SPOT) { - Transform modelview = (inverse_transform * light_transform).inverse(); + Transform3D modelview = (inverse_transform * light_transform).inverse(); CameraMatrix bias; bias.set_light_bias(); @@ -2747,9 +3260,11 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_enabled = false; } - li->light_index = index; + li->cull_mask = storage->light_get_cull_mask(base); - current_cluster_builder->add_light(type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_light(type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); + } r_positional_light_count++; } @@ -2768,8 +3283,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const } } -void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform) { - Transform uv_xform; +void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform) { + Transform3D uv_xform; uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0)); uv_xform.origin = Vector3(-1.0, 0.0, -1.0); @@ -2788,7 +3303,7 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const } RID decal = di->decal; - Transform xform = di->transform; + Transform3D xform = di->transform; real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; @@ -2813,11 +3328,18 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const sort_array.sort(cluster.decal_sort, cluster.decal_count); } + bool using_forward_ids = _uses_forward_ids(); for (uint32_t i = 0; i < cluster.decal_count; i++) { DecalInstance *di = cluster.decal_sort[i].instance; RID decal = di->decal; - Transform xform = di->transform; + if (using_forward_ids) { + _map_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id, i); + } + + di->cull_mask = storage->decal_get_cull_mask(decal); + + Transform3D xform = di->transform; float fade = 1.0; if (storage->decal_is_distance_fade_enabled(decal)) { @@ -2834,9 +3356,9 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Vector3 decal_extents = storage->decal_get_extents(decal); - Transform scale_xform; + Transform3D scale_xform; scale_xform.basis.scale(Vector3(decal_extents.x, decal_extents.y, decal_extents.z)); - Transform to_decal_xform = (p_camera_inverse_xform * di->transform * scale_xform * uv_xform).affine_inverse(); + Transform3D to_decal_xform = (p_camera_inverse_xform * di->transform * scale_xform * uv_xform).affine_inverse(); RendererStorageRD::store_transform(to_decal_xform, dd.xform); Vector3 normal = xform.basis.get_axis(Vector3::AXIS_Y).normalized(); @@ -2921,7 +3443,9 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const dd.upper_fade = storage->decal_get_upper_fade(decal); dd.lower_fade = storage->decal_get_lower_fade(decal); - current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, xform, decal_extents); + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, xform, decal_extents); + } } if (cluster.decal_count > 0) { @@ -2954,7 +3478,7 @@ void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { rb->volumetric_fog = nullptr; } -void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count) { +void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform3D &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count) { ERR_FAIL_COND(!is_clustered_enabled()); // can't use volumetric fog without clustered RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); @@ -2968,7 +3492,6 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e //validate if (!env || !env->volumetric_fog_enabled || rb->volumetric_fog->width != target_width || rb->volumetric_fog->height != target_height || rb->volumetric_fog->depth != volumetric_fog_depth) { _volumetric_fog_erase(rb); - _render_buffers_uniform_set_changed(p_render_buffers); } } @@ -3004,7 +3527,6 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); - _render_buffers_uniform_set_changed(p_render_buffers); Vector<RD::Uniform> uniforms; { @@ -3118,7 +3640,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 11; - u.ids.push_back(render_buffers_get_gi_probe_buffer(p_render_buffers)); + u.ids.push_back(render_buffers_get_voxel_gi_buffer(p_render_buffers)); uniforms.push_back(u); } @@ -3126,8 +3648,8 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 12; - for (int i = 0; i < RendererSceneGIRD::MAX_GIPROBES; i++) { - u.ids.push_back(rb->gi.giprobe_textures[i]); + for (int i = 0; i < RendererSceneGIRD::MAX_VOXEL_GI_INSTANCES; i++) { + u.ids.push_back(rb->gi.voxel_gi_textures[i]); } uniforms.push_back(u); } @@ -3252,10 +3774,10 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e params.cam_rotation[10] = p_cam_transform.basis[2][2]; params.cam_rotation[11] = 0; params.filter_axis = 0; - params.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + params.max_voxel_gi_instances = env->volumetric_fog_gi_inject > 0.001 ? p_voxel_gi_count : 0; params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; - Transform to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform; + Transform3D to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform; storage->store_transform(to_prev_cam_view, params.to_prev_view); params.use_temporal_reprojection = env->volumetric_fog_temporal_reprojection; @@ -3349,13 +3871,9 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->prev_cam_transform = p_cam_transform; } -uint32_t RendererSceneRenderRD::_get_render_state_directional_light_count() const { - return render_state.directional_light_count; -} - -bool RendererSceneRenderRD::_needs_post_prepass_render(bool p_use_gi) { - if (render_state.render_buffers.is_valid()) { - RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); +bool RendererSceneRenderRD::_needs_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); if (rb->sdfgi != nullptr) { return true; } @@ -3363,49 +3881,46 @@ bool RendererSceneRenderRD::_needs_post_prepass_render(bool p_use_gi) { return false; } -void RendererSceneRenderRD::_post_prepass_render(bool p_use_gi) { - if (render_state.render_buffers.is_valid()) { +void RendererSceneRenderRD::_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid()) { if (p_use_gi) { - RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); ERR_FAIL_COND(rb == nullptr); if (rb->sdfgi == nullptr) { return; } - RendererSceneEnvironmentRD *env = environment_owner.getornull(render_state.environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment); rb->sdfgi->update_probes(env, sky.sky_owner.getornull(env->sky)); } } } -void RendererSceneRenderRD::_pre_resolve_render(bool p_use_gi) { - if (render_state.render_buffers.is_valid()) { +void RendererSceneRenderRD::_pre_resolve_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid()) { if (p_use_gi) { RD::get_singleton()->compute_list_end(); } } } -void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer) { +void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer) { // Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time - if (render_state.render_buffers.is_valid() && p_use_gi) { - RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + if (p_render_data->render_buffers.is_valid() && p_use_gi) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); ERR_FAIL_COND(rb == nullptr); - if (rb->sdfgi == nullptr) { - return; + if (rb->sdfgi != nullptr) { + rb->sdfgi->store_probes(); } - - rb->sdfgi->store_probes(); } render_state.cube_shadows.clear(); render_state.shadows.clear(); render_state.directional_shadows.clear(); - Plane camera_plane(render_state.cam_transform.origin, -render_state.cam_transform.basis.get_axis(Vector3::AXIS_Z)); - float lod_distance_multiplier = render_state.cam_projection.get_lod_multiplier(); - + Plane camera_plane(p_render_data->cam_transform.origin, -p_render_data->cam_transform.basis.get_axis(Vector3::AXIS_Z)); + float lod_distance_multiplier = p_render_data->cam_projection.get_lod_multiplier(); { for (int i = 0; i < render_state.render_shadow_count; i++) { LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light); @@ -3421,7 +3936,7 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R //cube shadows are rendered in their own way for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, true, true, true); + _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, true, true, true, p_render_data->render_info); } if (render_state.directional_shadows.size()) { @@ -3435,7 +3950,7 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R // Render GI bool render_shadows = render_state.directional_shadows.size() || render_state.shadows.size(); - bool render_gi = render_state.render_buffers.is_valid() && p_use_gi; + bool render_gi = p_render_data->render_buffers.is_valid() && p_use_gi; if (render_shadows && render_gi) { RENDER_TIMESTAMP("Render GI + Render Shadows (parallel)"); @@ -3451,11 +3966,11 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R //render directional shadows for (uint32_t i = 0; i < render_state.directional_shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false); + _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false, p_render_data->render_info); } //render positional shadows for (uint32_t i = 0; i < render_state.shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true); + _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true, p_render_data->render_info); } _render_shadow_process(); @@ -3463,7 +3978,7 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R //start GI if (render_gi) { - gi.process_gi(render_state.render_buffers, p_normal_roughness_buffer, p_gi_probe_buffer, render_state.environment, render_state.cam_projection, render_state.cam_transform, *render_state.gi_probes, this); + gi.process_gi(p_render_data->render_buffers, p_normal_roughness_buffer, p_voxel_gi_buffer, p_render_data->environment, p_render_data->cam_projection, p_render_data->cam_transform, *p_render_data->voxel_gi_instances, this); } //Do shadow rendering (in parallel with GI) @@ -3475,9 +3990,9 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //use a later barrier } - if (render_state.render_buffers.is_valid()) { + if (p_render_data->render_buffers.is_valid()) { if (p_use_ssao) { - _process_ssao(render_state.render_buffers, render_state.environment, p_normal_roughness_buffer, render_state.cam_projection); + _process_ssao(p_render_data->render_buffers, p_render_data->environment, p_normal_roughness_buffer, p_render_data->cam_projection); } } @@ -3485,32 +4000,32 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL); if (current_cluster_builder) { - current_cluster_builder->begin(render_state.cam_transform, render_state.cam_projection, !render_state.reflection_probe.is_valid()); + current_cluster_builder->begin(p_render_data->cam_transform, p_render_data->cam_projection, !p_render_data->reflection_probe.is_valid()); } bool using_shadows = true; - if (render_state.reflection_probe.is_valid()) { - if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(render_state.reflection_probe))) { + if (p_render_data->reflection_probe.is_valid()) { + if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { using_shadows = false; } } else { //do not render reflections when rendering a reflection probe - _setup_reflections(*render_state.reflection_probes, render_state.cam_transform.affine_inverse(), render_state.environment); + _setup_reflections(*p_render_data->reflection_probes, p_render_data->cam_transform.affine_inverse(), p_render_data->environment); } uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*render_state.lights, render_state.cam_transform, render_state.shadow_atlas, using_shadows, directional_light_count, positional_light_count); - _setup_decals(*render_state.decals, render_state.cam_transform.affine_inverse()); + _setup_lights(*p_render_data->lights, p_render_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); + _setup_decals(*p_render_data->decals, p_render_data->cam_transform.affine_inverse()); - render_state.directional_light_count = directional_light_count; + p_render_data->directional_light_count = directional_light_count; if (current_cluster_builder) { current_cluster_builder->bake_cluster(); } - if (render_state.render_buffers.is_valid()) { + if (p_render_data->render_buffers.is_valid()) { bool directional_shadows = false; for (uint32_t i = 0; i < directional_light_count; i++) { if (cluster.directional_lights[i].shadow_enabled) { @@ -3519,52 +4034,75 @@ void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, R } } if (is_volumetric_supported()) { - _update_volumetric_fog(render_state.render_buffers, render_state.environment, render_state.cam_projection, render_state.cam_transform, render_state.shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.gi_probe_count); + _update_volumetric_fog(p_render_data->render_buffers, p_render_data->environment, p_render_data->cam_projection, p_render_data->cam_transform, p_render_data->shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.voxel_gi_count); } } } -void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data) { +void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data, RendererScene::RenderInfo *r_render_info) { // getting this here now so we can direct call a bunch of things more easily RenderBuffers *rb = nullptr; if (p_render_buffers.is_valid()) { rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); // !BAS! Do we fail here or skip the parts that won't work. can't really see a case why we would be rendering without buffers.... + ERR_FAIL_COND(!rb); } //assign render data + RenderDataRD render_data; { - render_state.render_buffers = p_render_buffers; - render_state.cam_transform = p_cam_transform; - render_state.cam_projection = p_cam_projection; - render_state.cam_ortogonal = p_cam_projection.is_orthogonal(); - render_state.instances = &p_instances; - render_state.lights = &p_lights; - render_state.reflection_probes = &p_reflection_probes; - render_state.gi_probes = &p_gi_probes; - render_state.decals = &p_decals; - render_state.lightmaps = &p_lightmaps; - render_state.environment = p_environment; - render_state.camera_effects = p_camera_effects; - render_state.shadow_atlas = p_shadow_atlas; - render_state.reflection_atlas = p_reflection_atlas; - render_state.reflection_probe = p_reflection_probe; - render_state.reflection_probe_pass = p_reflection_probe_pass; - render_state.screen_lod_threshold = p_screen_lod_threshold; + render_data.render_buffers = p_render_buffers; + + // Our first camera is used by default + render_data.cam_transform = p_camera_data->main_transform; + render_data.cam_projection = p_camera_data->main_projection; + render_data.view_projection[0] = p_camera_data->main_projection; + render_data.cam_ortogonal = p_camera_data->is_ortogonal; + + render_data.view_count = p_camera_data->view_count; + for (uint32_t v = 0; v < p_camera_data->view_count; v++) { + render_data.view_projection[v] = p_camera_data->view_projection[v]; + } + + render_data.z_near = p_camera_data->main_projection.get_z_near(); + render_data.z_far = p_camera_data->main_projection.get_z_far(); + + render_data.instances = &p_instances; + render_data.lights = &p_lights; + render_data.reflection_probes = &p_reflection_probes; + render_data.voxel_gi_instances = &p_voxel_gi_instances; + render_data.decals = &p_decals; + render_data.lightmaps = &p_lightmaps; + render_data.environment = p_environment; + render_data.camera_effects = p_camera_effects; + render_data.shadow_atlas = p_shadow_atlas; + render_data.reflection_atlas = p_reflection_atlas; + render_data.reflection_probe = p_reflection_probe; + render_data.reflection_probe_pass = p_reflection_probe_pass; + + // this should be the same for all cameras.. + render_data.lod_distance_multiplier = p_camera_data->main_projection.get_lod_multiplier(); + render_data.lod_camera_plane = Plane(p_camera_data->main_transform.get_origin(), -p_camera_data->main_transform.basis.get_axis(Vector3::AXIS_Z)); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { + render_data.screen_lod_threshold = 0.0; + } else { + render_data.screen_lod_threshold = p_screen_lod_threshold; + } render_state.render_shadows = p_render_shadows; render_state.render_shadow_count = p_render_shadow_count; render_state.render_sdfgi_regions = p_render_sdfgi_regions; render_state.render_sdfgi_region_count = p_render_sdfgi_region_count; render_state.sdfgi_update_data = p_sdfgi_update_data; + render_data.render_info = r_render_info; } PagedArray<RID> empty; if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { - render_state.lights = ∅ - render_state.reflection_probes = ∅ - render_state.gi_probes = ∅ + render_data.lights = ∅ + render_data.reflection_probes = ∅ + render_data.voxel_gi_instances = ∅ } //sdfgi first @@ -3584,19 +4122,21 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & clear_color = storage->get_default_clear_color(); } - //assign render indices to giprobes - for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { - RendererSceneGIRD::GIProbeInstance *giprobe_inst = gi.gi_probe_instance_owner.getornull(p_gi_probes[i]); - if (giprobe_inst) { - giprobe_inst->render_index = i; + //assign render indices to voxel_gi_instances + if (is_dynamic_gi_supported()) { + for (uint32_t i = 0; i < (uint32_t)p_voxel_gi_instances.size(); i++) { + RendererSceneGIRD::VoxelGIInstance *voxel_gi_inst = gi.voxel_gi_instance_owner.getornull(p_voxel_gi_instances[i]); + if (voxel_gi_inst) { + voxel_gi_inst->render_index = i; + } } } - if (render_buffers_owner.owns(render_state.render_buffers)) { - // render_state.render_buffers == p_render_buffers so we can use our already retrieved rb + if (render_buffers_owner.owns(render_data.render_buffers)) { + // render_data.render_buffers == p_render_buffers so we can use our already retrieved rb current_cluster_builder = rb->cluster_builder; - } else if (reflection_probe_instance_owner.owns(render_state.reflection_probe)) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_state.reflection_probe); + } else if (reflection_probe_instance_owner.owns(render_data.reflection_probe)) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_data.reflection_probe); ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas); if (!ra) { ERR_PRINT("reflection probe has no reflection atlas! Bug?"); @@ -3609,27 +4149,51 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & current_cluster_builder = nullptr; } - if (rb != nullptr && rb->sdfgi != nullptr) { - rb->sdfgi->update_cascades(); - - rb->sdfgi->pre_process_gi(p_cam_transform, this); - } + render_state.voxel_gi_count = 0; - render_state.gi_probe_count = 0; - if (rb != nullptr && rb->sdfgi != nullptr) { - gi.setup_giprobes(render_state.render_buffers, render_state.cam_transform, *render_state.gi_probes, render_state.gi_probe_count, this); + if (rb != nullptr && is_dynamic_gi_supported()) { + if (rb->sdfgi) { + rb->sdfgi->update_cascades(); + rb->sdfgi->pre_process_gi(render_data.cam_transform, &render_data, this); + rb->sdfgi->update_light(); + } - rb->sdfgi->update_light(); + gi.setup_voxel_gi_instances(render_data.render_buffers, render_data.cam_transform, *render_data.voxel_gi_instances, render_state.voxel_gi_count, this); } render_state.depth_prepass_used = false; //calls _pre_opaque_render between depth pre-pass and opaque pass - _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, *render_state.gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); + if (current_cluster_builder != nullptr) { + render_data.cluster_buffer = current_cluster_builder->get_cluster_buffer(); + render_data.cluster_size = current_cluster_builder->get_cluster_size(); + render_data.cluster_max_elements = current_cluster_builder->get_max_cluster_elements(); + } + + _render_scene(&render_data, clear_color); if (p_render_buffers.is_valid()) { - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { + /* + _debug_draw_cluster(p_render_buffers); + + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(&render_data); + */ + + _render_buffers_debug_draw(p_render_buffers, p_shadow_atlas, p_occluder_debug_tex); + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) { + rb->sdfgi->debug_draw(render_data.cam_projection, render_data.cam_transform, rb->width, rb->height, rb->render_target, rb->texture); + } + } +} + +void RendererSceneRenderRD::_debug_draw_cluster(RID p_render_buffers) { + if (p_render_buffers.is_valid() && current_cluster_builder != nullptr) { + RS::ViewportDebugDraw dd = get_debug_draw_mode(); + + if (dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX; - switch (debug_draw) { + switch (dd) { case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS: elem_type = ClusterBuilderRD::ELEMENT_TYPE_OMNI_LIGHT; break; @@ -3647,18 +4211,10 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & } current_cluster_builder->debug(elem_type); } - - RENDER_TIMESTAMP("Tonemap"); - - _render_buffers_post_process_and_tonemap(p_render_buffers, p_environment, p_camera_effects, p_cam_projection); - _render_buffers_debug_draw(p_render_buffers, p_shadow_atlas); - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) { - rb->sdfgi->debug_draw(p_cam_projection, p_cam_transform, rb->width, rb->height, rb->render_target, rb->texture); - } } } -void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region) { +void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region, RendererScene::RenderInfo *p_render_info) { LightInstance *light_instance = light_instance_owner.getornull(p_light); ERR_FAIL_COND(!light_instance); @@ -3668,6 +4224,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, bool using_dual_paraboloid = false; bool using_dual_paraboloid_flip = false; + Vector2i dual_paraboloid_offset; RID render_fb; RID render_texture; float zfar; @@ -3679,7 +4236,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, bool flip_y = false; CameraMatrix light_projection; - Transform light_transform; + Transform3D light_transform; if (storage->light_get_type(light_instance->light) == RS::LIGHT_DIRECTIONAL) { //set pssm stuff @@ -3761,6 +4318,9 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) { + bool wrap = (shadow + 1) % shadow_atlas->quadrants[quadrant].subdivision == 0; + dual_paraboloid_offset = wrap ? Vector2i(1 - shadow_atlas->quadrants[quadrant].subdivision, 1) : Vector2i(1, 0); + if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { ShadowCubemap *cubemap = _get_shadow_cubemap(shadow_size / 2); @@ -3780,12 +4340,16 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, } } else { + atlas_rect.position.x += 1; + atlas_rect.position.y += 1; + atlas_rect.size.x -= 2; + atlas_rect.size.y -= 2; + + atlas_rect.position += p_pass * atlas_rect.size * dual_paraboloid_offset; + light_projection = light_instance->shadow_transform[0].camera; light_transform = light_instance->shadow_transform[0].transform; - atlas_rect.size.height /= 2; - atlas_rect.position.y += p_pass * atlas_rect.size.height; - using_dual_paraboloid = true; using_dual_paraboloid_flip = p_pass == 1; render_fb = shadow_atlas->fb; @@ -3804,7 +4368,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, if (render_cubemap) { //rendering to cubemap - _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true, p_render_info); if (finalize_cubemap) { _render_shadow_process(); _render_shadow_end(); @@ -3814,10 +4378,9 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, atlas_rect_norm.position.y /= float(atlas_size); atlas_rect_norm.size.x /= float(atlas_size); atlas_rect_norm.size.y /= float(atlas_size); - atlas_rect_norm.size.height /= 2; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false); - atlas_rect_norm.position.y += atlas_rect_norm.size.height; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true); + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position += Vector2(dual_paraboloid_offset) * atlas_rect_norm.size; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), true); //restore transform so it can be properly used light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 0); @@ -3825,15 +4388,15 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, } else { //render shadow - _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass, p_render_info); } } -void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderRD::render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region); } -void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) { +void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<GeometryInstance *> &p_instances) { ERR_FAIL_COND(!storage->particles_collision_is_heightfield(p_collider)); Vector3 extents = storage->particles_collision_get_extents(p_collider) * p_transform.basis.get_scale(); CameraMatrix cm; @@ -3842,7 +4405,7 @@ void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, Vector3 cam_pos = p_transform.origin; cam_pos.y += extents.y; - Transform cam_xform; + Transform3D cam_xform; cam_xform.set_look_at(cam_pos, cam_pos - p_transform.basis.get_axis(Vector3::AXIS_Y), -p_transform.basis.get_axis(Vector3::AXIS_Z).normalized()); RID fb = storage->particles_collision_get_heightfield_framebuffer(p_collider); @@ -3881,27 +4444,29 @@ bool RendererSceneRenderRD::free(RID p_rid) { } reflection_atlas_owner.free(p_rid); } else if (reflection_probe_instance_owner.owns(p_rid)) { - //not much to delete, just free it - //ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid); + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid); + _free_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id); reflection_probe_release_atlas_index(p_rid); reflection_probe_instance_owner.free(p_rid); } else if (decal_instance_owner.owns(p_rid)) { + DecalInstance *di = decal_instance_owner.getornull(p_rid); + _free_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id); decal_instance_owner.free(p_rid); } else if (lightmap_instance_owner.owns(p_rid)) { lightmap_instance_owner.free(p_rid); - } else if (gi.gi_probe_instance_owner.owns(p_rid)) { - RendererSceneGIRD::GIProbeInstance *gi_probe = gi.gi_probe_instance_owner.getornull(p_rid); - if (gi_probe->texture.is_valid()) { - RD::get_singleton()->free(gi_probe->texture); - RD::get_singleton()->free(gi_probe->write_buffer); + } else if (gi.voxel_gi_instance_owner.owns(p_rid)) { + RendererSceneGIRD::VoxelGIInstance *voxel_gi = gi.voxel_gi_instance_owner.getornull(p_rid); + if (voxel_gi->texture.is_valid()) { + RD::get_singleton()->free(voxel_gi->texture); + RD::get_singleton()->free(voxel_gi->write_buffer); } - for (int i = 0; i < gi_probe->dynamic_maps.size(); i++) { - RD::get_singleton()->free(gi_probe->dynamic_maps[i].texture); - RD::get_singleton()->free(gi_probe->dynamic_maps[i].depth); + for (int i = 0; i < voxel_gi->dynamic_maps.size(); i++) { + RD::get_singleton()->free(voxel_gi->dynamic_maps[i].texture); + RD::get_singleton()->free(voxel_gi->dynamic_maps[i].depth); } - gi.gi_probe_instance_owner.free(p_rid); + gi.voxel_gi_instance_owner.free(p_rid); } else if (sky.sky_owner.owns(p_rid)) { sky.update_dirty_skys(); sky.free_sky(p_rid); @@ -3917,9 +4482,18 @@ bool RendererSceneRenderRD::free(RID p_rid) { uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK; shadow_atlas->quadrants[q].shadows.write[s].owner = RID(); + + if (key & ShadowAtlas::OMNI_LIGHT_FLAG) { + // Omni lights use two atlas spots, make sure to clear the other as well + shadow_atlas->quadrants[q].shadows.write[s + 1].owner = RID(); + } + shadow_atlas->shadow_owners.erase(p_rid); } + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + _free_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id); + } light_instance_owner.free(p_rid); } else if (shadow_atlas_owner.owns(p_rid)) { @@ -4024,7 +4598,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(albedo_alpha_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(albedo_alpha_tex); ret.push_back(img); @@ -4033,7 +4607,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(normal_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(normal_tex); ret.push_back(img); @@ -4042,7 +4616,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(orm_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(orm_tex); ret.push_back(img); @@ -4051,7 +4625,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(emission_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBAH, data); RD::get_singleton()->free(emission_tex); ret.push_back(img); @@ -4091,11 +4665,6 @@ int RendererSceneRenderRD::get_max_directional_lights() const { return cluster.max_directional_lights; } -bool RendererSceneRenderRD::is_low_end() const { - // by default we switch this on this (may be ignored in some implementations) - return GLOBAL_GET("rendering/driver/rd_renderer/use_low_end_renderer"); -} - bool RendererSceneRenderRD::is_dynamic_gi_supported() const { // usable by default (unless low end = true) return true; @@ -4111,30 +4680,28 @@ bool RendererSceneRenderRD::is_volumetric_supported() const { return true; } -RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { - max_cluster_elements = GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements"); +uint32_t RendererSceneRenderRD::get_max_elements() const { + return GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements"); +} +RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { storage = p_storage; singleton = this; +} + +void RendererSceneRenderRD::init() { + max_cluster_elements = get_max_elements(); directional_shadow.size = GLOBAL_GET("rendering/shadows/directional_shadow/size"); directional_shadow.use_16_bits = GLOBAL_GET("rendering/shadows/directional_shadow/16_bits"); - uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); - - low_end = is_low_end(); - - if (textures_per_stage < 48) { - low_end = true; - } - /* SKY SHADER */ sky.init(storage); /* GI */ - if (!low_end && is_dynamic_gi_supported()) { + if (is_dynamic_gi_supported()) { gi.init(storage, &sky); } @@ -4172,7 +4739,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { cluster.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); } - if (!low_end && is_volumetric_supported()) { + if (is_volumetric_supported()) { String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(cluster.max_directional_lights) + "\n"; Vector<String> volumetric_fog_modes; volumetric_fog_modes.push_back("\n#define MODE_DENSITY\n"); @@ -4218,6 +4785,9 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/environment/volumetric_fog/volume_size"), GLOBAL_GET("rendering/environment/volumetric_fog/volume_depth")); environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/environment/volumetric_fog/use_filter")); + decals_set_filter(RS::DecalFilter(int(GLOBAL_GET("rendering/textures/decals/filter")))); + light_projectors_set_filter(RS::LightProjectorFilter(int(GLOBAL_GET("rendering/textures/light_projectors/filter")))); + cull_argument.set_page_pool(&cull_argument_pool); } @@ -4230,7 +4800,7 @@ RendererSceneRenderRD::~RendererSceneRenderRD() { RD::get_singleton()->free(sky.sky_scene_state.uniform_set); } - if (!low_end) { + if (is_dynamic_gi_supported()) { gi.free(); volumetric_fog.shader.version_free(volumetric_fog.shader_version); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 264c0e4276..37533baecf 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -40,9 +40,51 @@ #include "servers/rendering/renderer_rd/renderer_scene_sky_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" #include "servers/rendering/renderer_rd/shaders/volumetric_fog.glsl.gen.h" +#include "servers/rendering/renderer_scene.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering/rendering_device.h" +struct RenderDataRD { + RID render_buffers = RID(); + + Transform3D cam_transform = Transform3D(); + CameraMatrix cam_projection = CameraMatrix(); + bool cam_ortogonal = false; + + // For stereo rendering + uint32_t view_count = 1; + CameraMatrix view_projection[RendererSceneRender::MAX_RENDER_VIEWS]; + + float z_near = 0.0; + float z_far = 0.0; + + const PagedArray<RendererSceneRender::GeometryInstance *> *instances = nullptr; + const PagedArray<RID> *lights = nullptr; + const PagedArray<RID> *reflection_probes = nullptr; + const PagedArray<RID> *voxel_gi_instances = nullptr; + const PagedArray<RID> *decals = nullptr; + const PagedArray<RID> *lightmaps = nullptr; + RID environment = RID(); + RID camera_effects = RID(); + RID shadow_atlas = RID(); + RID reflection_atlas = RID(); + RID reflection_probe = RID(); + int reflection_probe_pass = 0; + + float lod_distance_multiplier = 0.0; + Plane lod_camera_plane = Plane(); + float screen_lod_threshold = 0.0; + + RID cluster_buffer = RID(); + uint32_t cluster_size = 0; + uint32_t cluster_max_elements = 0; + + uint32_t directional_light_count = 0; + bool directional_light_soft_shadows = false; + + RendererScene::RenderInfo *render_info = nullptr; +}; + class RendererSceneRenderRD : public RendererSceneRender { friend RendererSceneSkyRD; friend RendererSceneGIRD; @@ -51,48 +93,52 @@ protected: RendererStorageRD *storage; double time; double time_step = 0; - bool low_end = false; // If true GI and Volumetric fog are disabled struct RenderBufferData { - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa) = 0; + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) = 0; virtual ~RenderBufferData() {} }; virtual RenderBufferData *_create_render_buffer_data() = 0; - void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); - void _setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform); - void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment); + void _setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows); + void _setup_decals(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform); + void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform3D &p_camera_inverse_transform, RID p_environment); - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_cluster_max_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_color) = 0; virtual void _render_shadow_begin() = 0; - virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true) = 0; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) = 0; virtual void _render_shadow_process() = 0; virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) = 0; - virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; + virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0; - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) = 0; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) = 0; void _debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); + void _debug_draw_cluster(RID p_render_buffers); RenderBufferData *render_buffers_get_data(RID p_render_buffers); virtual void _base_uniforms_changed() = 0; - virtual void _render_buffers_uniform_set_changed(RID p_render_buffers) = 0; virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) = 0; void _process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection); void _process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive); void _process_sss(RID p_render_buffers, const CameraMatrix &p_camera); - bool _needs_post_prepass_render(bool p_use_gi); - void _post_prepass_render(bool p_use_gi); - void _pre_resolve_render(bool p_use_gi); + bool _needs_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi); + void _post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi); + void _pre_resolve_render(RenderDataRD *p_render_data, bool p_use_gi); + + void _pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer); - void _pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer); - uint32_t _get_render_state_directional_light_count() const; + void _render_buffers_copy_screen_texture(const RenderDataRD *p_render_data); + void _render_buffers_copy_depth_texture(const RenderDataRD *p_render_data); + void _render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data); + void _post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data); + void _disable_clear_request(const RenderDataRD *p_render_data); // needed for a single argument calls (material and uv2) PagedArrayPool<GeometryInstance *> cull_argument_pool; @@ -107,7 +153,26 @@ protected: } else { return nullptr; } - } + }; + + //used for mobile renderer mostly + + typedef int32_t ForwardID; + + enum ForwardIDType { + FORWARD_ID_TYPE_OMNI_LIGHT, + FORWARD_ID_TYPE_SPOT_LIGHT, + FORWARD_ID_TYPE_REFLECTION_PROBE, + FORWARD_ID_TYPE_DECAL, + FORWARD_ID_MAX, + }; + + virtual ForwardID _allocate_forward_id(ForwardIDType p_type) { return -1; } + virtual void _free_forward_id(ForwardIDType p_type, ForwardID p_id) {} + virtual void _map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) {} + virtual bool _uses_forward_ids() const { return false; } + + virtual void _update_shader_quality_settings() {} private: RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; @@ -150,9 +215,11 @@ private: uint32_t render_step = 0; uint64_t last_pass = 0; - uint32_t render_index = 0; + uint32_t cull_mask = 0; + + ForwardID forward_id = -1; - Transform transform; + Transform3D transform; }; mutable RID_Owner<ReflectionProbeInstance> reflection_probe_instance_owner; @@ -161,7 +228,9 @@ private: struct DecalInstance { RID decal; - Transform transform; + Transform3D transform; + uint32_t cull_mask; + ForwardID forward_id = -1; }; mutable RID_Owner<DecalInstance> decal_instance_owner; @@ -170,7 +239,7 @@ private: struct LightmapInstance { RID lightmap; - Transform transform; + Transform3D transform; }; mutable RID_Owner<LightmapInstance> lightmap_instance_owner; @@ -186,7 +255,8 @@ private: struct ShadowAtlas { enum { QUADRANT_SHIFT = 27, - SHADOW_INDEX_MASK = (1 << QUADRANT_SHIFT) - 1, + OMNI_LIGHT_FLAG = 1 << 26, + SHADOW_INDEX_MASK = OMNI_LIGHT_FLAG - 1, SHADOW_INVALID = 0xFFFFFFFF }; @@ -230,7 +300,9 @@ private: void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + void _shadow_atlas_invalidate_shadow(RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, RendererSceneRenderRD::ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx); bool _shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); + bool _shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set RS::ShadowQuality directional_shadow_quality = RS::SHADOW_QUALITY_MAX; @@ -245,6 +317,8 @@ private: int directional_soft_shadow_samples = 0; int penumbra_shadow_samples = 0; int soft_shadow_samples = 0; + RS::DecalFilter decals_filter = RS::DECAL_FILTER_LINEAR_MIPMAPS; + RS::LightProjectorFilter light_projectors_filter = RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS; /* DIRECTIONAL SHADOW */ @@ -278,7 +352,7 @@ private: struct LightInstance { struct ShadowTransform { CameraMatrix camera; - Transform transform; + Transform3D transform; float farplane; float split; float bias_scale; @@ -295,7 +369,7 @@ private: AABB aabb; RID self; RID light; - Transform transform; + Transform3D transform; Vector3 light_vector; Vector3 spot_vector; @@ -305,17 +379,15 @@ private: uint64_t last_scene_pass = 0; uint64_t last_scene_shadow_pass = 0; uint64_t last_pass = 0; - uint32_t light_index = 0; + uint32_t cull_mask = 0; uint32_t light_directional_index = 0; - uint32_t current_shadow_atlas_key = 0; - - Vector2 dp; - Rect2 directional_rect; Set<RID> shadow_atlases; //shadow atlases where this light is registered + ForwardID forward_id = -1; + LightInstance() {} }; @@ -376,6 +448,7 @@ private: RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; RS::ViewportScreenSpaceAA screen_space_aa = RS::VIEWPORT_SCREEN_SPACE_AA_DISABLED; bool use_debanding = false; + uint32_t view_count = 1; RID render_target; @@ -383,6 +456,7 @@ private: RID texture; //main texture for rendering to, must be filled after done rendering RID depth_texture; //main depth texture + RID texture_fb; // framebuffer for the main texture, ONLY USED FOR MOBILE RENDERER POST EFFECTS, DO NOT USE FOR RENDERING 3D!!! RendererSceneGIRD::SDFGI *sdfgi = nullptr; VolumetricFog *volumetric_fog = nullptr; @@ -398,6 +472,11 @@ private: RID texture; int width; int height; + + // only used on mobile renderer + RID fb; + RID half_texture; + RID half_fb; }; Vector<Mipmap> mipmaps; @@ -405,9 +484,25 @@ private: Blur blur[2]; //the second one starts from the first mipmap + struct WeightBuffers { + RID weight; + RID fb; // FB with both texture and weight + }; + + // 2 full size, 2 half size + WeightBuffers weight_buffers[4]; // Only used in raster + RID base_weight_fb; // base buffer for weight + + RID depth_back_texture; + RID depth_back_fb; // only used on mobile + struct Luminance { Vector<RID> reduce; RID current; + + // used only on mobile renderer + Vector<RID> fb; + RID current_fb; } luminance; struct SSAO { @@ -419,6 +514,10 @@ private: Vector<RID> ao_pong_slices; RID ao_final; RID importance_map[2]; + + RID downsample_uniform_set; + RID gather_uniform_set; + RID importance_map_uniform_set; } ssao; struct SSR { @@ -440,16 +539,18 @@ private: void _free_render_buffer_data(RenderBuffers *rb); void _allocate_blur_textures(RenderBuffers *rb); + void _allocate_depth_backbuffer_textures(RenderBuffers *rb); void _allocate_luminance_textures(RenderBuffers *rb); - void _render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas); - void _render_buffers_post_process_and_tonemap(RID p_render_buffers, RID p_environment, RID p_camera_effects, const CameraMatrix &p_projection); + void _render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer); /* Cluster */ struct Cluster { /* Scene State UBO */ + // !BAS! Most data here is not just used by our clustering logic but also by other lighting implementations. Maybe rename this struct to something more appropriate + enum { REFLECTION_AMBIENT_DISABLED = 0, REFLECTION_AMBIENT_ENVIRONMENT = 1, @@ -463,8 +564,8 @@ private: uint32_t mask; float ambient[3]; // ambient color, float intensity; - bool exterior; - bool box_project; + uint32_t exterior; + uint32_t box_project; uint32_t ambient_mode; uint32_t pad; float local_matrix[16]; // up to here for spot and omni, rest is for directional @@ -473,7 +574,7 @@ private: struct LightData { float position[3]; float inv_radius; - float direction[3]; + float direction[3]; // in omni, x and y are used for dual paraboloid offset float size; float color[3]; @@ -493,7 +594,7 @@ private: float soft_shadow_scale; uint32_t mask; float shadow_volumetric_fog_fade; - uint32_t pad; + uint32_t bake_mode; float projector_rect[4]; }; @@ -510,7 +611,8 @@ private: uint32_t shadow_enabled; float fade_from; float fade_to; - uint32_t pad[3]; + uint32_t pad[2]; + uint32_t bake_mode; float shadow_volumetric_fog_fade; float shadow_bias[4]; float shadow_normal_bias[4]; @@ -587,38 +689,19 @@ private: } cluster; struct RenderState { - RID render_buffers; - Transform cam_transform; - CameraMatrix cam_projection; - bool cam_ortogonal = false; - const PagedArray<GeometryInstance *> *instances = nullptr; - const PagedArray<RID> *lights = nullptr; - const PagedArray<RID> *reflection_probes = nullptr; - const PagedArray<RID> *gi_probes = nullptr; - const PagedArray<RID> *decals = nullptr; - const PagedArray<RID> *lightmaps = nullptr; - RID environment; - RID camera_effects; - RID shadow_atlas; - RID reflection_atlas; - RID reflection_probe; - int reflection_probe_pass = 0; - float screen_lod_threshold = 0.0; - - const RenderShadowData *render_shadows = nullptr; + const RendererSceneRender::RenderShadowData *render_shadows = nullptr; int render_shadow_count = 0; - const RenderSDFGIData *render_sdfgi_regions = nullptr; + const RendererSceneRender::RenderSDFGIData *render_sdfgi_regions = nullptr; int render_sdfgi_region_count = 0; - const RenderSDFGIUpdateData *sdfgi_update_data = nullptr; + const RendererSceneRender::RenderSDFGIUpdateData *sdfgi_update_data = nullptr; - uint32_t directional_light_count = 0; - uint32_t gi_probe_count = 0; + uint32_t voxel_gi_count = 0; LocalVector<int> cube_shadows; LocalVector<int> shadows; LocalVector<int> directional_shadows; - bool depth_prepass_used; + bool depth_prepass_used; // this does not seem used anywhere... } render_state; struct VolumetricFog { @@ -644,7 +727,7 @@ private: int last_shadow_filter = -1; - Transform prev_cam_transform; + Transform3D prev_cam_transform; }; enum { @@ -673,7 +756,7 @@ private: float detail_spread; float gi_inject; - uint32_t max_gi_probes; + uint32_t max_voxel_gi_instances; uint32_t cluster_type_size; float screen_size[2]; @@ -702,7 +785,7 @@ private: bool volumetric_fog_filter_active = true; void _volumetric_fog_erase(RenderBuffers *rb); - void _update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count); + void _update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform3D &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count); RID shadow_sampler; @@ -718,18 +801,18 @@ private: uint32_t max_cluster_elements = 512; - void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true); + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true, RendererScene::RenderInfo *p_render_info = nullptr); public: - virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance) = 0; + virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) = 0; virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) = 0; /* SHADOW ATLAS API */ - RID shadow_atlas_create(); - void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false); - void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision); - bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version); + virtual RID shadow_atlas_create() override; + virtual void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false) override; + virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) override; + virtual bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) override; _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { ShadowAtlas *atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND_V(!atlas, false); @@ -748,9 +831,9 @@ public: return Size2(atlas->size, atlas->size); } - void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false); - int get_directional_light_shadow_size(RID p_light_intance); - void set_directional_shadow_count(int p_count); + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false) override; + virtual int get_directional_light_shadow_size(RID p_light_intance) override; + virtual void set_directional_shadow_count(int p_count) override; _FORCE_INLINE_ RID directional_shadow_get_texture() { return directional_shadow.depth; @@ -762,43 +845,43 @@ public: /* SDFGI UPDATE */ - virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position); - virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const; - virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const; - virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const; + virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) override; + virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const override; + virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const override; + virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const override; RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } /* SKY API */ - virtual RID sky_allocate(); - virtual void sky_initialize(RID p_rid); + virtual RID sky_allocate() override; + virtual void sky_initialize(RID p_rid) override; - void sky_set_radiance_size(RID p_sky, int p_radiance_size); - void sky_set_mode(RID p_sky, RS::SkyMode p_mode); - void sky_set_material(RID p_sky, RID p_material); - Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size); + virtual void sky_set_radiance_size(RID p_sky, int p_radiance_size) override; + virtual void sky_set_mode(RID p_sky, RS::SkyMode p_mode) override; + virtual void sky_set_material(RID p_sky, RID p_material) override; + virtual Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) override; /* ENVIRONMENT API */ - virtual RID environment_allocate(); - virtual void environment_initialize(RID p_rid); + virtual RID environment_allocate() override; + virtual void environment_initialize(RID p_rid) override; - void environment_set_background(RID p_env, RS::EnvironmentBG p_bg); - void environment_set_sky(RID p_env, RID p_sky); - void environment_set_sky_custom_fov(RID p_env, float p_scale); - void environment_set_sky_orientation(RID p_env, const Basis &p_orientation); - void environment_set_bg_color(RID p_env, const Color &p_color); - void environment_set_bg_energy(RID p_env, float p_energy); - void environment_set_canvas_max_layer(RID p_env, int p_max_layer); - void environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient = RS::ENV_AMBIENT_SOURCE_BG, float p_energy = 1.0, float p_sky_contribution = 0.0, RS::EnvironmentReflectionSource p_reflection_source = RS::ENV_REFLECTION_SOURCE_BG, const Color &p_ao_color = Color()); + virtual void environment_set_background(RID p_env, RS::EnvironmentBG p_bg) override; + virtual void environment_set_sky(RID p_env, RID p_sky) override; + virtual void environment_set_sky_custom_fov(RID p_env, float p_scale) override; + virtual void environment_set_sky_orientation(RID p_env, const Basis &p_orientation) override; + virtual void environment_set_bg_color(RID p_env, const Color &p_color) override; + virtual void environment_set_bg_energy(RID p_env, float p_energy) override; + virtual void environment_set_canvas_max_layer(RID p_env, int p_max_layer) override; + virtual void environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient = RS::ENV_AMBIENT_SOURCE_BG, float p_energy = 1.0, float p_sky_contribution = 0.0, RS::EnvironmentReflectionSource p_reflection_source = RS::ENV_REFLECTION_SOURCE_BG, const Color &p_ao_color = Color()) override; - RS::EnvironmentBG environment_get_background(RID p_env) const; + virtual RS::EnvironmentBG environment_get_background(RID p_env) const override; RID environment_get_sky(RID p_env) const; float environment_get_sky_custom_fov(RID p_env) const; Basis environment_get_sky_orientation(RID p_env) const; Color environment_get_bg_color(RID p_env) const; float environment_get_bg_energy(RID p_env) const; - int environment_get_canvas_max_layer(RID p_env) const; + virtual int environment_get_canvas_max_layer(RID p_env) const override; Color environment_get_ambient_light_color(RID p_env) const; RS::EnvironmentAmbientSource environment_get_ambient_source(RID p_env) const; float environment_get_ambient_light_energy(RID p_env) const; @@ -806,13 +889,13 @@ public: RS::EnvironmentReflectionSource environment_get_reflection_source(RID p_env) const; Color environment_get_ao_color(RID p_env) const; - bool is_environment(RID p_env) const; + virtual bool is_environment(RID p_env) const override; - void environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap); - void environment_glow_set_use_bicubic_upscale(bool p_enable); - void environment_glow_set_use_high_quality(bool p_enable); + virtual void environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap) override; + virtual void environment_glow_set_use_bicubic_upscale(bool p_enable) override; + virtual void environment_glow_set_use_high_quality(bool p_enable) override; - void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective); + virtual void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective) override; bool environment_is_fog_enabled(RID p_env) const; Color environment_get_fog_light_color(RID p_env) const; float environment_get_fog_light_energy(RID p_env) const; @@ -822,59 +905,65 @@ public: float environment_get_fog_height_density(RID p_env) const; float environment_get_fog_aerial_perspective(RID p_env) const; - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); + virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) override; - virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth); - virtual void environment_set_volumetric_fog_filter_active(bool p_enable); + virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) override; + virtual void environment_set_volumetric_fog_filter_active(bool p_enable) override; - void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); - void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); - void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + virtual void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) override; + virtual void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) override; + virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override; bool environment_is_ssao_enabled(RID p_env) const; float environment_get_ssao_ao_affect(RID p_env) const; float environment_get_ssao_light_affect(RID p_env) const; bool environment_is_ssr_enabled(RID p_env) const; bool environment_is_sdfgi_enabled(RID p_env) const; - virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); - virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count); - virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames); - virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update); + virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) override; + virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) override; + virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) override; + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) override; - void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); + virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) override; RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const; - void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale); - void environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction); + virtual void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) override; + virtual void environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) override; + + virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) override; - virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size); + virtual RID camera_effects_allocate() override; + virtual void camera_effects_initialize(RID p_rid) override; - virtual RID camera_effects_allocate(); - virtual void camera_effects_initialize(RID p_rid); + virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) override; + virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape) override; - virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter); - virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape); + virtual void camera_effects_set_dof_blur(RID p_camera_effects, bool p_far_enable, float p_far_distance, float p_far_transition, bool p_near_enable, float p_near_distance, float p_near_transition, float p_amount) override; + virtual void camera_effects_set_custom_exposure(RID p_camera_effects, bool p_enable, float p_exposure) override; - virtual void camera_effects_set_dof_blur(RID p_camera_effects, bool p_far_enable, float p_far_distance, float p_far_transition, bool p_near_enable, float p_near_distance, float p_near_transition, float p_amount); - virtual void camera_effects_set_custom_exposure(RID p_camera_effects, bool p_enable, float p_exposure); + bool camera_effects_uses_dof(RID p_camera_effects) { + CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects); - RID light_instance_create(RID p_light); - void light_instance_set_transform(RID p_light_instance, const Transform &p_transform); - void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb); - void light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2()); - void light_instance_mark_visible(RID p_light_instance); + return camfx && (camfx->dof_blur_near_enabled || camfx->dof_blur_far_enabled) && camfx->dof_blur_amount > 0.0; + } + + virtual RID light_instance_create(RID p_light) override; + virtual void light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform) override; + virtual void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) override; + virtual void light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2()) override; + virtual void light_instance_mark_visible(RID p_light_instance) override; _FORCE_INLINE_ RID light_instance_get_base_light(RID p_light_instance) { LightInstance *li = light_instance_owner.getornull(p_light_instance); return li->light; } - _FORCE_INLINE_ Transform light_instance_get_base_transform(RID p_light_instance) { + _FORCE_INLINE_ Transform3D light_instance_get_base_transform(RID p_light_instance) { LightInstance *li = light_instance_owner.getornull(p_light_instance); return li->transform; } - _FORCE_INLINE_ Rect2 light_instance_get_shadow_atlas_rect(RID p_light_instance, RID p_shadow_atlas) { + _FORCE_INLINE_ Rect2 light_instance_get_shadow_atlas_rect(RID p_light_instance, RID p_shadow_atlas, Vector2i &r_omni_offset) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); LightInstance *li = light_instance_owner.getornull(p_light_instance); uint32_t key = shadow_atlas->shadow_owners[li->self]; @@ -894,6 +983,16 @@ public: x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; + if (key & ShadowAtlas::OMNI_LIGHT_FLAG) { + if (((shadow + 1) % shadow_atlas->quadrants[quadrant].subdivision) == 0) { + r_omni_offset.x = 1 - int(shadow_atlas->quadrants[quadrant].subdivision); + r_omni_offset.y = 1; + } else { + r_omni_offset.x = 1; + r_omni_offset.y = 0; + } + } + uint32_t width = shadow_size; uint32_t height = shadow_size; @@ -926,7 +1025,7 @@ public: return float(1.0) / shadow_size; } - _FORCE_INLINE_ Transform + _FORCE_INLINE_ Transform3D light_instance_get_shadow_transform(RID p_light_instance, int p_index) { LightInstance *li = light_instance_owner.getornull(p_light_instance); return li->shadow_transform[p_index].transform; @@ -974,14 +1073,9 @@ public: return li->last_pass; } - _FORCE_INLINE_ void light_instance_set_index(RID p_light_instance, uint32_t p_index) { - LightInstance *li = light_instance_owner.getornull(p_light_instance); - li->light_index = p_index; - } - - _FORCE_INLINE_ uint32_t light_instance_get_index(RID p_light_instance) { + _FORCE_INLINE_ ForwardID light_instance_get_forward_id(RID p_light_instance) { LightInstance *li = light_instance_owner.getornull(p_light_instance); - return li->light_index; + return li->forward_id; } _FORCE_INLINE_ RS::LightType light_instance_get_type(RID p_light_instance) { @@ -989,9 +1083,9 @@ public: return li->light_type; } - virtual RID reflection_atlas_create(); - virtual void reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count); - virtual int reflection_atlas_get_size(RID p_ref_atlas) const; + virtual RID reflection_atlas_create() override; + virtual void reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count) override; + virtual int reflection_atlas_get_size(RID p_ref_atlas) const override; _FORCE_INLINE_ RID reflection_atlas_get_texture(RID p_ref_atlas) { ReflectionAtlas *atlas = reflection_atlas_owner.getornull(p_ref_atlas); @@ -999,13 +1093,14 @@ public: return atlas->reflection; } - virtual RID reflection_probe_instance_create(RID p_probe); - virtual void reflection_probe_instance_set_transform(RID p_instance, const Transform &p_transform); - virtual void reflection_probe_release_atlas_index(RID p_instance); - virtual bool reflection_probe_instance_needs_redraw(RID p_instance); - virtual bool reflection_probe_instance_has_reflection(RID p_instance); - virtual bool reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas); - virtual bool reflection_probe_instance_postprocess_step(RID p_instance); + virtual RID reflection_probe_instance_create(RID p_probe) override; + virtual void reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform) override; + virtual void reflection_probe_release_atlas_index(RID p_instance) override; + virtual bool reflection_probe_instance_needs_redraw(RID p_instance) override; + virtual bool reflection_probe_instance_has_reflection(RID p_instance) override; + virtual bool reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas) override; + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth); + virtual bool reflection_probe_instance_postprocess_step(RID p_instance) override; uint32_t reflection_probe_instance_get_resolution(RID p_instance); RID reflection_probe_instance_get_framebuffer(RID p_instance, int p_index); @@ -1018,17 +1113,11 @@ public: return rpi->probe; } - _FORCE_INLINE_ void reflection_probe_instance_set_render_index(RID p_instance, uint32_t p_render_index) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND(!rpi); - rpi->render_index = p_render_index; - } - - _FORCE_INLINE_ uint32_t reflection_probe_instance_get_render_index(RID p_instance) { + _FORCE_INLINE_ ForwardID reflection_probe_instance_get_forward_id(RID p_instance) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, 0); - return rpi->render_index; + return rpi->forward_id; } _FORCE_INLINE_ void reflection_probe_instance_set_render_pass(RID p_instance, uint32_t p_render_pass) { @@ -1044,9 +1133,9 @@ public: return rpi->last_pass; } - _FORCE_INLINE_ Transform reflection_probe_instance_get_transform(RID p_instance) { + _FORCE_INLINE_ Transform3D reflection_probe_instance_get_transform(RID p_instance) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND_V(!rpi, Transform()); + ERR_FAIL_COND_V(!rpi, Transform3D()); return rpi->transform; } @@ -1058,21 +1147,26 @@ public: return rpi->atlas_index; } - virtual RID decal_instance_create(RID p_decal); - virtual void decal_instance_set_transform(RID p_decal, const Transform &p_transform); + virtual RID decal_instance_create(RID p_decal) override; + virtual void decal_instance_set_transform(RID p_decal, const Transform3D &p_transform) override; _FORCE_INLINE_ RID decal_instance_get_base(RID p_decal) const { DecalInstance *decal = decal_instance_owner.getornull(p_decal); return decal->decal; } - _FORCE_INLINE_ Transform decal_instance_get_transform(RID p_decal) const { + _FORCE_INLINE_ ForwardID decal_instance_get_forward_id(RID p_decal) const { + DecalInstance *decal = decal_instance_owner.getornull(p_decal); + return decal->forward_id; + } + + _FORCE_INLINE_ Transform3D decal_instance_get_transform(RID p_decal) const { DecalInstance *decal = decal_instance_owner.getornull(p_decal); return decal->transform; } - virtual RID lightmap_instance_create(RID p_lightmap); - virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform); + virtual RID lightmap_instance_create(RID p_lightmap) override; + virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) override; _FORCE_INLINE_ bool lightmap_instance_is_valid(RID p_lightmap_instance) { return lightmap_instance_owner.getornull(p_lightmap_instance) != nullptr; } @@ -1081,29 +1175,32 @@ public: LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap_instance); return li->lightmap; } - _FORCE_INLINE_ Transform lightmap_instance_get_transform(RID p_lightmap_instance) { + _FORCE_INLINE_ Transform3D lightmap_instance_get_transform(RID p_lightmap_instance) { LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap_instance); return li->transform; } /* gi light probes */ - RID gi_probe_instance_create(RID p_base); - void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform); - bool gi_probe_needs_update(RID p_probe) const; - void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects); - void gi_probe_set_quality(RS::GIProbeQuality p_quality) { gi.gi_probe_quality = p_quality; } + virtual RID voxel_gi_instance_create(RID p_base) override; + virtual void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) override; + virtual bool voxel_gi_needs_update(RID p_probe) const override; + virtual void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects) override; + virtual void voxel_gi_set_quality(RS::VoxelGIQuality p_quality) override { gi.voxel_gi_quality = p_quality; } /* render buffers */ - RID render_buffers_create(); - void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding); - void gi_set_use_half_resolution(bool p_enable); + virtual RD::DataFormat _render_buffers_get_color_format(); + virtual bool _render_buffers_can_be_storage(); + virtual RID render_buffers_create() override; + virtual void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) override; + virtual void gi_set_use_half_resolution(bool p_enable) override; RID render_buffers_get_ao_texture(RID p_render_buffers); RID render_buffers_get_back_buffer_texture(RID p_render_buffers); - RID render_buffers_get_gi_probe_buffer(RID p_render_buffers); - RID render_buffers_get_default_gi_probe_buffer(); + RID render_buffers_get_back_depth_texture(RID p_render_buffers); + RID render_buffers_get_voxel_gi_buffer(RID p_render_buffers); + RID render_buffers_get_default_voxel_gi_buffer(); RID render_buffers_get_gi_ambient_texture(RID p_render_buffers); RID render_buffers_get_gi_reflection_texture(RID p_render_buffers); @@ -1126,30 +1223,34 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); - void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr); + virtual void render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr, RendererScene::RenderInfo *r_render_info = nullptr) override; - void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); + virtual void render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; - void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances); + virtual void render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<GeometryInstance *> &p_instances) override; - virtual void set_scene_pass(uint64_t p_pass) { + virtual void set_scene_pass(uint64_t p_pass) override { scene_pass = p_pass; } _FORCE_INLINE_ uint64_t get_scene_pass() { return scene_pass; } - virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit); - virtual bool screen_space_roughness_limiter_is_active() const; + virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) override; + virtual bool screen_space_roughness_limiter_is_active() const override; virtual float screen_space_roughness_limiter_get_amount() const; virtual float screen_space_roughness_limiter_get_limit() const; - virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality); + virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) override; RS::SubSurfaceScatteringQuality sub_surface_scattering_get_quality() const; - virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale); + virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) override; + + virtual void shadows_quality_set(RS::ShadowQuality p_quality) override; + virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality) override; + + virtual void decals_set_filter(RS::DecalFilter p_filter) override; + virtual void light_projectors_set_filter(RS::LightProjectorFilter p_filter) override; - virtual void shadows_quality_set(RS::ShadowQuality p_quality); - virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality); _FORCE_INLINE_ RS::ShadowQuality shadows_quality_get() const { return shadows_quality; } _FORCE_INLINE_ RS::ShadowQuality directional_shadow_quality_get() const { return directional_shadow_quality; } _FORCE_INLINE_ float shadows_quality_radius_get() const { return shadows_quality_radius; } @@ -1165,21 +1266,24 @@ public: _FORCE_INLINE_ int penumbra_shadow_samples_get() const { return penumbra_shadow_samples; } _FORCE_INLINE_ int soft_shadow_samples_get() const { return soft_shadow_samples; } + _FORCE_INLINE_ RS::LightProjectorFilter light_projectors_get_filter() const { return light_projectors_filter; } + _FORCE_INLINE_ RS::DecalFilter decals_get_filter() const { return decals_filter; } + int get_roughness_layers() const; bool is_using_radiance_cubemap_array() const; - virtual TypedArray<Image> bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size); + virtual TypedArray<Image> bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size) override; - virtual bool free(RID p_rid); + virtual bool free(RID p_rid) override; - virtual void update(); + virtual void update() override; - virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw); + virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) override; _FORCE_INLINE_ RS::ViewportDebugDraw get_debug_draw_mode() const { return debug_draw; } - void set_time(double p_time, double p_step); + virtual void set_time(double p_time, double p_step) override; RID get_reflection_probe_buffer(); RID get_omni_light_buffer(); @@ -1188,13 +1292,14 @@ public: RID get_decal_buffer(); int get_max_directional_lights() const; - void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir); - - virtual bool is_low_end() const; + virtual void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) override; virtual bool is_dynamic_gi_supported() const; virtual bool is_clustered_enabled() const; virtual bool is_volumetric_supported() const; + virtual uint32_t get_max_elements() const; + + void init(); RendererSceneRenderRD(RendererStorageRD *p_storage); ~RendererSceneRenderRD(); diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp index 769335ac16..9e85608f1e 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp @@ -30,7 +30,9 @@ #include "renderer_scene_sky_rd.h" #include "core/config/project_settings.h" +#include "core/math/math_defs.h" #include "renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" #include "servers/rendering/rendering_server_default.h" //////////////////////////////////////////////////////////////////////////////// @@ -50,6 +52,7 @@ void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { ShaderCompilerRD::GeneratedCode gen_code; ShaderCompilerRD::IdentifierActions actions; + actions.entry_point_stages["sky"] = ShaderCompilerRD::STAGE_FRAGMENT; uses_time = false; uses_half_res = false; @@ -89,8 +92,7 @@ void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; Error err = scene_singleton->sky.sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = scene_singleton->sky.sky_shader.shader.version_create(); @@ -110,7 +112,7 @@ void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { print_line("\n**light_code:\n" + gen_code.light); #endif - scene_singleton->sky.sky_shader.shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.fragment, gen_code.defines); + scene_singleton->sky.sky_shader.shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompilerRD::STAGE_VERTEX], gen_code.stage_globals[ShaderCompilerRD::STAGE_FRAGMENT], gen_code.defines); ERR_FAIL_COND(!scene_singleton->sky.sky_shader.shader.version_is_valid(version)); ubo_size = gen_code.uniform_total_size; @@ -124,8 +126,12 @@ void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { depth_stencil_state.enable_depth_test = true; depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; - RID shader_variant = scene_singleton->sky.sky_shader.shader.version_get_shader(version, i); - pipelines[i].setup(shader_variant, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), depth_stencil_state, RD::PipelineColorBlendState::create_disabled(), 0); + if (scene_singleton->sky.sky_shader.shader.is_variant_enabled(i)) { + RID shader_variant = scene_singleton->sky.sky_shader.shader.version_get_shader(version, i); + pipelines[i].setup(shader_variant, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), depth_stencil_state, RD::PipelineColorBlendState::create_disabled(), 0); + } else { + pipelines[i].clear(); + } } valid = true; @@ -223,96 +229,76 @@ RendererSceneSkyRD::SkyShaderData::~SkyShaderData() { //////////////////////////////////////////////////////////////////////////////// // Sky material -void RendererSceneSkyRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererSceneSkyRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; uniform_set_updated = true; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); +} - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } +RendererSceneSkyRD::SkyMaterialData::~SkyMaterialData() { + free_parameters_uniform_set(uniform_set); +} - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); +//////////////////////////////////////////////////////////////////////////////// +// Render sky + +static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_basis, float *p_array) { + p_array[0] = p_basis.elements[0][0]; + p_array[1] = p_basis.elements[1][0]; + p_array[2] = p_basis.elements[2][0]; + p_array[3] = 0; + p_array[4] = p_basis.elements[0][1]; + p_array[5] = p_basis.elements[1][1]; + p_array[6] = p_basis.elements[2][1]; + p_array[7] = 0; + p_array[8] = p_basis.elements[0][2]; + p_array[9] = p_basis.elements[1][2]; + p_array[10] = p_basis.elements[2][2]; + p_array[11] = 0; +} - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; +void RendererSceneSkyRD::_render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position) { + SkyPushConstant sky_push_constant; - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } + memset(&sky_push_constant, 0, sizeof(SkyPushConstant)); - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); + for (uint32_t v = 0; v < p_view_count; v++) { + // We only need key components of our projection matrix + sky_push_constant.projections[v][0] = p_projections[v].matrix[2][0]; + sky_push_constant.projections[v][1] = p_projections[v].matrix[0][0]; + sky_push_constant.projections[v][2] = p_projections[v].matrix[2][1]; + sky_push_constant.projections[v][3] = p_projections[v].matrix[1][1]; } + sky_push_constant.position[0] = p_position.x; + sky_push_constant.position[1] = p_position.y; + sky_push_constant.position[2] = p_position.z; + sky_push_constant.multiplier = p_multiplier; + sky_push_constant.time = p_time; + store_transform_3x3(p_orientation, sky_push_constant.orientation); - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } + RenderingDevice::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_fb); - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } + RD::DrawListID draw_list = p_list; - Vector<RD::Uniform> uniforms; + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format, false, RD::get_singleton()->draw_list_get_current_pass())); + // Update uniform sets. { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.uniform_set, 0); + if (RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_uniform_set, 1); } + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_texture_set, 2); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.fog_uniform_set, 3); } - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); -} + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); -RendererSceneSkyRD::SkyMaterialData::~SkyMaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } + RD::get_singleton()->draw_list_set_push_constant(draw_list, &sky_push_constant, sizeof(SkyPushConstant)); - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + RD::get_singleton()->draw_list_draw(draw_list, true); } //////////////////////////////////////////////////////////////////////////////// @@ -329,12 +315,16 @@ void RendererSceneSkyRD::ReflectionData::clear_reflection_data() { coefficient_buffer = RID(); } -void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers) { +void RendererSceneSkyRD::ReflectionData::update_reflection_data(RendererStorageRD *p_storage, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format) { //recreate radiance and all data int mipmaps = p_mipmaps; uint32_t w = p_size, h = p_size; + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + if (p_use_array) { int num_layers = p_low_quality ? 8 : p_roughness_layers; @@ -393,9 +383,9 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int } radiance_base_cubemap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, 0, RD::TEXTURE_SLICE_CUBEMAP); - + RD::get_singleton()->set_resource_name(radiance_base_cubemap, "radiance base cubemap"); RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = p_texture_format; tf.width = 64; // Always 64x64 tf.height = 64; tf.texture_type = RD::TEXTURE_TYPE_CUBE; @@ -404,6 +394,7 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(downsampled_radiance_cubemap, "downsampled radiance cubemap"); { uint32_t mmw = 64; uint32_t mmh = 64; @@ -413,6 +404,18 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int mm.size.width = mmw; mm.size.height = mmh; mm.view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, 0, j, RD::TEXTURE_SLICE_CUBEMAP); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip " + itos(j) + " "); + if (prefer_raster_effects) { + // we need a framebuffer for each side of our cubemap + + for (int k = 0; k < 6; k++) { + mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, k, j); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip: " + itos(j) + " Face: " + itos(k) + " "); + Vector<RID> fbtex; + fbtex.push_back(mm.views[k]); + mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex); + } + } mmw = MAX(1, mmw >> 1); mmh = MAX(1, mmh >> 1); @@ -421,50 +424,128 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int } void RendererSceneSkyRD::ReflectionData::create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays) { - p_storage->get_effects()->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + for (int k = 0; k < 6; k++) { + effects->cubemap_downsample_raster(radiance_base_cubemap, downsampled_layer.mipmaps[0].framebuffers[k], k, downsampled_layer.mipmaps[0].size); + } - for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { - p_storage->get_effects()->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); - } + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_downsample_raster(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].framebuffers[k], k, downsampled_layer.mipmaps[i].size); + } + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance - Vector<RID> views; - if (p_use_arrays) { - for (int i = 1; i < layers.size(); i++) { - views.push_back(layers[i].views[0]); + if (p_use_arrays) { + RD::get_singleton()->draw_command_begin_label("filter radiance map into array heads"); + for (int i = 0; i < layers.size(); i++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[i].mipmaps[0].framebuffers[k], k, i); + } + } + } else { + RD::get_singleton()->draw_command_begin_label("filter radiance map into mipmaps directly"); + for (int j = 0; j < layers[0].mipmaps.size(); j++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[0].mipmaps[j].framebuffers[k], k, j); + } + } } + RD::get_singleton()->draw_command_end_label(); // Filter radiance } else { - for (int i = 1; i < layers[0].views.size(); i++) { - views.push_back(layers[0].views[i]); + effects->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + effects->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); + } + + Vector<RID> views; + if (p_use_arrays) { + for (int i = 1; i < layers.size(); i++) { + views.push_back(layers[i].views[0]); + } + } else { + for (int i = 1; i < layers[0].views.size(); i++) { + views.push_back(layers[0].views[i]); + } } - } - p_storage->get_effects()->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); + effects->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); + } } void RendererSceneSkyRD::ReflectionData::create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality) { - if (p_use_arrays) { - //render directly to the layers - p_storage->get_effects()->cubemap_roughness(radiance_base_cubemap, layers[p_base_layer].views[0], p_cube_side, p_sky_ggx_samples_quality, float(p_base_layer) / (layers.size() - 1.0), layers[p_base_layer].mipmaps[0].size.x); + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + // Need to ask clayjohn but p_cube_side is set to 10, looks like in the compute shader we're doing all 6 sides in one call + // here we need to do them one by one so ignoring p_cube_side + if (p_use_arrays) { + for (int k = 0; k < 6; k++) { + effects->cubemap_roughness_raster( + radiance_base_cubemap, + layers[p_base_layer].mipmaps[0].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers.size() - 1.0), + layers[p_base_layer].mipmaps[0].size.x); + } + } else { + for (int k = 0; k < 6; k++) { + effects->cubemap_roughness_raster( + layers[0].views[p_base_layer - 1], + layers[0].mipmaps[p_base_layer].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } + } } else { - p_storage->get_effects()->cubemap_roughness( - layers[0].views[p_base_layer - 1], - layers[0].views[p_base_layer], - p_cube_side, - p_sky_ggx_samples_quality, - float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), - layers[0].mipmaps[p_base_layer].size.x); + if (p_use_arrays) { + //render directly to the layers + effects->cubemap_roughness(radiance_base_cubemap, layers[p_base_layer].views[0], p_cube_side, p_sky_ggx_samples_quality, float(p_base_layer) / (layers.size() - 1.0), layers[p_base_layer].mipmaps[0].size.x); + } else { + effects->cubemap_roughness( + layers[0].views[p_base_layer - 1], + layers[0].views[p_base_layer], + p_cube_side, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } } } void RendererSceneSkyRD::ReflectionData::update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end) { + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + RD::get_singleton()->draw_command_begin_label("Update Radiance Cubemap Array Mipmaps"); for (int i = p_start; i < p_end; i++) { for (int j = 0; j < layers[i].views.size() - 1; j++) { RID view = layers[i].views[j]; - RID texture = layers[i].views[j + 1]; Size2i size = layers[i].mipmaps[j + 1].size; - p_storage->get_effects()->cubemap_downsample(view, texture, size); + if (prefer_raster_effects) { + for (int k = 0; k < 6; k++) { + RID framebuffer = layers[i].mipmaps[j + 1].framebuffers[k]; + effects->cubemap_downsample_raster(view, framebuffer, k, size); + } + } else { + RID texture = layers[i].views[j + 1]; + effects->cubemap_downsample(view, texture, size); + } } } + RD::get_singleton()->draw_command_end_label(); } //////////////////////////////////////////////////////////////////////////////// @@ -621,7 +702,7 @@ Ref<Image> RendererSceneSkyRD::Sky::bake_panorama(RendererStorageRD *p_storage, RD::get_singleton()->free(rad_tex); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF, data); for (int i = 0; i < p_size.width; i++) { for (int j = 0; j < p_size.height; j++) { @@ -692,7 +773,18 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n"); // Cubemap sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_HALF_RES_PASS\n"); // Half Res Cubemap sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_QUARTER_RES_PASS\n"); // Quarter res Cubemap + + sky_modes.push_back("\n#define USE_MULTIVIEW\n"); // Full size multiview + sky_modes.push_back("\n#define USE_HALF_RES_PASS\n#define USE_MULTIVIEW\n"); // Half Res multiview + sky_modes.push_back("\n#define USE_QUARTER_RES_PASS\n#define USE_MULTIVIEW\n"); // Quarter res multiview + sky_shader.shader.initialize(sky_modes, defines); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + sky_shader.shader.set_variant_enabled(SKY_VERSION_BACKGROUND_MULTIVIEW, false); + sky_shader.shader.set_variant_enabled(SKY_VERSION_HALF_RES_MULTIVIEW, false); + sky_shader.shader.set_variant_enabled(SKY_VERSION_QUARTER_RES_MULTIVIEW, false); + } } // register our shader funds @@ -709,6 +801,9 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { actions.renames["SKY_COORDS"] = "panorama_coords"; actions.renames["SCREEN_UV"] = "uv"; actions.renames["TIME"] = "params.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); actions.renames["HALF_RES_COLOR"] = "half_res_color"; actions.renames["QUARTER_RES_COLOR"] = "quarter_res_color"; actions.renames["RADIANCE"] = "radiance"; @@ -759,7 +854,13 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_shader.default_shader = storage->shader_allocate(); storage->shader_initialize(sky_shader.default_shader); - storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void fragment() { COLOR = vec3(0.0); } \n"); + storage->shader_set_code(sky_shader.default_shader, R"( +shader_type sky; + +void sky() { + COLOR = vec3(0.0); +} +)"); sky_shader.default_material = storage->material_allocate(); storage->material_initialize(sky_shader.default_material); @@ -840,7 +941,15 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_scene_state.fog_shader = storage->shader_allocate(); storage->shader_initialize(sky_scene_state.fog_shader); - storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void fragment() { COLOR = clear_color.rgb; } \n"); + storage->shader_set_code(sky_scene_state.fog_shader, R"( +shader_type sky; + +uniform vec4 clear_color; + +void sky() { + COLOR = clear_color.rgb; +} +)"); sky_scene_state.fog_material = storage->material_allocate(); storage->material_initialize(sky_scene_state.fog_material); @@ -871,9 +980,48 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_scene_state.fog_only_texture_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES); } + + { //create index array for copy shaders + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + index_array = RD::get_singleton()->index_array_create(index_buffer, 0, 6); + } +} + +void RendererSceneSkyRD::set_texture_format(RD::DataFormat p_texture_format) { + texture_format = p_texture_format; +} + +RendererSceneSkyRD::~RendererSceneSkyRD() { + // TODO cleanup anything created in init... + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.uniform_set)) { + RD::get_singleton()->free(sky_scene_state.uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.default_fog_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.default_fog_uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.fog_only_texture_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.fog_only_texture_uniform_set); + } + + RD::get_singleton()->free(index_buffer); //array gets freed as dependency } -void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render) { +void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render) { ERR_FAIL_COND(!p_env); // I guess without an environment we also can't have a sky... SkyMaterialData *material = nullptr; @@ -1043,7 +1191,7 @@ void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_b RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); } -void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform &p_transform, double p_time) { +void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time) { ERR_FAIL_COND(!p_env); Sky *sky = get_sky(p_env->sky); @@ -1123,6 +1271,7 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM cm = correction * cm; if (shader_data->uses_quarter_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Quarter Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_QUARTER_RES]; Vector<Color> clear_colors; @@ -1130,17 +1279,18 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[2].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); } if (shader_data->uses_half_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Half Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_HALF_RES]; Vector<Color> clear_colors; @@ -1148,28 +1298,29 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP_HALF_RES, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[1].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); } RD::DrawListID cubemap_draw_list; PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP]; + RD::get_singleton()->draw_command_begin_label("Render Sky Cubemap"); for (int i = 0; i < 6; i++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[0].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); if (sky_mode == RS::SKY_MODE_REALTIME) { sky->reflection.create_reflection_fast_filter(storage, sky_use_cubemap_array); @@ -1208,9 +1359,12 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM } } -void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, const CameraMatrix &p_projection, const Transform &p_transform, double p_time) { +void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time) { ERR_FAIL_COND(!p_env); + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + Sky *sky = get_sky(p_env->sky); ERR_FAIL_COND(!sky); @@ -1252,24 +1406,28 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont float multiplier = p_env->bg_energy; float custom_fov = p_env->sky_custom_fov; + // Camera CameraMatrix camera; + uint32_t view_count = p_view_count; + const CameraMatrix *projections = p_projections; if (custom_fov) { - float near_plane = p_projection.get_z_near(); - float far_plane = p_projection.get_z_far(); - float aspect = p_projection.get_aspect(); + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); camera.set_perspective(custom_fov, aspect, near_plane, far_plane); - } else { - camera = p_projection; + view_count = 1; + projections = &camera; } sky_transform = p_transform.basis * sky_transform; if (shader_data->uses_quarter_res) { - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_QUARTER_RES]; + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_QUARTER_RES_MULTIVIEW : SKY_VERSION_QUARTER_RES]; RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_QUARTER_RES, sky_shader.default_shader_rd); @@ -1277,12 +1435,12 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont clear_colors.push_back(Color(0.0, 0.0, 0.0)); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); - storage->get_effects()->render_sky(draw_list, p_time, sky->quarter_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } if (shader_data->uses_half_res) { - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_HALF_RES]; + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_HALF_RES_MULTIVIEW : SKY_VERSION_HALF_RES]; RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_HALF_RES, sky_shader.default_shader_rd); @@ -1290,11 +1448,11 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont clear_colors.push_back(Color(0.0, 0.0, 0.0)); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); - storage->get_effects()->render_sky(draw_list, p_time, sky->half_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_BACKGROUND]; + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_BACKGROUND_MULTIVIEW : SKY_VERSION_BACKGROUND]; RID texture_uniform_set; if (sky) { @@ -1304,10 +1462,183 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont } RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb, RD::INITIAL_ACTION_CONTINUE, p_can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, p_can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); - storage->get_effects()->render_sky(draw_list, p_time, p_fb, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); RD::get_singleton()->draw_list_end(); } +void RendererSceneSkyRD::update_res_buffers(RendererSceneEnvironmentRD *p_env, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time) { + ERR_FAIL_COND(!p_env); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = p_env->sky_orientation; + sky_transform.invert(); + + float multiplier = p_env->bg_energy; + float custom_fov = p_env->sky_custom_fov; + + // Camera + CameraMatrix camera; + uint32_t view_count = p_view_count; + const CameraMatrix *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + if (shader_data->uses_quarter_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_QUARTER_RES_MULTIVIEW : SKY_VERSION_QUARTER_RES]; + + RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_QUARTER_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + + if (shader_data->uses_half_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_HALF_RES_MULTIVIEW : SKY_VERSION_HALF_RES]; + + RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_HALF_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } +} + +void RendererSceneSkyRD::draw(RD::DrawListID p_draw_list, RendererSceneEnvironmentRD *p_env, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time) { + ERR_FAIL_COND(!p_env); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = p_env->sky_orientation; + sky_transform.invert(); + + float multiplier = p_env->bg_energy; + float custom_fov = p_env->sky_custom_fov; + + // Camera + CameraMatrix camera; + uint32_t view_count = p_view_count; + const CameraMatrix *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_BACKGROUND_MULTIVIEW : SKY_VERSION_BACKGROUND]; + + RID texture_uniform_set; + if (sky) { + texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_BACKGROUND, sky_shader.default_shader_rd); + } else { + texture_uniform_set = sky_scene_state.fog_only_texture_uniform_set; + } + + _render_sky(p_draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); +} + void RendererSceneSkyRD::invalidate_sky(Sky *p_sky) { if (!p_sky->dirty) { p_sky->dirty = true; @@ -1339,7 +1670,7 @@ void RendererSceneSkyRD::update_dirty_skys() { //array (higher quality, 6 times more memory) RD::TextureFormat tf; tf.array_layers = layers * 6; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = texture_format; tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; tf.mipmaps = mipmaps; tf.width = w; @@ -1348,13 +1679,13 @@ void RendererSceneSkyRD::update_dirty_skys() { sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); - sky->reflection.update_reflection_data(sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers); + sky->reflection.update_reflection_data(storage, sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); } else { //regular cubemap, lower quality (aliasing, less memory) RD::TextureFormat tf; tf.array_layers = 6; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = texture_format; tf.texture_type = RD::TEXTURE_TYPE_CUBE; tf.mipmaps = MIN(mipmaps, layers); tf.width = w; @@ -1363,7 +1694,7 @@ void RendererSceneSkyRD::update_dirty_skys() { sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); - sky->reflection.update_reflection_data(sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers); + sky->reflection.update_reflection_data(storage, sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); } texture_set_dirty = true; } @@ -1371,7 +1702,7 @@ void RendererSceneSkyRD::update_dirty_skys() { // Create subpass buffers if they haven't been created already if (sky->half_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->half_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { RD::TextureFormat tformat; - tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tformat.format = texture_format; tformat.width = sky->screen_size.x / 2; tformat.height = sky->screen_size.y / 2; tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; @@ -1386,7 +1717,7 @@ void RendererSceneSkyRD::update_dirty_skys() { if (sky->quarter_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->quarter_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { RD::TextureFormat tformat; - tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tformat.format = texture_format; tformat.width = sky->screen_size.x / 4; tformat.height = sky->screen_size.y / 4; tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.h b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h index 73390a586b..7b670bddd5 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h @@ -43,9 +43,6 @@ class RendererSceneRenderRD; class RendererSceneSkyRD { -private: - RendererStorageRD *storage; - public: enum SkySet { SKY_SET_UNIFORMS, @@ -55,6 +52,23 @@ public: SKY_SET_MAX }; + // Skys need less info from Directional Lights than the normal shaders + struct SkyDirectionalLightData { + float direction[3]; + float energy; + float color[3]; + float size; + uint32_t enabled; + uint32_t pad[3]; + }; + +private: + RendererStorageRD *storage; + RD::DataFormat texture_format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + + RID index_buffer; + RID index_array; + enum SkyTextureSetVersion { SKY_TEXTURE_SET_BACKGROUND, SKY_TEXTURE_SET_HALF_RES, @@ -72,19 +86,61 @@ public: SKY_VERSION_CUBEMAP, SKY_VERSION_CUBEMAP_HALF_RES, SKY_VERSION_CUBEMAP_QUARTER_RES, + + SKY_VERSION_BACKGROUND_MULTIVIEW, + SKY_VERSION_HALF_RES_MULTIVIEW, + SKY_VERSION_QUARTER_RES_MULTIVIEW, + SKY_VERSION_MAX }; - // Skys need less info from Directional Lights than the normal shaders - struct SkyDirectionalLightData { - float direction[3]; - float energy; - float color[3]; - float size; - uint32_t enabled; - uint32_t pad[3]; + struct SkyPushConstant { + float orientation[12]; // 48 - 48 + float projections[RendererSceneRender::MAX_RENDER_VIEWS][4]; // 2 x 16 - 80 + float position[3]; // 12 - 92 + float multiplier; // 4 - 96 + float time; // 4 - 100 + float pad[3]; // 12 - 112 // Using pad to align on 16 bytes + // 128 is the max size of a push constant. We can replace "pad" but we can't add any more. }; + struct SkyShaderData : public RendererStorageRD::ShaderData { + bool valid; + RID version; + + PipelineCacheRD pipelines[SKY_VERSION_MAX]; + Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size; + + String path; + String code; + Map<StringName, RID> default_texture_params; + + bool uses_time; + bool uses_position; + bool uses_half_res; + bool uses_quarter_res; + bool uses_light; + + virtual void set_code(const String &p_Code); + virtual void set_default_texture_param(const StringName &p_name, RID p_texture); + virtual void get_param_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_param_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + SkyShaderData(); + virtual ~SkyShaderData(); + }; + + void _render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position); + +public: struct SkySceneState { struct UBO { uint32_t volumetric_fog_enabled; @@ -135,6 +191,10 @@ public: struct Mipmap { RID view; Size2i size; + + // for mobile only + RID views[6]; + RID framebuffers[6]; }; Vector<Mipmap> mipmaps; }; @@ -149,46 +209,12 @@ public: Vector<Layer> layers; void clear_reflection_data(); - void update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers); + void update_reflection_data(RendererStorageRD *p_storage, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format); void create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays); void create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality); void update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end); }; - struct SkyShaderData : public RendererStorageRD::ShaderData { - bool valid; - RID version; - - PipelineCacheRD pipelines[SKY_VERSION_MAX]; - Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; - Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; - - Vector<uint32_t> ubo_offsets; - uint32_t ubo_size; - - String path; - String code; - Map<StringName, RID> default_texture_params; - - bool uses_time; - bool uses_position; - bool uses_half_res; - bool uses_quarter_res; - bool uses_light; - - virtual void set_code(const String &p_Code); - virtual void set_default_texture_param(const StringName &p_name, RID p_texture); - virtual void get_param_list(List<PropertyInfo> *p_param_list) const; - virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; - virtual bool is_param_texture(const StringName &p_param) const; - virtual bool is_animated() const; - virtual bool casts_shadows() const; - virtual Variant get_default_parameter(const StringName &p_parameter) const; - virtual RS::ShaderNativeSourceCode get_native_source_code() const; - SkyShaderData(); - virtual ~SkyShaderData(); - }; - /* Sky shader */ struct SkyShader { @@ -203,15 +229,12 @@ public: struct SkyMaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; SkyShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; bool uniform_set_updated; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~SkyMaterialData(); }; @@ -265,12 +288,15 @@ public: static RendererStorageRD::MaterialData *_create_sky_material_funcs(RendererStorageRD::ShaderData *p_shader); RendererSceneSkyRD(); - void init(RendererStorageRD *p_storage); - - void setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render); - void update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform &p_transform, double p_time); - void draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, const CameraMatrix &p_projection, const Transform &p_transform, double p_time); + void set_texture_format(RD::DataFormat p_texture_format); + ~RendererSceneSkyRD(); + + void setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render); + void update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time); + void draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time); + void update_res_buffers(RendererSceneEnvironmentRD *p_env, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time); + void draw(RD::DrawListID p_draw_list, RendererSceneEnvironmentRD *p_env, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time); void invalidate_sky(Sky *p_sky); void update_dirty_skys(); diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index b984f850a0..8cc20618fc 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -33,7 +33,9 @@ #include "core/config/engine.h" #include "core/config/project_settings.h" #include "core/io/resource_loader.h" +#include "core/math/math_defs.h" #include "renderer_compositor_rd.h" +#include "servers/rendering/rendering_server_globals.h" #include "servers/rendering/shader_language.h" bool RendererStorageRD::can_create_resources_async() const { @@ -756,7 +758,7 @@ void RendererStorageRD::texture_3d_initialize(RID p_texture, Image::Format p_for for (int i = 0; i < p_data.size(); i++) { uint32_t s = images[i]->get_data().size(); - copymem(&all_data.write[offset], images[i]->get_data().ptr(), s); + memcpy(&all_data.write[offset], images[i]->get_data().ptr(), s); { Texture::BufferSlice3D slice; slice.size.width = images[i]->get_width(); @@ -882,10 +884,6 @@ void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_im RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); } -void RendererStorageRD::texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer) { - _texture_2d_update(p_texture, p_image, p_layer, true); -} - void RendererStorageRD::texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer) { _texture_2d_update(p_texture, p_image, p_layer, false); } @@ -919,7 +917,7 @@ void RendererStorageRD::texture_3d_update(RID p_texture, const Vector<Ref<Image> for (int i = 0; i < p_data.size(); i++) { uint32_t s = images[i]->get_data().size(); - copymem(&all_data.write[offset], images[i]->get_data().ptr(), s); + memcpy(&all_data.write[offset], images[i]->get_data().ptr(), s); offset += s; } } @@ -971,7 +969,7 @@ void RendererStorageRD::texture_2d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -987,7 +985,7 @@ void RendererStorageRD::texture_2d_layered_placeholder_initialize(RID p_texture, //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -1013,7 +1011,7 @@ void RendererStorageRD::texture_3d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -1043,7 +1041,7 @@ Ref<Image> RendererStorageRD::texture_2d_get(RID p_texture) const { Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, 0); ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); Ref<Image> image; - image.instance(); + image.instantiate(); image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); if (tex->format != tex->validated_format) { @@ -1066,7 +1064,7 @@ Ref<Image> RendererStorageRD::texture_2d_layer_get(RID p_texture, int p_layer) c Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, p_layer); ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); Ref<Image> image; - image.instance(); + image.instantiate(); image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); if (tex->format != tex->validated_format) { @@ -1094,7 +1092,7 @@ Vector<Ref<Image>> RendererStorageRD::texture_3d_get(RID p_texture) const { Vector<uint8_t> sub_region = all_data.subarray(bs.offset, bs.offset + bs.buffer_size - 1); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(bs.size.width, bs.size.height, false, tex->validated_format, sub_region); ERR_FAIL_COND_V(img->is_empty(), Vector<Ref<Image>>()); if (tex->format != tex->validated_format) { @@ -1233,7 +1231,7 @@ RID RendererStorageRD::canvas_texture_allocate() { return canvas_texture_owner.allocate_rid(); } void RendererStorageRD::canvas_texture_initialize(RID p_rid) { - canvas_texture_owner.initialize_rid(p_rid, memnew(CanvasTexture)); + canvas_texture_owner.initialize_rid(p_rid); } void RendererStorageRD::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { @@ -1439,8 +1437,10 @@ void RendererStorageRD::shader_set_code(RID p_shader, const String &p_code) { material->shader_type = new_type; } - for (Map<StringName, RID>::Element *E = shader->default_texture_parameter.front(); E; E = E->next()) { - shader->data->set_default_texture_param(E->key(), E->get()); + if (shader->data) { + for (Map<StringName, RID>::Element *E = shader->default_texture_parameter.front(); E; E = E->next()) { + shader->data->set_default_texture_param(E->key(), E->get()); + } } } @@ -1526,27 +1526,18 @@ RID RendererStorageRD::material_allocate() { return material_owner.allocate_rid(); } void RendererStorageRD::material_initialize(RID p_rid) { - Material material; - material.data = nullptr; - material.shader = nullptr; - material.shader_type = SHADER_TYPE_MAX; - material.update_next = nullptr; - material.update_requested = false; - material.uniform_dirty = false; - material.texture_dirty = false; - material.priority = 0; - material.self = p_rid; - material_owner.initialize_rid(p_rid, material); + material_owner.initialize_rid(p_rid); + Material *material = material_owner.getornull(p_rid); + material->self = p_rid; } void RendererStorageRD::_material_queue_update(Material *material, bool p_uniform, bool p_texture) { - if (material->update_requested) { + if (material->update_element.in_list()) { return; } - material->update_next = material_update_list; - material_update_list = material; - material->update_requested = true; + material_update_list.add(&material->update_element); + material->uniform_dirty = material->uniform_dirty || p_uniform; material->texture_dirty = material->texture_dirty || p_texture; } @@ -1601,6 +1592,7 @@ void RendererStorageRD::material_set_param(RID p_material, const StringName &p_p if (p_value.get_type() == Variant::NIL) { material->params.erase(p_param); } else { + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); //object not allowed material->params[p_param] = p_value; } @@ -1877,8 +1869,8 @@ _FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataTy gui[1] = v.position.y; gui[2] = v.size.x; gui[3] = v.size.y; - } else if (value.get_type() == Variant::QUAT) { - Quat v = value; + } else if (value.get_type() == Variant::QUATERNION) { + Quaternion v = value; gui[0] = v.x; gui[1] = v.y; @@ -1925,7 +1917,7 @@ _FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataTy gui[11] = 0; } break; case ShaderLanguage::TYPE_MAT4: { - Transform v = value; + Transform3D v = value; float *gui = (float *)data; gui[0] = v.basis.elements[0][0]; @@ -2108,13 +2100,13 @@ _FORCE_INLINE_ static void _fill_std140_ubo_empty(ShaderLanguage::DataType type, case ShaderLanguage::TYPE_INT: case ShaderLanguage::TYPE_UINT: case ShaderLanguage::TYPE_FLOAT: { - zeromem(data, 4); + memset(data, 0, 4); } break; case ShaderLanguage::TYPE_BVEC2: case ShaderLanguage::TYPE_IVEC2: case ShaderLanguage::TYPE_UVEC2: case ShaderLanguage::TYPE_VEC2: { - zeromem(data, 8); + memset(data, 0, 8); } break; case ShaderLanguage::TYPE_BVEC3: case ShaderLanguage::TYPE_IVEC3: @@ -2124,16 +2116,16 @@ _FORCE_INLINE_ static void _fill_std140_ubo_empty(ShaderLanguage::DataType type, case ShaderLanguage::TYPE_IVEC4: case ShaderLanguage::TYPE_UVEC4: case ShaderLanguage::TYPE_VEC4: { - zeromem(data, 16); + memset(data, 0, 16); } break; case ShaderLanguage::TYPE_MAT2: { - zeromem(data, 32); + memset(data, 0, 32); } break; case ShaderLanguage::TYPE_MAT3: { - zeromem(data, 48); + memset(data, 0, 48); } break; case ShaderLanguage::TYPE_MAT4: { - zeromem(data, 64); + memset(data, 0, 64); } break; default: { @@ -2232,6 +2224,10 @@ RendererStorageRD::MaterialData::~MaterialData() { //unregister material from those using global textures rs->global_variables.materials_using_texture.erase(global_texture_E); } + + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + } } void RendererStorageRD::MaterialData::update_textures(const Map<StringName, Variant> &p_parameters, const Map<StringName, RID> &p_default_textures, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, RID *p_textures, bool p_use_linear_color) { @@ -2381,6 +2377,105 @@ void RendererStorageRD::MaterialData::update_textures(const Map<StringName, Vari } } +void RendererStorageRD::MaterialData::free_parameters_uniform_set(RID p_uniform_set) { + if (p_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(p_uniform_set, nullptr, nullptr); + RD::get_singleton()->free(p_uniform_set); + } +} + +bool RendererStorageRD::MaterialData::update_parameters_uniform_set(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const Map<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, const Map<StringName, RID> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, uint32_t p_barrier) { + if ((uint32_t)ubo_data.size() != p_ubo_size) { + p_uniform_dirty = true; + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + uniform_buffer = RID(); + } + + ubo_data.resize(p_ubo_size); + if (ubo_data.size()) { + uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); + memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear + } + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + //check whether buffer changed + if (p_uniform_dirty && ubo_data.size()) { + update_uniform_buffer(p_uniforms, p_uniform_offsets, p_parameters, ubo_data.ptrw(), ubo_data.size(), false); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), p_barrier); + } + + uint32_t tex_uniform_count = p_texture_uniforms.size(); + + if ((uint32_t)texture_cache.size() != tex_uniform_count || p_textures_dirty) { + texture_cache.resize(tex_uniform_count); + p_textures_dirty = true; + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + if (p_textures_dirty && tex_uniform_count) { + update_textures(p_parameters, p_default_texture_params, p_texture_uniforms, texture_cache.ptrw(), true); + } + + if (p_ubo_size == 0 && p_texture_uniforms.size() == 0) { + // This material does not require an uniform set, so don't create it. + return false; + } + + if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + //no reason to update uniform set, only UBO (or nothing) was needed to update + return false; + } + + Vector<RD::Uniform> uniforms; + + { + if (p_ubo_size) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 0; + u.ids.push_back(uniform_buffer); + uniforms.push_back(u); + } + + const RID *textures = texture_cache.ptrw(); + for (uint32_t i = 0; i < tex_uniform_count; i++) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1 + i; + u.ids.push_back(textures[i]); + uniforms.push_back(u); + } + } + + uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_shader_uniform_set); + + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, _material_uniform_set_erased, &self); + + return true; +} + +void RendererStorageRD::_material_uniform_set_erased(const RID &p_set, void *p_material) { + RID rid = *(RID *)p_material; + Material *material = base_singleton->material_owner.getornull(rid); + if (material) { + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); + } +} + void RendererStorageRD::material_force_update_textures(RID p_material, ShaderType p_shader_type) { Material *material = material_owner.getornull(p_material); if (material->shader_type != p_shader_type) { @@ -2392,20 +2487,23 @@ void RendererStorageRD::material_force_update_textures(RID p_material, ShaderTyp } void RendererStorageRD::_update_queued_materials() { - Material *material = material_update_list; - while (material) { - Material *next = material->update_next; + while (material_update_list.first()) { + Material *material = material_update_list.first()->self(); + bool uniforms_changed = false; if (material->data) { - material->data->update_parameters(material->params, material->uniform_dirty, material->texture_dirty); + uniforms_changed = material->data->update_parameters(material->params, material->uniform_dirty, material->texture_dirty); } - material->update_requested = false; material->texture_dirty = false; material->uniform_dirty = false; - material->update_next = nullptr; - material = next; + + material_update_list.remove(&material->update_element); + + if (uniforms_changed) { + //some implementations such as 3D renderer cache the matreial uniform set, so update is required + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); + } } - material_update_list = nullptr; } /* MESH API */ @@ -2433,6 +2531,8 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); + ERR_FAIL_COND(mesh->surface_count == RS::MAX_MESH_SURFACES); + #ifdef DEBUG_ENABLED //do a validation, to catch errors first { @@ -2460,7 +2560,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su } break; case RS::ARRAY_COLOR: { - attrib_stride += sizeof(int16_t) * 4; + attrib_stride += sizeof(uint32_t); } break; case RS::ARRAY_TEX_UV: { attrib_stride += sizeof(float) * 2; @@ -2548,6 +2648,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su s->lods[i].index_buffer = RD::get_singleton()->index_buffer_create(indices, is_index_16 ? RD::INDEX_BUFFER_FORMAT_UINT16 : RD::INDEX_BUFFER_FORMAT_UINT32, p_surface.lods[i].index_data); s->lods[i].index_array = RD::get_singleton()->index_array_create(s->lods[i].index_buffer, 0, indices); s->lods[i].edge_length = p_surface.lods[i].edge_length; + s->lods[i].index_count = indices; } } } @@ -2615,9 +2716,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su mesh->surfaces[mesh->surface_count] = s; mesh->surface_count++; - for (List<MeshInstance *>::Element *E = mesh->instances.front(); E; E = E->next()) { - //update instances - MeshInstance *mi = E->get(); + for (MeshInstance *mi : mesh->instances) { _mesh_instance_add_surface(mi, mesh, mesh->surface_count - 1); } @@ -2652,7 +2751,7 @@ RS::BlendShapeMode RendererStorageRD::mesh_get_blend_shape_mode(RID p_mesh) cons return mesh->blend_shape_mode; } -void RendererStorageRD::mesh_surface_update_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { +void RendererStorageRD::mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); @@ -2663,6 +2762,30 @@ void RendererStorageRD::mesh_surface_update_region(RID p_mesh, int p_surface, in RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->vertex_buffer, p_offset, data_size, r); } +void RendererStorageRD::mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->attribute_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->attribute_buffer, p_offset, data_size, r); +} + +void RendererStorageRD::mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->skin_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->skin_buffer, p_offset, data_size, r); +} + void RendererStorageRD::mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); @@ -2775,7 +2898,7 @@ AABB RendererStorageRD::mesh_get_aabb(RID p_mesh, RID p_skeleton) { const float *dataptr = baseptr + j * 8; - Transform mtx; + Transform3D mtx; mtx.basis.elements[0].x = dataptr[0]; mtx.basis.elements[1].x = dataptr[1]; @@ -2802,7 +2925,7 @@ AABB RendererStorageRD::mesh_get_aabb(RID p_mesh, RID p_skeleton) { const float *dataptr = baseptr + j * 12; - Transform mtx; + Transform3D mtx; mtx.basis.elements[0][0] = dataptr[0]; mtx.basis.elements[0][1] = dataptr[1]; @@ -2904,8 +3027,7 @@ void RendererStorageRD::mesh_clear(RID p_mesh) { mesh->surface_count = 0; mesh->material_cache.clear(); //clear instance data - for (List<MeshInstance *>::Element *E = mesh->instances.front(); E; E = E->next()) { - MeshInstance *mi = E->get(); + for (MeshInstance *mi : mesh->instances) { _mesh_instance_clear(mi); } mesh->has_bone_weights = false; @@ -2931,7 +3053,8 @@ RID RendererStorageRD::mesh_instance_create(RID p_base) { Mesh *mesh = mesh_owner.getornull(p_base); ERR_FAIL_COND_V(!mesh, RID()); - MeshInstance *mi = memnew(MeshInstance); + RID rid = mesh_instance_owner.make_rid(); + MeshInstance *mi = mesh_instance_owner.getornull(rid); mi->mesh = mesh; @@ -2943,7 +3066,7 @@ RID RendererStorageRD::mesh_instance_create(RID p_base) { mi->dirty = true; - return mesh_instance_owner.make_rid(mi); + return rid; } void RendererStorageRD::mesh_instance_set_skeleton(RID p_mesh_instance, RID p_skeleton) { MeshInstance *mi = mesh_instance_owner.getornull(p_mesh_instance); @@ -3230,8 +3353,8 @@ void RendererStorageRD::_mesh_surface_generate_version_for_input_mask(Mesh::Surf case RS::ARRAY_COLOR: { vd.offset = attribute_stride; - vd.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - attribute_stride += sizeof(int16_t) * 4; + vd.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + attribute_stride += sizeof(int8_t) * 4; buffer = s->attribute_buffer; } break; case RS::ARRAY_TEX_UV: { @@ -3330,6 +3453,7 @@ void RendererStorageRD::multimesh_allocate_data(RID p_multimesh, int p_instances if (multimesh->buffer.is_valid()) { RD::get_singleton()->free(multimesh->buffer); multimesh->buffer = RID(); + multimesh->uniform_set_2d = RID(); //cleared by dependency multimesh->uniform_set_3d = RID(); //cleared by dependency } @@ -3412,10 +3536,10 @@ void RendererStorageRD::_multimesh_make_local(MultiMesh *multimesh) const { Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); { const uint8_t *r = buffer.ptr(); - copymem(w, r, buffer.size()); + memcpy(w, r, buffer.size()); } } else { - zeromem(w, multimesh->instances * multimesh->stride_cache * sizeof(float)); + memset(w, 0, multimesh->instances * multimesh->stride_cache * sizeof(float)); } } uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; @@ -3477,7 +3601,7 @@ void RendererStorageRD::_multimesh_re_create_aabb(MultiMesh *multimesh, const fl AABB mesh_aabb = mesh_get_aabb(multimesh->mesh); for (int i = 0; i < p_instances; i++) { const float *data = p_data + multimesh->stride_cache * i; - Transform t; + Transform3D t; if (multimesh->xform_format == RS::MULTIMESH_TRANSFORM_3D) { t.basis.elements[0][0] = data[0]; @@ -3513,7 +3637,7 @@ void RendererStorageRD::_multimesh_re_create_aabb(MultiMesh *multimesh, const fl multimesh->aabb = aabb; } -void RendererStorageRD::multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform &p_transform) { +void RendererStorageRD::multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform3D &p_transform) { MultiMesh *multimesh = multimesh_owner.getornull(p_multimesh); ERR_FAIL_COND(!multimesh); ERR_FAIL_INDEX(p_index, multimesh->instances); @@ -3620,15 +3744,15 @@ RID RendererStorageRD::multimesh_get_mesh(RID p_multimesh) const { return multimesh->mesh; } -Transform RendererStorageRD::multimesh_instance_get_transform(RID p_multimesh, int p_index) const { +Transform3D RendererStorageRD::multimesh_instance_get_transform(RID p_multimesh, int p_index) const { MultiMesh *multimesh = multimesh_owner.getornull(p_multimesh); - ERR_FAIL_COND_V(!multimesh, Transform()); - ERR_FAIL_INDEX_V(p_index, multimesh->instances, Transform()); - ERR_FAIL_COND_V(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_3D, Transform()); + ERR_FAIL_COND_V(!multimesh, Transform3D()); + ERR_FAIL_INDEX_V(p_index, multimesh->instances, Transform3D()); + ERR_FAIL_COND_V(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_3D, Transform3D()); _multimesh_make_local(multimesh); - Transform t; + Transform3D t; { const float *r = multimesh->data_cache.ptr(); @@ -3771,7 +3895,7 @@ Vector<float> RendererStorageRD::multimesh_get_buffer(RID p_multimesh) const { { float *w = ret.ptrw(); const uint8_t *r = buffer.ptr(); - copymem(w, r, buffer.size()); + memcpy(w, r, buffer.size()); } return ret; @@ -3822,7 +3946,7 @@ void RendererStorageRD::_update_dirty_multimeshes() { if (multimesh->data_cache_used_dirty_regions) { uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; - uint32_t visible_region_count = (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); @@ -3873,6 +3997,18 @@ void RendererStorageRD::particles_initialize(RID p_rid) { particles_owner.initialize_rid(p_rid, Particles()); } +void RendererStorageRD::particles_set_mode(RID p_particles, RS::ParticlesMode p_mode) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + if (particles->mode == p_mode) { + return; + } + + _particles_free_data(particles); + + particles->mode = p_mode; +} + void RendererStorageRD::particles_set_emitting(RID p_particles, bool p_emitting) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -3881,6 +4017,7 @@ void RendererStorageRD::particles_set_emitting(RID p_particles, bool p_emitting) } bool RendererStorageRD::particles_get_emitting(RID p_particles) { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, false); @@ -3888,22 +4025,37 @@ bool RendererStorageRD::particles_get_emitting(RID p_particles) { } void RendererStorageRD::_particles_free_data(Particles *particles) { - if (!particles->particle_buffer.is_valid()) { - return; + if (particles->particle_buffer.is_valid()) { + RD::get_singleton()->free(particles->particle_buffer); + particles->particle_buffer = RID(); + RD::get_singleton()->free(particles->particle_instance_buffer); + particles->particle_instance_buffer = RID(); + } + + if (particles->frame_params_buffer.is_valid()) { + RD::get_singleton()->free(particles->frame_params_buffer); + particles->frame_params_buffer = RID(); } - RD::get_singleton()->free(particles->particle_buffer); - RD::get_singleton()->free(particles->frame_params_buffer); - RD::get_singleton()->free(particles->particle_instance_buffer); particles->particles_transforms_buffer_uniform_set = RID(); - particles->particle_buffer = RID(); + if (RD::get_singleton()->uniform_set_is_valid(particles->trail_bind_pose_uniform_set)) { + RD::get_singleton()->free(particles->trail_bind_pose_uniform_set); + } + particles->trail_bind_pose_uniform_set = RID(); + + if (particles->trail_bind_pose_buffer.is_valid()) { + RD::get_singleton()->free(particles->trail_bind_pose_buffer); + particles->trail_bind_pose_buffer = RID(); + } if (RD::get_singleton()->uniform_set_is_valid(particles->collision_textures_uniform_set)) { RD::get_singleton()->free(particles->collision_textures_uniform_set); } + particles->collision_textures_uniform_set = RID(); if (particles->particles_sort_buffer.is_valid()) { RD::get_singleton()->free(particles->particles_sort_buffer); particles->particles_sort_buffer = RID(); + particles->particles_sort_uniform_set = RID(); } if (particles->emission_buffer != nullptr) { @@ -3912,6 +4064,12 @@ void RendererStorageRD::_particles_free_data(Particles *particles) { RD::get_singleton()->free(particles->emission_storage_buffer); particles->emission_storage_buffer = RID(); } + + if (RD::get_singleton()->uniform_set_is_valid(particles->particles_material_uniform_set)) { + //will need to be re-created + RD::get_singleton()->free(particles->particles_material_uniform_set); + } + particles->particles_material_uniform_set = RID(); } void RendererStorageRD::particles_set_amount(RID p_particles, int p_amount) { @@ -3926,41 +4084,15 @@ void RendererStorageRD::particles_set_amount(RID p_particles, int p_amount) { particles->amount = p_amount; - if (particles->amount > 0) { - particles->particle_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ParticleData) * p_amount); - particles->frame_params_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ParticlesFrameParams) * 1); - particles->particle_instance_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * 4 * (3 + 1 + 1) * p_amount); - //needs to clear it - - { - Vector<RD::Uniform> uniforms; - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 1; - u.ids.push_back(particles->particle_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 2; - u.ids.push_back(particles->particle_instance_buffer); - uniforms.push_back(u); - } - - particles->particles_copy_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, 0), 0); - } - } - particles->prev_ticks = 0; particles->phase = 0; particles->prev_phase = 0; particles->clear = true; + + particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); } -void RendererStorageRD::particles_set_lifetime(RID p_particles, float p_lifetime) { +void RendererStorageRD::particles_set_lifetime(RID p_particles, double p_lifetime) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->lifetime = p_lifetime; @@ -3972,17 +4104,17 @@ void RendererStorageRD::particles_set_one_shot(RID p_particles, bool p_one_shot) particles->one_shot = p_one_shot; } -void RendererStorageRD::particles_set_pre_process_time(RID p_particles, float p_time) { +void RendererStorageRD::particles_set_pre_process_time(RID p_particles, double p_time) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->pre_process_time = p_time; } -void RendererStorageRD::particles_set_explosiveness_ratio(RID p_particles, float p_ratio) { +void RendererStorageRD::particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->explosiveness = p_ratio; } -void RendererStorageRD::particles_set_randomness_ratio(RID p_particles, float p_ratio) { +void RendererStorageRD::particles_set_randomness_ratio(RID p_particles, real_t p_ratio) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->randomness = p_ratio; @@ -3995,7 +4127,7 @@ void RendererStorageRD::particles_set_custom_aabb(RID p_particles, const AABB &p particles->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } -void RendererStorageRD::particles_set_speed_scale(RID p_particles, float p_scale) { +void RendererStorageRD::particles_set_speed_scale(RID p_particles, double p_scale) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -4013,6 +4145,22 @@ void RendererStorageRD::particles_set_fixed_fps(RID p_particles, int p_fps) { ERR_FAIL_COND(!particles); particles->fixed_fps = p_fps; + + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); +} + +void RendererStorageRD::particles_set_interpolate(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + + particles->interpolate = p_enable; } void RendererStorageRD::particles_set_fractional_delta(RID p_particles, bool p_enable) { @@ -4022,13 +4170,56 @@ void RendererStorageRD::particles_set_fractional_delta(RID p_particles, bool p_e particles->fractional_delta = p_enable; } -void RendererStorageRD::particles_set_collision_base_size(RID p_particles, float p_size) { +void RendererStorageRD::particles_set_trails(RID p_particles, bool p_enable, double p_length) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_COND(p_length < 0.1); + p_length = MIN(10.0, p_length); + + particles->trails_enabled = p_enable; + particles->trail_length = p_length; + + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); +} + +void RendererStorageRD::particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + if (particles->trail_bind_pose_buffer.is_valid() && particles->trail_bind_poses.size() != p_bind_poses.size()) { + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + } + particles->trail_bind_poses = p_bind_poses; + particles->trail_bind_poses_dirty = true; + + particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); +} + +void RendererStorageRD::particles_set_collision_base_size(RID p_particles, real_t p_size) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->collision_base_size = p_size; } +void RendererStorageRD::particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + + particles->transform_align = p_transform_align; +} + void RendererStorageRD::particles_set_process_material(RID p_particles, RID p_material) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -4068,7 +4259,7 @@ void RendererStorageRD::_particles_allocate_emission_buffer(Particles *particles ERR_FAIL_COND(particles->emission_buffer != nullptr); particles->emission_buffer_data.resize(sizeof(ParticleEmissionBuffer::Data) * particles->amount + sizeof(uint32_t) * 4); - zeromem(particles->emission_buffer_data.ptrw(), particles->emission_buffer_data.size()); + memset(particles->emission_buffer_data.ptrw(), 0, particles->emission_buffer_data.size()); particles->emission_buffer = (ParticleEmissionBuffer *)particles->emission_buffer_data.ptrw(); particles->emission_buffer->particle_max = particles->amount; @@ -4094,7 +4285,7 @@ void RendererStorageRD::particles_set_subemitter(RID p_particles, RID p_subemitt } } -void RendererStorageRD::particles_emit(RID p_particles, const Transform &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) { +void RendererStorageRD::particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); ERR_FAIL_COND(particles->amount == 0); @@ -4149,21 +4340,31 @@ void RendererStorageRD::particles_request_process(RID p_particles) { } AABB RendererStorageRD::particles_get_current_aabb(RID p_particles) { + if (RSG::threaded) { + WARN_PRINT_ONCE("Calling this function with threaded rendering enabled stalls the renderer, use with care."); + } + const Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, AABB()); + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + Vector<ParticleData> data; - data.resize(particles->amount); + data.resize(total_amount); Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(particles->particle_buffer); - Transform inv = particles->emission_transform.affine_inverse(); + Transform3D inv = particles->emission_transform.affine_inverse(); AABB aabb; if (buffer.size()) { bool first = true; + const ParticleData *particle_data = (const ParticleData *)data.ptr(); - for (int i = 0; i < particles->amount; i++) { + for (int i = 0; i < total_amount; i++) { if (particle_data[i].active) { Vector3 pos = Vector3(particle_data[i].xform[12], particle_data[i].xform[13], particle_data[i].xform[14]); if (!particles->use_local_coords) { @@ -4199,7 +4400,7 @@ AABB RendererStorageRD::particles_get_aabb(RID p_particles) const { return particles->custom_aabb; } -void RendererStorageRD::particles_set_emission_transform(RID p_particles, const Transform &p_transform) { +void RendererStorageRD::particles_set_emission_transform(RID p_particles, const Transform3D &p_transform) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -4224,18 +4425,25 @@ RID RendererStorageRD::particles_get_draw_pass_mesh(RID p_particles, int p_pass) void RendererStorageRD::particles_add_collision(RID p_particles, RID p_particles_collision_instance) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); - particles->collisions.insert(p_particles_collision_instance); } void RendererStorageRD::particles_remove_collision(RID p_particles, RID p_particles_collision_instance) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); - particles->collisions.erase(p_particles_collision_instance); } -void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta) { +void RendererStorageRD::particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND(!particles); + particles->has_sdf_collision = p_enable; + particles->sdf_collision_transform = p_xform; + particles->sdf_collision_to_screen = p_to_screen; + particles->sdf_collision_texture = p_texture; +} + +void RendererStorageRD::_particles_process(Particles *p_particles, double p_delta) { if (p_particles->particles_material_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(p_particles->particles_material_uniform_set)) { Vector<RD::Uniform> uniforms; @@ -4284,9 +4492,14 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta p_particles->particles_material_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.default_shader_rd, 1); } - float new_phase = Math::fmod((float)p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, (float)1.0); + double new_phase = Math::fmod((double)p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, 1.0); - ParticlesFrameParams &frame_params = p_particles->frame_params; + //move back history (if there is any) + for (uint32_t i = p_particles->frame_history.size() - 1; i > 0; i--) { + p_particles->frame_history[i] = p_particles->frame_history[i - 1]; + } + //update current frame + ParticlesFrameParams &frame_params = p_particles->frame_history[0]; if (p_particles->clear) { p_particles->cycle_number = 0; @@ -4311,12 +4524,16 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta frame_params.randomness = p_particles->randomness; if (p_particles->use_local_coords) { - store_transform(Transform(), frame_params.emission_transform); + store_transform(Transform3D(), frame_params.emission_transform); } else { store_transform(p_particles->emission_transform, frame_params.emission_transform); } frame_params.cycle = p_particles->cycle_number; + frame_params.frame = p_particles->frame_counter++; + frame_params.pad0 = 0; + frame_params.pad1 = 0; + frame_params.pad2 = 0; { //collision and attractors @@ -4327,10 +4544,54 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RID collision_3d_textures[ParticlesFrameParams::MAX_3D_TEXTURES]; RID collision_heightmap_texture; - Transform to_particles; + Transform3D to_particles; if (p_particles->use_local_coords) { to_particles = p_particles->emission_transform.affine_inverse(); } + + if (p_particles->has_sdf_collision && RD::get_singleton()->texture_is_valid(p_particles->sdf_collision_texture)) { + //2D collision + + Transform2D xform = p_particles->sdf_collision_transform; //will use dotproduct manually so invert beforehand + Transform2D revert = xform.affine_inverse(); + frame_params.collider_count = 1; + frame_params.colliders[0].transform[0] = xform.elements[0][0]; + frame_params.colliders[0].transform[1] = xform.elements[0][1]; + frame_params.colliders[0].transform[2] = 0; + frame_params.colliders[0].transform[3] = xform.elements[2][0]; + + frame_params.colliders[0].transform[4] = xform.elements[1][0]; + frame_params.colliders[0].transform[5] = xform.elements[1][1]; + frame_params.colliders[0].transform[6] = 0; + frame_params.colliders[0].transform[7] = xform.elements[2][1]; + + frame_params.colliders[0].transform[8] = revert.elements[0][0]; + frame_params.colliders[0].transform[9] = revert.elements[0][1]; + frame_params.colliders[0].transform[10] = 0; + frame_params.colliders[0].transform[11] = revert.elements[2][0]; + + frame_params.colliders[0].transform[12] = revert.elements[1][0]; + frame_params.colliders[0].transform[13] = revert.elements[1][1]; + frame_params.colliders[0].transform[14] = 0; + frame_params.colliders[0].transform[15] = revert.elements[2][1]; + + frame_params.colliders[0].extents[0] = p_particles->sdf_collision_to_screen.size.x; + frame_params.colliders[0].extents[1] = p_particles->sdf_collision_to_screen.size.y; + frame_params.colliders[0].extents[2] = p_particles->sdf_collision_to_screen.position.x; + frame_params.colliders[0].scale = p_particles->sdf_collision_to_screen.position.y; + frame_params.colliders[0].texture_index = 0; + frame_params.colliders[0].type = ParticlesFrameParams::COLLISION_TYPE_2D_SDF; + + collision_heightmap_texture = p_particles->sdf_collision_texture; + + //replace in all other history frames where used because parameters are no longer valid if screen moves + for (uint32_t i = 1; i < p_particles->frame_history.size(); i++) { + if (p_particles->frame_history[i].collider_count > 0 && p_particles->frame_history[i].colliders[0].type == ParticlesFrameParams::COLLISION_TYPE_2D_SDF) { + p_particles->frame_history[i].colliders[0] = frame_params.colliders[0]; + } + } + } + uint32_t collision_3d_textures_used = 0; for (const Set<RID>::Element *E = p_particles->collisions.front(); E; E = E->next()) { ParticlesCollisionInstance *pci = particles_collision_instance_owner.getornull(E->get()); @@ -4340,7 +4601,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta ParticlesCollision *pc = particles_collision_owner.getornull(pci->collision); ERR_CONTINUE(!pc); - Transform to_collider = pci->transform; + Transform3D to_collider = pci->transform; if (p_particles->use_local_coords) { to_collider = to_particles * to_collider; } @@ -4515,12 +4776,18 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta ParticlesShader::PushConstant push_constant; + int process_amount = p_particles->amount; + + if (p_particles->trails_enabled && p_particles->trail_bind_poses.size() > 1) { + process_amount *= p_particles->trail_bind_poses.size(); + } push_constant.clear = p_particles->clear; push_constant.total_particles = p_particles->amount; push_constant.lifetime = p_particles->lifetime; - push_constant.trail_size = 1; + push_constant.trail_size = p_particles->trail_params.size(); push_constant.use_fractional_delta = p_particles->fractional_delta; push_constant.sub_emitter_mode = !p_particles->emitting && p_particles->emission_buffer && (p_particles->emission_buffer->particle_count > 0 || p_particles->force_sub_emit); + push_constant.trail_pass = false; p_particles->force_sub_emit = false; //reset @@ -4553,7 +4820,17 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta p_particles->clear = false; - RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params); + if (p_particles->trail_params.size() > 1) { + //fill the trail params + for (uint32_t i = 0; i < p_particles->trail_params.size(); i++) { + uint32_t src_idx = i * p_particles->frame_history.size() / p_particles->trail_params.size(); + p_particles->trail_params[i] = p_particles->frame_history[src_idx]; + } + } else { + p_particles->trail_params[0] = p_particles->frame_history[0]; + } + + RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams) * p_particles->trail_params.size(), p_particles->trail_params.ptr()); ParticlesMaterialData *m = (ParticlesMaterialData *)material_get_data(p_particles->process_material, SHADER_TYPE_PARTICLES); if (!m) { @@ -4562,6 +4839,8 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta ERR_FAIL_COND(!m); + p_particles->has_collision_cache = m->shader_data->uses_collision; + //todo should maybe compute all particle systems together? RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, m->shader_data->pipeline); @@ -4575,27 +4854,45 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1); + if (p_particles->trails_enabled && p_particles->trail_bind_poses.size() > 1) { + //trails requires two passes in order to catch particle starts + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount / p_particles->trail_bind_poses.size(), 1, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + push_constant.trail_pass = true; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount - p_particles->amount, 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount, 1, 1); + } RD::get_singleton()->compute_list_end(); } -void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 &p_axis) { +void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); - if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH) { - return; //uninteresting for other modes + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + return; + } + + if (particles->particle_buffer.is_null()) { + return; //particles have not processed yet } + bool do_sort = particles->draw_order == RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH; + //copy to sort buffer - if (particles->particles_sort_buffer == RID()) { + if (do_sort && particles->particles_sort_buffer == RID()) { uint32_t size = particles->amount; if (size & 1) { size++; //make multiple of 16 } size *= sizeof(float) * 2; particles->particles_sort_buffer = RD::get_singleton()->storage_buffer_create(size); + { Vector<RD::Uniform> uniforms; @@ -4611,41 +4908,114 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & } } + ParticlesShader::CopyPushConstant copy_push_constant; + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + int fixed_fps = 60.0; + if (particles->fixed_fps > 0) { + fixed_fps = particles->fixed_fps; + } + + copy_push_constant.trail_size = particles->trail_bind_poses.size(); + copy_push_constant.trail_total = particles->frame_history.size(); + copy_push_constant.frame_delta = 1.0 / fixed_fps; + } else { + copy_push_constant.trail_size = 1; + copy_push_constant.trail_total = 1; + copy_push_constant.frame_delta = 0.0; + } + + copy_push_constant.order_by_lifetime = (particles->draw_order == RS::PARTICLES_DRAW_ORDER_LIFETIME || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME); + copy_push_constant.lifetime_split = MIN(particles->amount * particles->phase, particles->amount - 1); + copy_push_constant.lifetime_reverse = particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME; + + copy_push_constant.frame_remainder = particles->interpolate ? particles->frame_remainder : 0.0; + copy_push_constant.total_particles = particles->amount; + Vector3 axis = -p_axis; // cameras look to z negative if (particles->use_local_coords) { axis = particles->emission_transform.basis.xform_inv(axis).normalized(); } - ParticlesShader::CopyPushConstant copy_push_constant; - copy_push_constant.total_particles = particles->amount; copy_push_constant.sort_direction[0] = axis.x; copy_push_constant.sort_direction[1] = axis.y; copy_push_constant.sort_direction[2] = axis.z; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_SORT_BUFFER]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); + copy_push_constant.align_up[0] = p_up_axis.x; + copy_push_constant.align_up[1] = p_up_axis.y; + copy_push_constant.align_up[2] = p_up_axis.z; - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); + copy_push_constant.align_mode = particles->transform_align; - RD::get_singleton()->compute_list_end(); + if (do_sort) { + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_SORT_BUFFER]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); + + RD::get_singleton()->compute_list_end(); + effects->sort_buffer(particles->particles_sort_uniform_set, particles->amount); + } - effects.sort_buffer(particles->particles_sort_uniform_set, particles->amount); + copy_push_constant.total_particles *= copy_push_constant.total_particles; - compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_INSTANCES_WITH_SORT_BUFFER]); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[do_sort ? ParticlesShader::COPY_MODE_FILL_INSTANCES_WITH_SORT_BUFFER : (particles->mode == RS::PARTICLES_MODE_2D ? ParticlesShader::COPY_MODE_FILL_INSTANCES_2D : ParticlesShader::COPY_MODE_FILL_INSTANCES)]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); + if (do_sort) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); + } + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, copy_push_constant.total_particles, 1, 1); RD::get_singleton()->compute_list_end(); } +void RendererStorageRD::_particles_update_buffers(Particles *particles) { + if (particles->amount > 0 && particles->particle_buffer.is_null()) { + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + + uint32_t xform_size = particles->mode == RS::PARTICLES_MODE_2D ? 2 : 3; + + particles->particle_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ParticleData) * total_amount); + + particles->particle_instance_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * 4 * (xform_size + 1 + 1) * total_amount); + //needs to clear it + + { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.ids.push_back(particles->particle_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(particles->particle_instance_buffer); + uniforms.push_back(u); + } + + particles->particles_copy_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, 0), 0); + } + } +} void RendererStorageRD::update_particles() { while (particle_update_list) { //use transform feedback to process particles @@ -4657,6 +5027,8 @@ void RendererStorageRD::update_particles() { particles->update_list = nullptr; particles->dirty = false; + _particles_update_buffers(particles); + if (particles->restart_request) { particles->prev_ticks = 0; particles->phase = 0; @@ -4688,17 +5060,86 @@ void RendererStorageRD::update_particles() { } } +#ifndef _MSC_VER +#warning Should use display refresh rate for all this +#endif + + float screen_hz = 60; + + int fixed_fps = 0; + if (particles->fixed_fps > 0) { + fixed_fps = particles->fixed_fps; + } else if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + fixed_fps = screen_hz; + } + { + //update trails + int history_size = 1; + int trail_steps = 1; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + history_size = MAX(1, int(particles->trail_length * fixed_fps)); + trail_steps = particles->trail_bind_poses.size(); + } + + if (uint32_t(history_size) != particles->frame_history.size()) { + particles->frame_history.resize(history_size); + memset(particles->frame_history.ptr(), 0, sizeof(ParticlesFrameParams) * history_size); + } + + if (uint32_t(trail_steps) != particles->trail_params.size() || particles->frame_params_buffer.is_null()) { + particles->trail_params.resize(trail_steps); + if (particles->frame_params_buffer.is_valid()) { + RD::get_singleton()->free(particles->frame_params_buffer); + } + particles->frame_params_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ParticlesFrameParams) * trail_steps); + } + + if (particles->trail_bind_poses.size() > 1 && particles->trail_bind_pose_buffer.is_null()) { + particles->trail_bind_pose_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * 16 * particles->trail_bind_poses.size()); + particles->trail_bind_poses_dirty = true; + } + + if (particles->trail_bind_pose_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + if (particles->trail_bind_pose_buffer.is_valid()) { + u.ids.push_back(particles->trail_bind_pose_buffer); + } else { + u.ids.push_back(default_rd_storage_buffer); + } + uniforms.push_back(u); + } + + particles->trail_bind_pose_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, 0), 2); + } + + if (particles->trail_bind_pose_buffer.is_valid() && particles->trail_bind_poses_dirty) { + if (particles_shader.pose_update_buffer.size() < uint32_t(particles->trail_bind_poses.size()) * 16) { + particles_shader.pose_update_buffer.resize(particles->trail_bind_poses.size() * 16); + } + + for (int i = 0; i < particles->trail_bind_poses.size(); i++) { + store_transform(particles->trail_bind_poses[i], &particles_shader.pose_update_buffer[i * 16]); + } + + RD::get_singleton()->buffer_update(particles->trail_bind_pose_buffer, 0, particles->trail_bind_poses.size() * 16 * sizeof(float), particles_shader.pose_update_buffer.ptr()); + } + } + bool zero_time_scale = Engine::get_singleton()->get_time_scale() <= 0.0; if (particles->clear && particles->pre_process_time > 0.0) { - float frame_time; - if (particles->fixed_fps > 0) { - frame_time = 1.0 / particles->fixed_fps; + double frame_time; + if (fixed_fps > 0) { + frame_time = 1.0 / fixed_fps; } else { frame_time = 1.0 / 30.0; } - float todo = particles->pre_process_time; + double todo = particles->pre_process_time; while (todo >= 0) { _particles_process(particles, frame_time); @@ -4706,23 +5147,23 @@ void RendererStorageRD::update_particles() { } } - if (particles->fixed_fps > 0) { - float frame_time; - float decr; + if (fixed_fps > 0) { + double frame_time; + double decr; if (zero_time_scale) { frame_time = 0.0; - decr = 1.0 / particles->fixed_fps; + decr = 1.0 / fixed_fps; } else { - frame_time = 1.0 / particles->fixed_fps; + frame_time = 1.0 / fixed_fps; decr = frame_time; } - float delta = RendererCompositorRD::singleton->get_frame_delta_time(); + double delta = RendererCompositorRD::singleton->get_frame_delta_time(); if (delta > 0.1) { //avoid recursive stalls if fps goes below 10 delta = 0.1; } else if (delta <= 0.0) { //unlikely but.. delta = 0.001; } - float todo = particles->frame_remainder + delta; + double todo = particles->frame_remainder + delta; while (todo >= frame_time) { _particles_process(particles, frame_time); @@ -4741,16 +5182,43 @@ void RendererStorageRD::update_particles() { //copy particles to instance buffer - if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH) { + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + //does not need view dependent operation, do copy here ParticlesShader::CopyPushConstant copy_push_constant; - copy_push_constant.total_particles = particles->amount; + + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + + copy_push_constant.total_particles = total_amount; + copy_push_constant.frame_remainder = particles->interpolate ? particles->frame_remainder : 0.0; + copy_push_constant.align_mode = particles->transform_align; + copy_push_constant.align_up[0] = 0; + copy_push_constant.align_up[1] = 0; + copy_push_constant.align_up[2] = 0; + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + copy_push_constant.trail_size = particles->trail_bind_poses.size(); + copy_push_constant.trail_total = particles->frame_history.size(); + copy_push_constant.frame_delta = 1.0 / fixed_fps; + } else { + copy_push_constant.trail_size = 1; + copy_push_constant.trail_total = 1; + copy_push_constant.frame_delta = 0.0; + } + + copy_push_constant.order_by_lifetime = (particles->draw_order == RS::PARTICLES_DRAW_ORDER_LIFETIME || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME); + copy_push_constant.lifetime_split = MIN(particles->amount * particles->phase, particles->amount - 1); + copy_push_constant.lifetime_reverse = particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_INSTANCES]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[particles->mode == RS::PARTICLES_MODE_2D ? ParticlesShader::COPY_MODE_FILL_INSTANCES_2D : ParticlesShader::COPY_MODE_FILL_INSTANCES]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, total_amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4760,6 +5228,7 @@ void RendererStorageRD::update_particles() { } bool RendererStorageRD::particles_is_inactive(RID p_particles) const { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); const Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, false); return !particles->emitting && particles->inactive; @@ -4774,6 +5243,7 @@ void RendererStorageRD::ParticlesShaderData::set_code(const String &p_code) { valid = false; ubo_size = 0; uniforms.clear(); + uses_collision = false; if (code == String()) { return; //just invalid, but no error @@ -4781,6 +5251,8 @@ void RendererStorageRD::ParticlesShaderData::set_code(const String &p_code) { ShaderCompilerRD::GeneratedCode gen_code; ShaderCompilerRD::IdentifierActions actions; + actions.entry_point_stages["start"] = ShaderCompilerRD::STAGE_COMPUTE; + actions.entry_point_stages["process"] = ShaderCompilerRD::STAGE_COMPUTE; /* uses_time = false; @@ -4791,17 +5263,18 @@ void RendererStorageRD::ParticlesShaderData::set_code(const String &p_code) { actions.usage_flag_pointers["TIME"] = &uses_time; */ + actions.usage_flag_pointers["COLLIDED"] = &uses_collision; + actions.uniforms = &uniforms; Error err = base_singleton->particles_shader.compiler.compile(RS::SHADER_PARTICLES, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = base_singleton->particles_shader.shader.version_create(); } - base_singleton->particles_shader.shader.version_set_compute_code(version, gen_code.uniforms, gen_code.compute_global, gen_code.compute, gen_code.defines); + base_singleton->particles_shader.shader.version_set_compute_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompilerRD::STAGE_COMPUTE], gen_code.defines); ERR_FAIL_COND(!base_singleton->particles_shader.shader.version_is_valid(version)); ubo_size = gen_code.uniform_total_size; @@ -4905,94 +5378,14 @@ RendererStorageRD::ShaderData *RendererStorageRD::_create_particles_shader_func( return shader_data; } -void RendererStorageRD::ParticlesMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererStorageRD::ParticlesMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { uniform_set_updated = true; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, base_singleton->particles_shader.shader.version_get_shader(shader_data->version, 0), 3); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, base_singleton->particles_shader.shader.version_get_shader(shader_data->version, 0), 3); } RendererStorageRD::ParticlesMaterialData::~ParticlesMaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *RendererStorageRD::_create_particles_material_func(ParticlesShaderData *p_shader) { @@ -5070,7 +5463,7 @@ void RendererStorageRD::particles_collision_set_cull_mask(RID p_particles_collis particles_collision->cull_mask = p_cull_mask; } -void RendererStorageRD::particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius) { +void RendererStorageRD::particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); @@ -5086,21 +5479,21 @@ void RendererStorageRD::particles_collision_set_box_extents(RID p_particles_coll particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } -void RendererStorageRD::particles_collision_set_attractor_strength(RID p_particles_collision, float p_strength) { +void RendererStorageRD::particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); particles_collision->attractor_strength = p_strength; } -void RendererStorageRD::particles_collision_set_attractor_directionality(RID p_particles_collision, float p_directionality) { +void RendererStorageRD::particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); particles_collision->attractor_directionality = p_directionality; } -void RendererStorageRD::particles_collision_set_attractor_attenuation(RID p_particles_collision, float p_curve) { +void RendererStorageRD::particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); @@ -5177,7 +5570,7 @@ RID RendererStorageRD::particles_collision_instance_create(RID p_collision) { pci.collision = p_collision; return particles_collision_instance_owner.make_rid(pci); } -void RendererStorageRD::particles_collision_instance_set_transform(RID p_collision_instance, const Transform &p_transform) { +void RendererStorageRD::particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform) { ParticlesCollisionInstance *pci = particles_collision_instance_owner.getornull(p_collision_instance); ERR_FAIL_COND(!pci); pci->transform = p_transform; @@ -5188,6 +5581,59 @@ void RendererStorageRD::particles_collision_instance_set_active(RID p_collision_ pci->active = p_active; } +/* VISIBILITY NOTIFIER */ + +RID RendererStorageRD::visibility_notifier_allocate() { + return visibility_notifier_owner.allocate_rid(); +} +void RendererStorageRD::visibility_notifier_initialize(RID p_notifier) { + visibility_notifier_owner.initialize_rid(p_notifier, VisibilityNotifier()); +} +void RendererStorageRD::visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + vn->aabb = p_aabb; + vn->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); +} +void RendererStorageRD::visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + vn->enter_callback = p_enter_callbable; + vn->exit_callback = p_exit_callable; +} + +AABB RendererStorageRD::visibility_notifier_get_aabb(RID p_notifier) const { + const VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND_V(!vn, AABB()); + return vn->aabb; +} +void RendererStorageRD::visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + + if (p_enter) { + if (!vn->enter_callback.is_null()) { + if (p_deferred) { + vn->enter_callback.call_deferred(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->enter_callback.call(nullptr, 0, r, ce); + } + } + } else { + if (!vn->exit_callback.is_null()) { + if (p_deferred) { + vn->exit_callback.call_deferred(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->exit_callback.call(nullptr, 0, r, ce); + } + } + } +} + /* SKELETON API */ RID RendererStorageRD::skeleton_allocate() { @@ -5228,7 +5674,7 @@ void RendererStorageRD::skeleton_allocate_data(RID p_skeleton, int p_bones, bool if (skeleton->size) { skeleton->data.resize(skeleton->size * (skeleton->use_2d ? 8 : 12)); skeleton->buffer = RD::get_singleton()->storage_buffer_create(skeleton->data.size() * sizeof(float)); - zeromem(skeleton->data.ptrw(), skeleton->data.size() * sizeof(float)); + memset(skeleton->data.ptrw(), 0, skeleton->data.size() * sizeof(float)); _skeleton_make_dirty(skeleton); @@ -5255,7 +5701,7 @@ int RendererStorageRD::skeleton_get_bone_count(RID p_skeleton) const { return skeleton->size; } -void RendererStorageRD::skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform &p_transform) { +void RendererStorageRD::skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform3D &p_transform) { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); ERR_FAIL_COND(!skeleton); @@ -5280,16 +5726,16 @@ void RendererStorageRD::skeleton_bone_set_transform(RID p_skeleton, int p_bone, _skeleton_make_dirty(skeleton); } -Transform RendererStorageRD::skeleton_bone_get_transform(RID p_skeleton, int p_bone) const { +Transform3D RendererStorageRD::skeleton_bone_get_transform(RID p_skeleton, int p_bone) const { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); - ERR_FAIL_COND_V(!skeleton, Transform()); - ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform()); - ERR_FAIL_COND_V(skeleton->use_2d, Transform()); + ERR_FAIL_COND_V(!skeleton, Transform3D()); + ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform3D()); + ERR_FAIL_COND_V(skeleton->use_2d, Transform3D()); const float *dataptr = skeleton->data.ptr() + p_bone * 12; - Transform t; + Transform3D t; t.basis.elements[0][0] = dataptr[0]; t.basis.elements[0][1] = dataptr[1]; @@ -5439,6 +5885,10 @@ void RendererStorageRD::light_set_param(RID p_light, RS::LightParam p_param, flo ERR_FAIL_COND(!light); ERR_FAIL_INDEX(p_param, RS::LIGHT_PARAM_MAX); + if (light->param[p_param] == p_value) { + return; + } + switch (p_param) { case RS::LIGHT_PARAM_RANGE: case RS::LIGHT_PARAM_SPOT_ANGLE: @@ -5452,6 +5902,12 @@ void RendererStorageRD::light_set_param(RID p_light, RS::LightParam p_param, flo light->version++; light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } break; + case RS::LIGHT_PARAM_SIZE: { + if ((light->param[p_param] > CMP_EPSILON) != (p_value > CMP_EPSILON)) { + //changing from no size to size and the opposite + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); + } + } break; default: { } } @@ -5488,8 +5944,11 @@ void RendererStorageRD::light_set_projector(RID p_light, RID p_texture) { light->projector = p_texture; - if (light->type != RS::LIGHT_DIRECTIONAL && light->projector.is_valid()) { - texture_add_to_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + if (light->type != RS::LIGHT_DIRECTIONAL) { + if (light->projector.is_valid()) { + texture_add_to_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + } + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); } } @@ -5603,20 +6062,6 @@ RS::LightDirectionalShadowMode RendererStorageRD::light_directional_get_shadow_m return light->directional_shadow_mode; } -void RendererStorageRD::light_directional_set_shadow_depth_range_mode(RID p_light, RS::LightDirectionalShadowDepthRangeMode p_range_mode) { - Light *light = light_owner.getornull(p_light); - ERR_FAIL_COND(!light); - - light->directional_range_mode = p_range_mode; -} - -RS::LightDirectionalShadowDepthRangeMode RendererStorageRD::light_directional_get_shadow_depth_range_mode(RID p_light) const { - const Light *light = light_owner.getornull(p_light); - ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE); - - return light->directional_range_mode; -} - uint32_t RendererStorageRD::light_get_max_sdfgi_cascade(RID p_light) { const Light *light = light_owner.getornull(p_light); ERR_FAIL_COND_V(!light, 0); @@ -5979,36 +6424,36 @@ AABB RendererStorageRD::decal_get_aabb(RID p_decal) const { return AABB(-decal->extents, decal->extents * 2.0); } -RID RendererStorageRD::gi_probe_allocate() { - return gi_probe_owner.allocate_rid(); +RID RendererStorageRD::voxel_gi_allocate() { + return voxel_gi_owner.allocate_rid(); } -void RendererStorageRD::gi_probe_initialize(RID p_gi_probe) { - gi_probe_owner.initialize_rid(p_gi_probe, GIProbe()); +void RendererStorageRD::voxel_gi_initialize(RID p_voxel_gi) { + voxel_gi_owner.initialize_rid(p_voxel_gi, VoxelGI()); } -void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_allocate_data(RID p_voxel_gi, const Transform3D &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - if (gi_probe->octree_buffer.is_valid()) { - RD::get_singleton()->free(gi_probe->octree_buffer); - RD::get_singleton()->free(gi_probe->data_buffer); - if (gi_probe->sdf_texture.is_valid()) { - RD::get_singleton()->free(gi_probe->sdf_texture); + if (voxel_gi->octree_buffer.is_valid()) { + RD::get_singleton()->free(voxel_gi->octree_buffer); + RD::get_singleton()->free(voxel_gi->data_buffer); + if (voxel_gi->sdf_texture.is_valid()) { + RD::get_singleton()->free(voxel_gi->sdf_texture); } - gi_probe->sdf_texture = RID(); - gi_probe->octree_buffer = RID(); - gi_probe->data_buffer = RID(); - gi_probe->octree_buffer_size = 0; - gi_probe->data_buffer_size = 0; - gi_probe->cell_count = 0; + voxel_gi->sdf_texture = RID(); + voxel_gi->octree_buffer = RID(); + voxel_gi->data_buffer = RID(); + voxel_gi->octree_buffer_size = 0; + voxel_gi->data_buffer_size = 0; + voxel_gi->cell_count = 0; } - gi_probe->to_cell_xform = p_to_cell_xform; - gi_probe->bounds = p_aabb; - gi_probe->octree_size = p_octree_size; - gi_probe->level_counts = p_level_counts; + voxel_gi->to_cell_xform = p_to_cell_xform; + voxel_gi->bounds = p_aabb; + voxel_gi->octree_size = p_octree_size; + voxel_gi->level_counts = p_level_counts; if (p_octree_cells.size()) { ERR_FAIL_COND(p_octree_cells.size() % 32 != 0); //cells size must be a multiple of 32 @@ -6017,42 +6462,42 @@ void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform & ERR_FAIL_COND(p_data_cells.size() != (int)cell_count * 16); //see that data size matches - gi_probe->cell_count = cell_count; - gi_probe->octree_buffer = RD::get_singleton()->storage_buffer_create(p_octree_cells.size(), p_octree_cells); - gi_probe->octree_buffer_size = p_octree_cells.size(); - gi_probe->data_buffer = RD::get_singleton()->storage_buffer_create(p_data_cells.size(), p_data_cells); - gi_probe->data_buffer_size = p_data_cells.size(); + voxel_gi->cell_count = cell_count; + voxel_gi->octree_buffer = RD::get_singleton()->storage_buffer_create(p_octree_cells.size(), p_octree_cells); + voxel_gi->octree_buffer_size = p_octree_cells.size(); + voxel_gi->data_buffer = RD::get_singleton()->storage_buffer_create(p_data_cells.size(), p_data_cells); + voxel_gi->data_buffer_size = p_data_cells.size(); if (p_distance_field.size()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8_UNORM; - tf.width = gi_probe->octree_size.x; - tf.height = gi_probe->octree_size.y; - tf.depth = gi_probe->octree_size.z; + tf.width = voxel_gi->octree_size.x; + tf.height = voxel_gi->octree_size.y; + tf.depth = voxel_gi->octree_size.z; tf.texture_type = RD::TEXTURE_TYPE_3D; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; Vector<Vector<uint8_t>> s; s.push_back(p_distance_field); - gi_probe->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView(), s); + voxel_gi->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView(), s); } #if 0 { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8_UNORM; - tf.width = gi_probe->octree_size.x; - tf.height = gi_probe->octree_size.y; - tf.depth = gi_probe->octree_size.z; + tf.width = voxel_gi->octree_size.x; + tf.height = voxel_gi->octree_size.y; + tf.depth = voxel_gi->octree_size.z; tf.type = RD::TEXTURE_TYPE_3D; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; tf.shareable_formats.push_back(RD::DATA_FORMAT_R8_UNORM); tf.shareable_formats.push_back(RD::DATA_FORMAT_R8_UINT); - gi_probe->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + voxel_gi->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); } RID shared_tex; { RD::TextureView tv; tv.format_override = RD::DATA_FORMAT_R8_UINT; - shared_tex = RD::get_singleton()->texture_create_shared(tv, gi_probe->sdf_texture); + shared_tex = RD::get_singleton()->texture_create_shared(tv, voxel_gi->sdf_texture); } //update SDF texture Vector<RD::Uniform> uniforms; @@ -6060,14 +6505,14 @@ void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform & RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 1; - u.ids.push_back(gi_probe->octree_buffer); + u.ids.push_back(voxel_gi->octree_buffer); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 2; - u.ids.push_back(gi_probe->data_buffer); + u.ids.push_back(voxel_gi->data_buffer); uniforms.push_back(u); } { @@ -6078,24 +6523,24 @@ void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform & uniforms.push_back(u); } - RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_sdf_shader_version_shader, 0); + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, voxel_gi_sdf_shader_version_shader, 0); { uint32_t push_constant[4] = { 0, 0, 0, 0 }; - for (int i = 0; i < gi_probe->level_counts.size() - 1; i++) { - push_constant[0] += gi_probe->level_counts[i]; + for (int i = 0; i < voxel_gi->level_counts.size() - 1; i++) { + push_constant[0] += voxel_gi->level_counts[i]; } - push_constant[1] = push_constant[0] + gi_probe->level_counts[gi_probe->level_counts.size() - 1]; + push_constant[1] = push_constant[0] + voxel_gi->level_counts[voxel_gi->level_counts.size() - 1]; print_line("offset: " + itos(push_constant[0])); print_line("size: " + itos(push_constant[1])); //create SDF RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_sdf_shader_pipeline); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, voxel_gi_sdf_shader_pipeline); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, push_constant, sizeof(uint32_t) * 4); - RD::get_singleton()->compute_list_dispatch(compute_list, gi_probe->octree_size.x / 4, gi_probe->octree_size.y / 4, gi_probe->octree_size.z / 4); + RD::get_singleton()->compute_list_dispatch(compute_list, voxel_gi->octree_size.x / 4, voxel_gi->octree_size.y / 4, voxel_gi->octree_size.z / 4); RD::get_singleton()->compute_list_end(); } @@ -6105,232 +6550,206 @@ void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform & #endif } - gi_probe->version++; - gi_probe->data_version++; + voxel_gi->version++; + voxel_gi->data_version++; - gi_probe->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); + voxel_gi->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } -AABB RendererStorageRD::gi_probe_get_bounds(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, AABB()); +AABB RendererStorageRD::voxel_gi_get_bounds(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, AABB()); - return gi_probe->bounds; + return voxel_gi->bounds; } -Vector3i RendererStorageRD::gi_probe_get_octree_size(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Vector3i()); - return gi_probe->octree_size; +Vector3i RendererStorageRD::voxel_gi_get_octree_size(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector3i()); + return voxel_gi->octree_size; } -Vector<uint8_t> RendererStorageRD::gi_probe_get_octree_cells(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Vector<uint8_t>()); +Vector<uint8_t> RendererStorageRD::voxel_gi_get_octree_cells(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); - if (gi_probe->octree_buffer.is_valid()) { - return RD::get_singleton()->buffer_get_data(gi_probe->octree_buffer); + if (voxel_gi->octree_buffer.is_valid()) { + return RD::get_singleton()->buffer_get_data(voxel_gi->octree_buffer); } return Vector<uint8_t>(); } -Vector<uint8_t> RendererStorageRD::gi_probe_get_data_cells(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Vector<uint8_t>()); +Vector<uint8_t> RendererStorageRD::voxel_gi_get_data_cells(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); - if (gi_probe->data_buffer.is_valid()) { - return RD::get_singleton()->buffer_get_data(gi_probe->data_buffer); + if (voxel_gi->data_buffer.is_valid()) { + return RD::get_singleton()->buffer_get_data(voxel_gi->data_buffer); } return Vector<uint8_t>(); } -Vector<uint8_t> RendererStorageRD::gi_probe_get_distance_field(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Vector<uint8_t>()); +Vector<uint8_t> RendererStorageRD::voxel_gi_get_distance_field(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); - if (gi_probe->data_buffer.is_valid()) { - return RD::get_singleton()->texture_get_data(gi_probe->sdf_texture, 0); + if (voxel_gi->data_buffer.is_valid()) { + return RD::get_singleton()->texture_get_data(voxel_gi->sdf_texture, 0); } return Vector<uint8_t>(); } -Vector<int> RendererStorageRD::gi_probe_get_level_counts(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Vector<int>()); +Vector<int> RendererStorageRD::voxel_gi_get_level_counts(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<int>()); - return gi_probe->level_counts; + return voxel_gi->level_counts; } -Transform RendererStorageRD::gi_probe_get_to_cell_xform(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, Transform()); +Transform3D RendererStorageRD::voxel_gi_get_to_cell_xform(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Transform3D()); - return gi_probe->to_cell_xform; + return voxel_gi->to_cell_xform; } -void RendererStorageRD::gi_probe_set_dynamic_range(RID p_gi_probe, float p_range) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_dynamic_range(RID p_voxel_gi, float p_range) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->dynamic_range = p_range; - gi_probe->version++; + voxel_gi->dynamic_range = p_range; + voxel_gi->version++; } -float RendererStorageRD::gi_probe_get_dynamic_range(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); +float RendererStorageRD::voxel_gi_get_dynamic_range(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); - return gi_probe->dynamic_range; + return voxel_gi->dynamic_range; } -void RendererStorageRD::gi_probe_set_propagation(RID p_gi_probe, float p_range) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_propagation(RID p_voxel_gi, float p_range) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->propagation = p_range; - gi_probe->version++; + voxel_gi->propagation = p_range; + voxel_gi->version++; } -float RendererStorageRD::gi_probe_get_propagation(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->propagation; +float RendererStorageRD::voxel_gi_get_propagation(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->propagation; } -void RendererStorageRD::gi_probe_set_energy(RID p_gi_probe, float p_energy) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_energy(RID p_voxel_gi, float p_energy) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->energy = p_energy; + voxel_gi->energy = p_energy; } -float RendererStorageRD::gi_probe_get_energy(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->energy; +float RendererStorageRD::voxel_gi_get_energy(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->energy; } -void RendererStorageRD::gi_probe_set_ao(RID p_gi_probe, float p_ao) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_bias(RID p_voxel_gi, float p_bias) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->ao = p_ao; + voxel_gi->bias = p_bias; } -float RendererStorageRD::gi_probe_get_ao(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->ao; +float RendererStorageRD::voxel_gi_get_bias(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->bias; } -void RendererStorageRD::gi_probe_set_ao_size(RID p_gi_probe, float p_strength) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_normal_bias(RID p_voxel_gi, float p_normal_bias) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->ao_size = p_strength; + voxel_gi->normal_bias = p_normal_bias; } -float RendererStorageRD::gi_probe_get_ao_size(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->ao_size; +float RendererStorageRD::voxel_gi_get_normal_bias(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->normal_bias; } -void RendererStorageRD::gi_probe_set_bias(RID p_gi_probe, float p_bias) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_anisotropy_strength(RID p_voxel_gi, float p_strength) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->bias = p_bias; + voxel_gi->anisotropy_strength = p_strength; } -float RendererStorageRD::gi_probe_get_bias(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->bias; +float RendererStorageRD::voxel_gi_get_anisotropy_strength(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->anisotropy_strength; } -void RendererStorageRD::gi_probe_set_normal_bias(RID p_gi_probe, float p_normal_bias) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); - - gi_probe->normal_bias = p_normal_bias; -} +void RendererStorageRD::voxel_gi_set_interior(RID p_voxel_gi, bool p_enable) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); -float RendererStorageRD::gi_probe_get_normal_bias(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->normal_bias; + voxel_gi->interior = p_enable; } -void RendererStorageRD::gi_probe_set_anisotropy_strength(RID p_gi_probe, float p_strength) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); +void RendererStorageRD::voxel_gi_set_use_two_bounces(RID p_voxel_gi, bool p_enable) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); - gi_probe->anisotropy_strength = p_strength; + voxel_gi->use_two_bounces = p_enable; + voxel_gi->version++; } -float RendererStorageRD::gi_probe_get_anisotropy_strength(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->anisotropy_strength; +bool RendererStorageRD::voxel_gi_is_using_two_bounces(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, false); + return voxel_gi->use_two_bounces; } -void RendererStorageRD::gi_probe_set_interior(RID p_gi_probe, bool p_enable) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); - - gi_probe->interior = p_enable; -} - -void RendererStorageRD::gi_probe_set_use_two_bounces(RID p_gi_probe, bool p_enable) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); - - gi_probe->use_two_bounces = p_enable; - gi_probe->version++; -} - -bool RendererStorageRD::gi_probe_is_using_two_bounces(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, false); - return gi_probe->use_two_bounces; -} - -bool RendererStorageRD::gi_probe_is_interior(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->interior; +bool RendererStorageRD::voxel_gi_is_interior(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->interior; } -uint32_t RendererStorageRD::gi_probe_get_version(RID p_gi_probe) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->version; +uint32_t RendererStorageRD::voxel_gi_get_version(RID p_voxel_gi) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->version; } -uint32_t RendererStorageRD::gi_probe_get_data_version(RID p_gi_probe) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, 0); - return gi_probe->data_version; +uint32_t RendererStorageRD::voxel_gi_get_data_version(RID p_voxel_gi) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->data_version; } -RID RendererStorageRD::gi_probe_get_octree_buffer(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, RID()); - return gi_probe->octree_buffer; +RID RendererStorageRD::voxel_gi_get_octree_buffer(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->octree_buffer; } -RID RendererStorageRD::gi_probe_get_data_buffer(RID p_gi_probe) const { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, RID()); - return gi_probe->data_buffer; +RID RendererStorageRD::voxel_gi_get_data_buffer(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->data_buffer; } -RID RendererStorageRD::gi_probe_get_sdf_texture(RID p_gi_probe) { - GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); - ERR_FAIL_COND_V(!gi_probe, RID()); +RID RendererStorageRD::voxel_gi_get_sdf_texture(RID p_voxel_gi) { + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); - return gi_probe->sdf_texture; + return voxel_gi->sdf_texture; } /* LIGHTMAP API */ @@ -6566,9 +6985,13 @@ void RendererStorageRD::_update_render_target(RenderTarget *rt) { rd_format.width = rt->size.width; rd_format.height = rt->size.height; rd_format.depth = 1; - rd_format.array_layers = 1; + rd_format.array_layers = rt->view_count; // for stereo we create two (or more) layers, need to see if we can make fallback work like this too if we don't have multiview rd_format.mipmaps = 1; - rd_format.texture_type = RD::TEXTURE_TYPE_2D; + if (rd_format.array_layers > 1) { // why are we not using rt->texture_type ?? + rd_format.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } else { + rd_format.texture_type = RD::TEXTURE_TYPE_2D; + } rd_format.samples = RD::TEXTURE_SAMPLES_1; rd_format.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; rd_format.shareable_formats.push_back(rt->color_format); @@ -6580,7 +7003,7 @@ void RendererStorageRD::_update_render_target(RenderTarget *rt) { Vector<RID> fb_textures; fb_textures.push_back(rt->color); - rt->framebuffer = RD::get_singleton()->framebuffer_create(fb_textures); + rt->framebuffer = RD::get_singleton()->framebuffer_create(fb_textures, RenderingDevice::INVALID_ID, rt->view_count); if (rt->framebuffer.is_null()) { _clear_render_target(rt); ERR_FAIL_COND(rt->framebuffer.is_null()); @@ -6694,12 +7117,15 @@ void RendererStorageRD::render_target_set_position(RID p_render_target, int p_x, //unused for this render target } -void RendererStorageRD::render_target_set_size(RID p_render_target, int p_width, int p_height) { +void RendererStorageRD::render_target_set_size(RID p_render_target, int p_width, int p_height, uint32_t p_view_count) { RenderTarget *rt = render_target_owner.getornull(p_render_target); ERR_FAIL_COND(!rt); - rt->size.x = p_width; - rt->size.y = p_height; - _update_render_target(rt); + if (rt->size.x != p_width || rt->size.y != p_height || rt->view_count != p_view_count) { + rt->size.x = p_width; + rt->size.y = p_height; + rt->view_count = p_view_count; + _update_render_target(rt); + } } RID RendererStorageRD::render_target_get_texture(RID p_render_target) { @@ -6856,6 +7282,20 @@ Rect2i RendererStorageRD::render_target_get_sdf_rect(RID p_render_target) const return _render_target_get_sdf_rect(rt); } +void RendererStorageRD::render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) { + RenderTarget *rt = render_target_owner.getornull(p_render_target); + ERR_FAIL_COND(!rt); + + rt->sdf_enabled = p_enabled; +} + +bool RendererStorageRD::render_target_is_sdf_enabled(RID p_render_target) const { + const RenderTarget *rt = render_target_owner.getornull(p_render_target); + ERR_FAIL_COND_V(!rt, false); + + return rt->sdf_enabled; +} + RID RendererStorageRD::render_target_get_sdf_texture(RID p_render_target) { RenderTarget *rt = render_target_owner.getornull(p_render_target); ERR_FAIL_COND_V(!rt, RID()); @@ -6870,7 +7310,7 @@ RID RendererStorageRD::render_target_get_sdf_texture(RID p_render_target) { Vector<uint8_t> pv; pv.resize(16 * 4); - zeromem(pv.ptrw(), 16 * 4); + memset(pv.ptrw(), 0, 16 * 4); Vector<Vector<uint8_t>> vpv; rt->sdf_buffer_read = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); @@ -6923,7 +7363,7 @@ void RendererStorageRD::_render_target_allocate_sdf(RenderTarget *rt) { rt->process_size.x = MAX(rt->process_size.x, 1); rt->process_size.y = MAX(rt->process_size.y, 1); - tformat.format = RD::DATA_FORMAT_R16G16_UINT; + tformat.format = RD::DATA_FORMAT_R16G16_SINT; tformat.width = rt->process_size.width; tformat.height = rt->process_size.height; tformat.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; @@ -6931,7 +7371,7 @@ void RendererStorageRD::_render_target_allocate_sdf(RenderTarget *rt) { rt->sdf_buffer_process[0] = RD::get_singleton()->texture_create(tformat, RD::TextureView()); rt->sdf_buffer_process[1] = RD::get_singleton()->texture_create(tformat, RD::TextureView()); - tformat.format = RD::DATA_FORMAT_R16_UNORM; + tformat.format = RD::DATA_FORMAT_R16_SNORM; tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rt->sdf_buffer_read = RD::get_singleton()->texture_create(tformat, RD::TextureView()); @@ -7095,7 +7535,7 @@ void RendererStorageRD::render_target_copy_to_back_buffer(RID p_render_target, c //single texture copy for backbuffer //RD::get_singleton()->texture_copy(rt->color, rt->backbuffer_mipmap0, Vector3(region.position.x, region.position.y, 0), Vector3(region.position.x, region.position.y, 0), Vector3(region.size.x, region.size.y, 1), 0, 0, 0, 0, true); - effects.copy_to_rect(rt->color, rt->backbuffer_mipmap0, region, false, false, false, true, true); + effects->copy_to_rect(rt->color, rt->backbuffer_mipmap0, region, false, false, false, true, true); if (!p_gen_mipmaps) { return; @@ -7111,7 +7551,7 @@ void RendererStorageRD::render_target_copy_to_back_buffer(RID p_render_target, c region.size.y = MAX(1, region.size.y >> 1); const RenderTarget::BackbufferMipmap &mm = rt->backbuffer_mipmaps[i]; - effects.gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); + effects->gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); prev_texture = mm.mipmap; } } @@ -7134,7 +7574,7 @@ void RendererStorageRD::render_target_clear_back_buffer(RID p_render_target, con } //single texture copy for backbuffer - effects.set_color(rt->backbuffer_mipmap0, p_color, region, true); + effects->set_color(rt->backbuffer_mipmap0, p_color, region, true); } void RendererStorageRD::render_target_gen_back_buffer_mipmaps(RID p_render_target, const Rect2i &p_region) { @@ -7164,7 +7604,7 @@ void RendererStorageRD::render_target_gen_back_buffer_mipmaps(RID p_render_targe region.size.y = MAX(1, region.size.y >> 1); const RenderTarget::BackbufferMipmap &mm = rt->backbuffer_mipmaps[i]; - effects.gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); + effects->gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); prev_texture = mm.mipmap; } } @@ -7207,8 +7647,8 @@ void RendererStorageRD::base_update_dependency(RID p_base, DependencyTracker *p_ } else if (decal_owner.owns(p_base)) { Decal *decal = decal_owner.getornull(p_base); p_instance->update_dependency(&decal->dependency); - } else if (gi_probe_owner.owns(p_base)) { - GIProbe *gip = gi_probe_owner.getornull(p_base); + } else if (voxel_gi_owner.owns(p_base)) { + VoxelGI *gip = voxel_gi_owner.getornull(p_base); p_instance->update_dependency(&gip->dependency); } else if (lightmap_owner.owns(p_base)) { Lightmap *lm = lightmap_owner.getornull(p_base); @@ -7222,6 +7662,9 @@ void RendererStorageRD::base_update_dependency(RID p_base, DependencyTracker *p_ } else if (particles_collision_owner.owns(p_base)) { ParticlesCollision *pc = particles_collision_owner.getornull(p_base); p_instance->update_dependency(&pc->dependency); + } else if (visibility_notifier_owner.owns(p_base)) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_base); + p_instance->update_dependency(&vn->dependency); } } @@ -7245,8 +7688,8 @@ RS::InstanceType RendererStorageRD::get_base_type(RID p_rid) const { if (decal_owner.owns(p_rid)) { return RS::INSTANCE_DECAL; } - if (gi_probe_owner.owns(p_rid)) { - return RS::INSTANCE_GI_PROBE; + if (voxel_gi_owner.owns(p_rid)) { + return RS::INSTANCE_VOXEL_GI; } if (light_owner.owns(p_rid)) { return RS::INSTANCE_LIGHT; @@ -7260,6 +7703,9 @@ RS::InstanceType RendererStorageRD::get_base_type(RID p_rid) const { if (particles_collision_owner.owns(p_rid)) { return RS::INSTANCE_PARTICLES_COLLISION; } + if (visibility_notifier_owner.owns(p_rid)) { + return RS::INSTANCE_VISIBLITY_NOTIFIER; + } return RS::INSTANCE_NONE; } @@ -7357,7 +7803,7 @@ void RendererStorageRD::_update_decal_atlas() { v_offsetsv.resize(base_size); int *v_offsets = v_offsetsv.ptrw(); - zeromem(v_offsets, sizeof(int) * base_size); + memset(v_offsets, 0, sizeof(int) * base_size); int max_height = 0; @@ -7479,14 +7925,14 @@ void RendererStorageRD::_update_decal_atlas() { while ((K = decal_atlas.textures.next(K))) { DecalAtlas::Texture *t = decal_atlas.textures.getptr(*K); Texture *src_tex = texture_owner.getornull(*K); - effects.copy_to_atlas_fb(src_tex->rd_texture, mm.fb, t->uv_rect, draw_list, false, t->panorama_to_dp_users > 0); + effects->copy_to_atlas_fb(src_tex->rd_texture, mm.fb, t->uv_rect, draw_list, false, t->panorama_to_dp_users > 0); } RD::get_singleton()->draw_list_end(); prev_texture = mm.texture; } else { - effects.copy_to_fb_rect(prev_texture, mm.fb, Rect2i(Point2i(), mm.size)); + effects->copy_to_fb_rect(prev_texture, mm.fb, Rect2i(Point2i(), mm.size)); prev_texture = mm.texture; } } else { @@ -7770,7 +8216,7 @@ void RendererStorageRD::_global_variable_store_in_buffer(int32_t p_index, RS::Gl } break; case RS::GLOBAL_VAR_TYPE_TRANSFORM: { GlobalVariables::Value *bv = &global_variables.buffer_values[p_index]; - Transform v = p_value; + Transform3D v = p_value; bv[0].x = v.basis.elements[0][0]; bv[0].y = v.basis.elements[1][0]; bv[0].z = v.basis.elements[2][0]; @@ -7906,6 +8352,9 @@ void RendererStorageRD::global_variable_set_override(const StringName &p_name, c if (!global_variables.variables.has(p_name)) { return; //variable may not exist } + + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); + GlobalVariables::Variable &gv = global_variables.variables[p_name]; gv.override = p_value; @@ -7921,7 +8370,6 @@ void RendererStorageRD::global_variable_set_override(const StringName &p_name, c _global_variable_mark_buffer_dirty(gv.buffer_index, gv.buffer_elements); } else { //texture - //texture for (Set<RID>::Element *E = gv.texture_materials.front(); E; E = E->next()) { Material *material = material_owner.getornull(E->get()); ERR_CONTINUE(!material); @@ -7962,10 +8410,10 @@ void RendererStorageRD::global_variables_load_settings(bool p_load_textures) { List<PropertyInfo> settings; ProjectSettings::get_singleton()->get_property_list(&settings); - for (List<PropertyInfo>::Element *E = settings.front(); E; E = E->next()) { - if (E->get().name.begins_with("shader_globals/")) { - StringName name = E->get().name.get_slice("/", 1); - Dictionary d = ProjectSettings::get_singleton()->get(E->get().name); + for (const PropertyInfo &E : settings) { + if (E.name.begins_with("shader_globals/")) { + StringName name = E.name.get_slice("/", 1); + Dictionary d = ProjectSettings::get_singleton()->get(E.name); ERR_CONTINUE(!d.has("type")); ERR_CONTINUE(!d.has("value")); @@ -8114,7 +8562,7 @@ void RendererStorageRD::_update_global_variables() { if (total_regions / global_variables.buffer_dirty_region_count <= 4) { // 25% of regions dirty, just update all buffer RD::get_singleton()->buffer_update(global_variables.buffer, 0, sizeof(GlobalVariables::Value) * global_variables.buffer_size, global_variables.buffer_values); - zeromem(global_variables.buffer_dirty_regions, sizeof(bool) * total_regions); + memset(global_variables.buffer_dirty_regions, 0, sizeof(bool) * total_regions); } else { uint32_t region_byte_size = sizeof(GlobalVariables::Value) * GlobalVariables::BUFFER_DIRTY_REGION_SIZE; @@ -8133,8 +8581,8 @@ void RendererStorageRD::_update_global_variables() { if (global_variables.must_update_buffer_materials) { // only happens in the case of a buffer variable added or removed, // so not often. - for (List<RID>::Element *E = global_variables.materials_using_buffer.front(); E; E = E->next()) { - Material *material = material_owner.getornull(E->get()); + for (const RID &E : global_variables.materials_using_buffer) { + Material *material = material_owner.getornull(E); ERR_CONTINUE(!material); //wtf _material_queue_update(material, true, false); @@ -8146,8 +8594,8 @@ void RendererStorageRD::_update_global_variables() { if (global_variables.must_update_texture_materials) { // only happens in the case of a buffer variable added or removed, // so not often. - for (List<RID>::Element *E = global_variables.materials_using_texture.front(); E; E = E->next()) { - Material *material = material_owner.getornull(E->get()); + for (const RID &E : global_variables.materials_using_texture) { + Material *material = material_owner.getornull(E); ERR_CONTINUE(!material); //wtf _material_queue_update(material, false, true); @@ -8194,6 +8642,7 @@ bool RendererStorageRD::free(RID p_rid) { if (texture_owner.owns(p_rid)) { Texture *t = texture_owner.getornull(p_rid); + ERR_FAIL_COND_V(!t, false); ERR_FAIL_COND_V(t->is_render_target, false); if (RD::get_singleton()->texture_is_valid(t->rd_texture_srgb)) { @@ -8230,8 +8679,6 @@ bool RendererStorageRD::free(RID p_rid) { texture_owner.free(p_rid); } else if (canvas_texture_owner.owns(p_rid)) { - CanvasTexture *ct = canvas_texture_owner.getornull(p_rid); - memdelete(ct); canvas_texture_owner.free(p_rid); } else if (shader_owner.owns(p_rid)) { Shader *shader = shader_owner.getornull(p_rid); @@ -8247,9 +8694,6 @@ bool RendererStorageRD::free(RID p_rid) { } else if (material_owner.owns(p_rid)) { Material *material = material_owner.getornull(p_rid); - if (material->update_requested) { - _update_queued_materials(); - } material_set_shader(p_rid, RID()); //clean up shader material->dependency.deleted_notify(p_rid); @@ -8277,7 +8721,6 @@ bool RendererStorageRD::free(RID p_rid) { mi->I = nullptr; mesh_instance_owner.free(p_rid); - memdelete(mi); } else if (multimesh_owner.owns(p_rid)) { _update_dirty_multimeshes(); @@ -8304,11 +8747,11 @@ bool RendererStorageRD::free(RID p_rid) { } decal->dependency.deleted_notify(p_rid); decal_owner.free(p_rid); - } else if (gi_probe_owner.owns(p_rid)) { - gi_probe_allocate_data(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate - GIProbe *gi_probe = gi_probe_owner.getornull(p_rid); - gi_probe->dependency.deleted_notify(p_rid); - gi_probe_owner.free(p_rid); + } else if (voxel_gi_owner.owns(p_rid)) { + voxel_gi_allocate_data(p_rid, Transform3D(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate + VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_rid); + voxel_gi->dependency.deleted_notify(p_rid); + voxel_gi_owner.free(p_rid); } else if (lightmap_owner.owns(p_rid)) { lightmap_set_textures(p_rid, RID(), false); Lightmap *lightmap = lightmap_owner.getornull(p_rid); @@ -8323,9 +8766,10 @@ bool RendererStorageRD::free(RID p_rid) { light_owner.free(p_rid); } else if (particles_owner.owns(p_rid)) { + update_particles(); Particles *particles = particles_owner.getornull(p_rid); - _particles_free_data(particles); particles->dependency.deleted_notify(p_rid); + _particles_free_data(particles); particles_owner.free(p_rid); } else if (particles_collision_owner.owns(p_rid)) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_rid); @@ -8335,6 +8779,10 @@ bool RendererStorageRD::free(RID p_rid) { } particles_collision->dependency.deleted_notify(p_rid); particles_collision_owner.free(p_rid); + } else if (visibility_notifier_owner.owns(p_rid)) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_rid); + vn->dependency.deleted_notify(p_rid); + visibility_notifier_owner.free(p_rid); } else if (particles_collision_instance_owner.owns(p_rid)) { particles_collision_instance_owner.free(p_rid); } else if (render_target_owner.owns(p_rid)) { @@ -8356,8 +8804,13 @@ bool RendererStorageRD::free(RID p_rid) { return true; } +void RendererStorageRD::init_effects(bool p_prefer_raster_effects) { + effects = memnew(EffectsRD(p_prefer_raster_effects)); +} + EffectsRD *RendererStorageRD::get_effects() { - return &effects; + ERR_FAIL_NULL_V_MSG(effects, nullptr, "Effects haven't been initialised yet."); + return effects; } void RendererStorageRD::capture_timestamps_begin() { @@ -8388,6 +8841,29 @@ String RendererStorageRD::get_captured_timestamp_name(uint32_t p_index) const { return RD::get_singleton()->get_captured_timestamp_name(p_index); } +void RendererStorageRD::update_memory_info() { + texture_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TEXTURES); + buffer_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_BUFFERS); + total_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TOTAL); +} +uint64_t RendererStorageRD::get_rendering_info(RS::RenderingInfo p_info) { + if (p_info == RS::RENDERING_INFO_TEXTURE_MEM_USED) { + return texture_mem_cache; + } else if (p_info == RS::RENDERING_INFO_BUFFER_MEM_USED) { + return buffer_mem_cache; + } else if (p_info == RS::RENDERING_INFO_VIDEO_MEM_USED) { + return total_mem_cache; + } + return 0; +} + +String RendererStorageRD::get_video_adapter_name() const { + return RenderingDevice::get_singleton()->get_device_name(); +} +String RendererStorageRD::get_video_adapter_vendor() const { + return RenderingDevice::get_singleton()->get_device_vendor_name(); +} + RendererStorageRD *RendererStorageRD::base_singleton = nullptr; RendererStorageRD::RendererStorageRD() { @@ -8402,13 +8878,12 @@ RendererStorageRD::RendererStorageRD() { global_variables.buffer_size = GLOBAL_GET("rendering/limits/global_shader_variables/buffer_size"); global_variables.buffer_size = MAX(4096, global_variables.buffer_size); global_variables.buffer_values = memnew_arr(GlobalVariables::Value, global_variables.buffer_size); - zeromem(global_variables.buffer_values, sizeof(GlobalVariables::Value) * global_variables.buffer_size); + memset(global_variables.buffer_values, 0, sizeof(GlobalVariables::Value) * global_variables.buffer_size); global_variables.buffer_usage = memnew_arr(GlobalVariables::ValueUsage, global_variables.buffer_size); global_variables.buffer_dirty_regions = memnew_arr(bool, global_variables.buffer_size / GlobalVariables::BUFFER_DIRTY_REGION_SIZE); - zeromem(global_variables.buffer_dirty_regions, sizeof(bool) * global_variables.buffer_size / GlobalVariables::BUFFER_DIRTY_REGION_SIZE); + memset(global_variables.buffer_dirty_regions, 0, sizeof(bool) * global_variables.buffer_size / GlobalVariables::BUFFER_DIRTY_REGION_SIZE); global_variables.buffer = RD::get_singleton()->storage_buffer_create(sizeof(GlobalVariables::Value) * global_variables.buffer_size); - material_update_list = nullptr; { //create default textures RD::TextureFormat tformat; @@ -8653,26 +9128,42 @@ RendererStorageRD::RendererStorageRD() { } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS: { sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; - sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS: { sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC: { sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; - sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } sampler_state.use_anisotropy = true; sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } sampler_state.use_anisotropy = true; sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); @@ -8819,16 +9310,6 @@ RendererStorageRD::RendererStorageRD() { } } - { - Vector<String> sdf_versions; - sdf_versions.push_back(""); //one only - giprobe_sdf_shader.initialize(sdf_versions); - giprobe_sdf_shader_version = giprobe_sdf_shader.version_create(); - giprobe_sdf_shader.version_set_compute_code(giprobe_sdf_shader_version, "", "", "", Vector<String>()); - giprobe_sdf_shader_version_shader = giprobe_sdf_shader.version_get_shader(giprobe_sdf_shader_version, 0); - giprobe_sdf_shader_pipeline = RD::get_singleton()->compute_pipeline_create(giprobe_sdf_shader_version_shader); - } - using_lightmap_array = true; // high end if (using_lightmap_array) { uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); @@ -8863,14 +9344,17 @@ RendererStorageRD::RendererStorageRD() { actions.renames["COLOR"] = "PARTICLE.color"; actions.renames["VELOCITY"] = "PARTICLE.velocity"; //actions.renames["MASS"] = "mass"; ? - actions.renames["ACTIVE"] = "PARTICLE.is_active"; + actions.renames["ACTIVE"] = "particle_active"; actions.renames["RESTART"] = "restart"; actions.renames["CUSTOM"] = "PARTICLE.custom"; actions.renames["TRANSFORM"] = "PARTICLE.xform"; actions.renames["TIME"] = "FRAME.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); actions.renames["LIFETIME"] = "params.lifetime"; actions.renames["DELTA"] = "local_delta"; - actions.renames["NUMBER"] = "particle"; + actions.renames["NUMBER"] = "particle_number"; actions.renames["INDEX"] = "index"; //actions.renames["GRAVITY"] = "current_gravity"; actions.renames["EMISSION_TRANSFORM"] = "FRAME.emission_transform"; @@ -8913,7 +9397,13 @@ RendererStorageRD::RendererStorageRD() { // default material and shader for particles shader particles_shader.default_shader = shader_allocate(); shader_initialize(particles_shader.default_shader); - shader_set_code(particles_shader.default_shader, "shader_type particles; void compute() { COLOR = vec4(1.0); } \n"); + shader_set_code(particles_shader.default_shader, R"( +shader_type particles; + +void process() { + COLOR = vec4(1.0); +} +)"); particles_shader.default_material = material_allocate(); material_initialize(particles_shader.default_material); material_set_shader(particles_shader.default_material, particles_shader.default_shader); @@ -8960,6 +9450,7 @@ RendererStorageRD::RendererStorageRD() { { Vector<String> copy_modes; copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n"); + copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n#define MODE_2D\n"); copy_modes.push_back("\n#define MODE_FILL_SORT_BUFFER\n#define USE_SORT_BUFFER\n"); copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n#define USE_SORT_BUFFER\n"); @@ -9038,7 +9529,6 @@ RendererStorageRD::~RendererStorageRD() { RD::get_singleton()->free(mesh_default_rd_buffers[i]); } - giprobe_sdf_shader.version_free(giprobe_sdf_shader_version); particles_shader.copy_shader.version_free(particles_shader.copy_shader_version); rt_sdf.shader.version_free(rt_sdf.shader_version); @@ -9056,4 +9546,9 @@ RendererStorageRD::~RendererStorageRD() { if (decal_atlas.texture.is_valid()) { RD::get_singleton()->free(decal_atlas.texture); } + + if (effects) { + memdelete(effects); + effects = NULL; + } } diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index 6405bb75b0..02395a884f 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -38,15 +38,15 @@ #include "servers/rendering/renderer_rd/effects_rd.h" #include "servers/rendering/renderer_rd/shader_compiler_rd.h" #include "servers/rendering/renderer_rd/shaders/canvas_sdf.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/particles.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/particles_copy.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/skeleton.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl.gen.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering/rendering_device.h" class RendererStorageRD : public RendererStorage { public: - static _FORCE_INLINE_ void store_transform(const Transform &p_mtx, float *p_array) { + static _FORCE_INLINE_ void store_transform(const Transform3D &p_mtx, float *p_array) { p_array[0] = p_mtx.basis.elements[0][0]; p_array[1] = p_mtx.basis.elements[1][0]; p_array[2] = p_mtx.basis.elements[2][0]; @@ -95,7 +95,7 @@ public: p_array[11] = 0; } - static _FORCE_INLINE_ void store_transform_transposed_3x4(const Transform &p_mtx, float *p_array) { + static _FORCE_INLINE_ void store_transform_transposed_3x4(const Transform3D &p_mtx, float *p_array) { p_array[0] = p_mtx.basis.elements[0][0]; p_array[1] = p_mtx.basis.elements[0][1]; p_array[2] = p_mtx.basis.elements[0][2]; @@ -155,9 +155,13 @@ public: virtual void set_render_priority(int p_priority) = 0; virtual void set_next_pass(RID p_pass) = 0; - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) = 0; + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) = 0; virtual ~MaterialData(); + //to be used internally by update_parameters, in the most common configuration of material parameters + bool update_parameters_uniform_set(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const Map<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, const Map<StringName, RID> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void free_parameters_uniform_set(RID p_uniform_set); + private: friend class RendererStorageRD; RID self; @@ -165,8 +169,14 @@ public: List<RID>::Element *global_texture_E = nullptr; uint64_t global_textures_pass = 0; Map<StringName, uint64_t> used_global_textures; + + //internally by update_parameters_uniform_set + Vector<uint8_t> ubo_data; + RID uniform_buffer; + Vector<RID> texture_cache; }; typedef MaterialData *(*MaterialDataRequestFunction)(ShaderData *); + static void _material_uniform_set_erased(const RID &p_set, void *p_material); enum DefaultRDTexture { DEFAULT_RD_TEXTURE_WHITE, @@ -221,7 +231,7 @@ private: ~CanvasTexture(); }; - RID_PtrOwner<CanvasTexture, true> canvas_texture_owner; + RID_Owner<CanvasTexture, true> canvas_texture_owner; /* TEXTURE API */ struct Texture { @@ -373,25 +383,28 @@ private: struct Material { RID self; - MaterialData *data; - Shader *shader; + MaterialData *data = nullptr; + Shader *shader = nullptr; //shortcut to shader data and type - ShaderType shader_type; + ShaderType shader_type = SHADER_TYPE_MAX; uint32_t shader_id = 0; - bool update_requested; - bool uniform_dirty; - bool texture_dirty; - Material *update_next; + bool uniform_dirty = false; + bool texture_dirty = false; Map<StringName, Variant> params; - int32_t priority; + int32_t priority = 0; RID next_pass; + SelfList<Material> update_element; + Dependency dependency; + + Material() : + update_element(this) {} }; MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; mutable RID_Owner<Material, true> material_owner; - Material *material_update_list; + SelfList<Material>::List material_update_list; void _material_queue_update(Material *material, bool p_uniform, bool p_texture); void _update_queued_materials(); @@ -434,6 +447,7 @@ private: struct LOD { float edge_length = 0.0; + uint32_t index_count = 0; RID index_buffer; RID index_array; }; @@ -513,7 +527,7 @@ private: void _mesh_instance_clear(MeshInstance *mi); void _mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface); - mutable RID_PtrOwner<MeshInstance> mesh_instance_owner; + mutable RID_Owner<MeshInstance> mesh_instance_owner; SelfList<MeshInstance>::List dirty_mesh_instance_weights; SelfList<MeshInstance>::List dirty_mesh_instance_arrays; @@ -580,6 +594,7 @@ private: RID buffer; //storage buffer RID uniform_set_3d; + RID uniform_set_2d; bool dirty = false; MultiMesh *dirty_list = nullptr; @@ -637,7 +652,9 @@ private: COLLISION_TYPE_SPHERE, COLLISION_TYPE_BOX, COLLISION_TYPE_SDF, - COLLISION_TYPE_HEIGHT_FIELD + COLLISION_TYPE_HEIGHT_FIELD, + COLLISION_TYPE_2D_SDF, + }; struct Collider { @@ -646,7 +663,7 @@ private: uint32_t type; uint32_t texture_index; //texture index for vector field - float scale; + real_t scale; uint32_t pad[2]; }; @@ -655,11 +672,16 @@ private: float prev_system_phase; uint32_t cycle; - float explosiveness; - float randomness; + real_t explosiveness; + real_t randomness; float time; float delta; + uint32_t frame; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + uint32_t random_seed; uint32_t attractor_count; uint32_t collider_count; @@ -691,23 +713,37 @@ private: }; struct Particles { + RS::ParticlesMode mode = RS::PARTICLES_MODE_3D; bool inactive = true; - float inactive_time = 0.0; + double inactive_time = 0.0; bool emitting = false; bool one_shot = false; int amount = 0; - float lifetime = 1.0; - float pre_process_time = 0.0; - float explosiveness = 0.0; - float randomness = 0.0; + double lifetime = 1.0; + double pre_process_time = 0.0; + real_t explosiveness = 0.0; + real_t randomness = 0.0; bool restart_request = false; AABB custom_aabb = AABB(Vector3(-4, -4, -4), Vector3(8, 8, 8)); bool use_local_coords = true; + bool has_collision_cache = false; + + bool has_sdf_collision = false; + Transform2D sdf_collision_transform; + Rect2 sdf_collision_to_screen; + RID sdf_collision_texture; + RID process_material; + uint32_t frame_counter = 0; + RS::ParticlesTransformAlign transform_align = RS::PARTICLES_TRANSFORM_ALIGN_DISABLED; RS::ParticlesDrawOrder draw_order = RS::PARTICLES_DRAW_ORDER_INDEX; Vector<RID> draw_passes; + Vector<Transform3D> trail_bind_poses; + bool trail_bind_poses_dirty = false; + RID trail_bind_pose_buffer; + RID trail_bind_pose_uniform_set; RID particle_buffer; RID particle_instance_buffer; @@ -730,25 +766,26 @@ private: RID sub_emitter; - float phase = 0.0; - float prev_phase = 0.0; + double phase = 0.0; + double prev_phase = 0.0; uint64_t prev_ticks = 0; uint32_t random_seed = 0; uint32_t cycle_number = 0; - float speed_scale = 1.0; + double speed_scale = 1.0; - int fixed_fps = 0; + int fixed_fps = 30; + bool interpolate = true; bool fractional_delta = false; - float frame_remainder = 0; - float collision_base_size = 0.01; + double frame_remainder = 0; + real_t collision_base_size = 0.01; bool clear = true; bool force_sub_emit = false; - Transform emission_transform; + Transform3D emission_transform; Vector<uint8_t> emission_buffer_data; @@ -759,12 +796,19 @@ private: Dependency dependency; - ParticlesFrameParams frame_params; + double trail_length = 1.0; + bool trails_enabled = false; + LocalVector<ParticlesFrameParams> frame_history; + LocalVector<ParticlesFrameParams> trail_params; + + Particles() { + } }; - void _particles_process(Particles *p_particles, float p_delta); + void _particles_process(Particles *p_particles, double p_delta); void _particles_allocate_emission_buffer(Particles *particles); void _particles_free_data(Particles *particles); + void _particles_update_buffers(Particles *particles); struct ParticlesShader { struct PushConstant { @@ -776,7 +820,7 @@ private: uint32_t use_fractional_delta; uint32_t sub_emitter_mode; uint32_t can_emit; - uint32_t pad; + uint32_t trail_pass; }; ParticlesShaderRD shader; @@ -791,10 +835,24 @@ private: struct CopyPushConstant { float sort_direction[3]; uint32_t total_particles; + + uint32_t trail_size; + uint32_t trail_total; + float frame_delta; + float frame_remainder; + + float align_up[3]; + uint32_t align_mode; + + uint32_t order_by_lifetime; + uint32_t lifetime_split; + uint32_t lifetime_reverse; + uint32_t pad; }; enum { COPY_MODE_FILL_INSTANCES, + COPY_MODE_FILL_INSTANCES_2D, COPY_MODE_FILL_SORT_BUFFER, COPY_MODE_FILL_INSTANCES_WITH_SORT_BUFFER, COPY_MODE_MAX, @@ -804,6 +862,8 @@ private: RID copy_shader_version; RID copy_pipelines[COPY_MODE_MAX]; + LocalVector<float> pose_update_buffer; + } particles_shader; Particles *particle_update_list = nullptr; @@ -811,6 +871,7 @@ private: struct ParticlesShaderData : public ShaderData { bool valid; RID version; + bool uses_collision = false; //PipelineCacheRD pipelines[SKY_VERSION_MAX]; Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; @@ -847,17 +908,14 @@ private: } struct ParticlesMaterialData : public MaterialData { - uint64_t last_frame; - ParticlesShaderData *shader_data; - RID uniform_buffer; + uint64_t last_frame = 0; + ParticlesShaderData *shader_data = nullptr; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; - bool uniform_set_updated; + bool uniform_set_updated = false; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~ParticlesMaterialData(); }; @@ -894,12 +952,23 @@ private: struct ParticlesCollisionInstance { RID collision; - Transform transform; + Transform3D transform; bool active = false; }; mutable RID_Owner<ParticlesCollisionInstance> particles_collision_instance_owner; + /* visibility_notifier */ + + struct VisibilityNotifier { + AABB aabb; + Callable enter_callback; + Callable exit_callback; + Dependency dependency; + }; + + mutable RID_Owner<VisibilityNotifier> visibility_notifier_owner; + /* Skeleton */ struct Skeleton { @@ -944,7 +1013,6 @@ private: uint32_t cull_mask = 0xFFFFFFFF; RS::LightOmniShadowMode omni_shadow_mode = RS::LIGHT_OMNI_SHADOW_DUAL_PARABOLOID; RS::LightDirectionalShadowMode directional_shadow_mode = RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL; - RS::LightDirectionalShadowDepthRangeMode directional_range_mode = RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE; bool directional_blend_splits = false; bool directional_sky_only = false; uint64_t version = 0; @@ -998,9 +1066,9 @@ private: mutable RID_Owner<Decal, true> decal_owner; - /* GI PROBE */ + /* VOXEL GI */ - struct GIProbe { + struct VoxelGI { RID octree_buffer; RID data_buffer; RID sdf_texture; @@ -1012,14 +1080,12 @@ private: int cell_count = 0; - Transform to_cell_xform; + Transform3D to_cell_xform; AABB bounds; Vector3i octree_size; float dynamic_range = 4.0; float energy = 1.0; - float ao = 0.0; - float ao_size = 0.5; float bias = 1.4; float normal_bias = 0.0; float propagation = 0.7; @@ -1034,12 +1100,7 @@ private: Dependency dependency; }; - GiprobeSdfShaderRD giprobe_sdf_shader; - RID giprobe_sdf_shader_version; - RID giprobe_sdf_shader_version_shader; - RID giprobe_sdf_shader_pipeline; - - mutable RID_Owner<GIProbe, true> gi_probe_owner; + mutable RID_Owner<VoxelGI, true> voxel_gi_owner; /* REFLECTION PROBE */ @@ -1078,6 +1139,7 @@ private: struct RenderTarget { Size2i size; + uint32_t view_count; RID framebuffer; RID color; @@ -1088,6 +1150,8 @@ private: bool flags[RENDER_TARGET_FLAG_MAX]; + bool sdf_enabled = false; + RID backbuffer; //used for effects RID backbuffer_fb; RID backbuffer_mipmap0; @@ -1221,7 +1285,7 @@ private: void _update_global_variables(); /* EFFECTS */ - EffectsRD effects; + EffectsRD *effects = NULL; public: virtual bool can_create_resources_async() const; @@ -1237,7 +1301,6 @@ public: virtual void _texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate); - virtual void texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer = 0); //mostly used for video and streaming virtual void texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0); virtual void texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data); virtual void texture_proxy_update(RID p_texture, RID p_proxy_to); @@ -1396,7 +1459,9 @@ public: virtual void mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode); virtual RS::BlendShapeMode mesh_get_blend_shape_mode(RID p_mesh) const; - virtual void mesh_surface_update_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); virtual void mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material); virtual RID mesh_surface_get_material(RID p_mesh, int p_surface) const; @@ -1465,10 +1530,18 @@ public: return s->lod_count > 0; } - _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_vertices_drawn_count(void *p_surface) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + return s->index_count ? s->index_count : s->vertex_count; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold, uint32_t *r_index_count = nullptr) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); int32_t current_lod = -1; + if (r_index_count) { + *r_index_count = s->index_count; + } for (uint32_t i = 0; i < s->lod_count; i++) { float screen_size = s->lods[i].edge_length * p_model_scale / p_distance_threshold; if (screen_size > p_lod_threshold) { @@ -1479,6 +1552,9 @@ public: if (current_lod == -1) { return 0; } else { + if (r_index_count) { + *r_index_count = s->lods[current_lod].index_count; + } return current_lod + 1; } } @@ -1612,14 +1688,14 @@ public: int multimesh_get_instance_count(RID p_multimesh) const; void multimesh_set_mesh(RID p_multimesh, RID p_mesh); - void multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform &p_transform); + void multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform3D &p_transform); void multimesh_instance_set_transform_2d(RID p_multimesh, int p_index, const Transform2D &p_transform); void multimesh_instance_set_color(RID p_multimesh, int p_index, const Color &p_color); void multimesh_instance_set_custom_data(RID p_multimesh, int p_index, const Color &p_color); RID multimesh_get_mesh(RID p_multimesh) const; - Transform multimesh_instance_get_transform(RID p_multimesh, int p_index) const; + Transform3D multimesh_instance_get_transform(RID p_multimesh, int p_index) const; Transform2D multimesh_instance_get_transform_2d(RID p_multimesh, int p_index) const; Color multimesh_instance_get_color(RID p_multimesh, int p_index) const; Color multimesh_instance_get_custom_data(RID p_multimesh, int p_index) const; @@ -1670,23 +1746,20 @@ public: return multimesh->uniform_set_3d; } - /* IMMEDIATE API */ - - RID immediate_allocate() { return RID(); } - void immediate_initialize(RID p_immediate) {} + _FORCE_INLINE_ RID multimesh_get_2d_uniform_set(RID p_multimesh, RID p_shader, uint32_t p_set) const { + MultiMesh *multimesh = multimesh_owner.getornull(p_multimesh); + if (!multimesh->uniform_set_2d.is_valid()) { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.ids.push_back(multimesh->buffer); + uniforms.push_back(u); + multimesh->uniform_set_2d = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); + } - virtual void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} - virtual void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} - virtual void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} - virtual void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} - virtual void immediate_color(RID p_immediate, const Color &p_color) {} - virtual void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} - virtual void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} - virtual void immediate_end(RID p_immediate) {} - virtual void immediate_clear(RID p_immediate) {} - virtual void immediate_set_material(RID p_immediate, RID p_material) {} - virtual RID immediate_get_material(RID p_immediate) const { return RID(); } - virtual AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } + return multimesh->uniform_set_2d; + } /* SKELETON API */ @@ -1695,10 +1768,10 @@ public: void skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton = false); void skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform); - void skeleton_set_world_transform(RID p_skeleton, bool p_enable, const Transform &p_world_transform); + void skeleton_set_world_transform(RID p_skeleton, bool p_enable, const Transform3D &p_world_transform); int skeleton_get_bone_count(RID p_skeleton) const; - void skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform &p_transform); - Transform skeleton_bone_get_transform(RID p_skeleton, int p_bone) const; + void skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform3D &p_transform); + Transform3D skeleton_bone_get_transform(RID p_skeleton, int p_bone) const; void skeleton_bone_set_transform_2d(RID p_skeleton, int p_bone, const Transform2D &p_transform); Transform2D skeleton_bone_get_transform_2d(RID p_skeleton, int p_bone) const; @@ -1756,8 +1829,6 @@ public: bool light_directional_get_blend_splits(RID p_light) const; void light_directional_set_sky_only(RID p_light, bool p_sky_only); bool light_directional_is_sky_only(RID p_light) const; - void light_directional_set_shadow_depth_range_mode(RID p_light, RS::LightDirectionalShadowDepthRangeMode p_range_mode); - RS::LightDirectionalShadowDepthRangeMode light_directional_get_shadow_depth_range_mode(RID p_light) const; RS::LightDirectionalShadowMode light_directional_get_shadow_mode(RID p_light); RS::LightOmniShadowMode light_omni_get_shadow_mode(RID p_light); @@ -1812,6 +1883,13 @@ public: return light->shadow; } + _FORCE_INLINE_ bool light_has_projector(RID p_light) const { + const Light *light = light_owner.getornull(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return texture_owner.owns(light->projector); + } + _FORCE_INLINE_ bool light_is_negative(RID p_light) const { const Light *light = light_owner.getornull(p_light); ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); @@ -1955,59 +2033,53 @@ public: virtual AABB decal_get_aabb(RID p_decal) const; - /* GI PROBE API */ - - RID gi_probe_allocate(); - void gi_probe_initialize(RID p_gi_probe); - - void gi_probe_allocate_data(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts); + /* VOXEL GI API */ - AABB gi_probe_get_bounds(RID p_gi_probe) const; - Vector3i gi_probe_get_octree_size(RID p_gi_probe) const; - Vector<uint8_t> gi_probe_get_octree_cells(RID p_gi_probe) const; - Vector<uint8_t> gi_probe_get_data_cells(RID p_gi_probe) const; - Vector<uint8_t> gi_probe_get_distance_field(RID p_gi_probe) const; + RID voxel_gi_allocate(); + void voxel_gi_initialize(RID p_voxel_gi); - Vector<int> gi_probe_get_level_counts(RID p_gi_probe) const; - Transform gi_probe_get_to_cell_xform(RID p_gi_probe) const; + void voxel_gi_allocate_data(RID p_voxel_gi, const Transform3D &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts); - void gi_probe_set_dynamic_range(RID p_gi_probe, float p_range); - float gi_probe_get_dynamic_range(RID p_gi_probe) const; + AABB voxel_gi_get_bounds(RID p_voxel_gi) const; + Vector3i voxel_gi_get_octree_size(RID p_voxel_gi) const; + Vector<uint8_t> voxel_gi_get_octree_cells(RID p_voxel_gi) const; + Vector<uint8_t> voxel_gi_get_data_cells(RID p_voxel_gi) const; + Vector<uint8_t> voxel_gi_get_distance_field(RID p_voxel_gi) const; - void gi_probe_set_propagation(RID p_gi_probe, float p_range); - float gi_probe_get_propagation(RID p_gi_probe) const; + Vector<int> voxel_gi_get_level_counts(RID p_voxel_gi) const; + Transform3D voxel_gi_get_to_cell_xform(RID p_voxel_gi) const; - void gi_probe_set_energy(RID p_gi_probe, float p_energy); - float gi_probe_get_energy(RID p_gi_probe) const; + void voxel_gi_set_dynamic_range(RID p_voxel_gi, float p_range); + float voxel_gi_get_dynamic_range(RID p_voxel_gi) const; - void gi_probe_set_ao(RID p_gi_probe, float p_ao); - float gi_probe_get_ao(RID p_gi_probe) const; + void voxel_gi_set_propagation(RID p_voxel_gi, float p_range); + float voxel_gi_get_propagation(RID p_voxel_gi) const; - void gi_probe_set_ao_size(RID p_gi_probe, float p_strength); - float gi_probe_get_ao_size(RID p_gi_probe) const; + void voxel_gi_set_energy(RID p_voxel_gi, float p_energy); + float voxel_gi_get_energy(RID p_voxel_gi) const; - void gi_probe_set_bias(RID p_gi_probe, float p_bias); - float gi_probe_get_bias(RID p_gi_probe) const; + void voxel_gi_set_bias(RID p_voxel_gi, float p_bias); + float voxel_gi_get_bias(RID p_voxel_gi) const; - void gi_probe_set_normal_bias(RID p_gi_probe, float p_range); - float gi_probe_get_normal_bias(RID p_gi_probe) const; + void voxel_gi_set_normal_bias(RID p_voxel_gi, float p_range); + float voxel_gi_get_normal_bias(RID p_voxel_gi) const; - void gi_probe_set_interior(RID p_gi_probe, bool p_enable); - bool gi_probe_is_interior(RID p_gi_probe) const; + void voxel_gi_set_interior(RID p_voxel_gi, bool p_enable); + bool voxel_gi_is_interior(RID p_voxel_gi) const; - void gi_probe_set_use_two_bounces(RID p_gi_probe, bool p_enable); - bool gi_probe_is_using_two_bounces(RID p_gi_probe) const; + void voxel_gi_set_use_two_bounces(RID p_voxel_gi, bool p_enable); + bool voxel_gi_is_using_two_bounces(RID p_voxel_gi) const; - void gi_probe_set_anisotropy_strength(RID p_gi_probe, float p_strength); - float gi_probe_get_anisotropy_strength(RID p_gi_probe) const; + void voxel_gi_set_anisotropy_strength(RID p_voxel_gi, float p_strength); + float voxel_gi_get_anisotropy_strength(RID p_voxel_gi) const; - uint32_t gi_probe_get_version(RID p_probe); - uint32_t gi_probe_get_data_version(RID p_probe); + uint32_t voxel_gi_get_version(RID p_probe); + uint32_t voxel_gi_get_data_version(RID p_probe); - RID gi_probe_get_octree_buffer(RID p_gi_probe) const; - RID gi_probe_get_data_buffer(RID p_gi_probe) const; + RID voxel_gi_get_octree_buffer(RID p_voxel_gi) const; + RID voxel_gi_get_data_buffer(RID p_voxel_gi) const; - RID gi_probe_get_sdf_texture(RID p_gi_probe); + RID voxel_gi_get_sdf_texture(RID p_voxel_gi); /* LIGHTMAP CAPTURE */ @@ -2064,22 +2136,30 @@ public: RID particles_allocate(); void particles_initialize(RID p_particles_collision); + void particles_set_mode(RID p_particles, RS::ParticlesMode p_mode); void particles_set_emitting(RID p_particles, bool p_emitting); void particles_set_amount(RID p_particles, int p_amount); - void particles_set_lifetime(RID p_particles, float p_lifetime); + void particles_set_lifetime(RID p_particles, double p_lifetime); void particles_set_one_shot(RID p_particles, bool p_one_shot); - void particles_set_pre_process_time(RID p_particles, float p_time); - void particles_set_explosiveness_ratio(RID p_particles, float p_ratio); - void particles_set_randomness_ratio(RID p_particles, float p_ratio); + void particles_set_pre_process_time(RID p_particles, double p_time); + void particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio); + void particles_set_randomness_ratio(RID p_particles, real_t p_ratio); void particles_set_custom_aabb(RID p_particles, const AABB &p_aabb); - void particles_set_speed_scale(RID p_particles, float p_scale); + void particles_set_speed_scale(RID p_particles, double p_scale); void particles_set_use_local_coordinates(RID p_particles, bool p_enable); void particles_set_process_material(RID p_particles, RID p_material); void particles_set_fixed_fps(RID p_particles, int p_fps); + void particles_set_interpolate(RID p_particles, bool p_enable); void particles_set_fractional_delta(RID p_particles, bool p_enable); - void particles_set_collision_base_size(RID p_particles, float p_size); + void particles_set_collision_base_size(RID p_particles, real_t p_size); + void particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align); + + void particles_set_trails(RID p_particles, bool p_enable, double p_length); + void particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses); + void particles_restart(RID p_particles); - void particles_emit(RID p_particles, const Transform &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags); + void particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags); + void particles_set_subemitter(RID p_particles, RID p_subemitter_particles); void particles_set_draw_order(RID p_particles, RS::ParticlesDrawOrder p_order); @@ -2091,21 +2171,40 @@ public: AABB particles_get_current_aabb(RID p_particles); AABB particles_get_aabb(RID p_particles) const; - void particles_set_emission_transform(RID p_particles, const Transform &p_transform); + void particles_set_emission_transform(RID p_particles, const Transform3D &p_transform); bool particles_get_emitting(RID p_particles); int particles_get_draw_passes(RID p_particles) const; RID particles_get_draw_pass_mesh(RID p_particles, int p_pass) const; - void particles_set_view_axis(RID p_particles, const Vector3 &p_axis); + void particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis); virtual bool particles_is_inactive(RID p_particles) const; - _FORCE_INLINE_ uint32_t particles_get_amount(RID p_particles) { + _FORCE_INLINE_ RS::ParticlesMode particles_get_mode(RID p_particles) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND_V(!particles, RS::PARTICLES_MODE_2D); + return particles->mode; + } + + _FORCE_INLINE_ uint32_t particles_get_amount(RID p_particles, uint32_t &r_trail_divisor) { + Particles *particles = particles_owner.getornull(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + r_trail_divisor = particles->trail_bind_poses.size(); + } else { + r_trail_divisor = 1; + } + + return particles->amount * r_trail_divisor; + } + + _FORCE_INLINE_ bool particles_has_collision(RID p_particles) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, 0); - return particles->amount; + return particles->has_collision_cache; } _FORCE_INLINE_ uint32_t particles_is_using_local_coords(RID p_particles) { @@ -2119,6 +2218,8 @@ public: Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, RID()); if (particles->particles_transforms_buffer_uniform_set.is_null()) { + _particles_update_buffers(particles); + Vector<RD::Uniform> uniforms; { @@ -2137,6 +2238,7 @@ public: virtual void particles_add_collision(RID p_particles, RID p_particles_collision_instance); virtual void particles_remove_collision(RID p_particles, RID p_particles_collision_instance); + virtual void particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture); /* PARTICLES COLLISION */ @@ -2145,11 +2247,11 @@ public: virtual void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type); virtual void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask); - virtual void particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius); //for spheres + virtual void particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius); //for spheres virtual void particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents); //for non-spheres - virtual void particles_collision_set_attractor_strength(RID p_particles_collision, float p_strength); - virtual void particles_collision_set_attractor_directionality(RID p_particles_collision, float p_directionality); - virtual void particles_collision_set_attractor_attenuation(RID p_particles_collision, float p_curve); + virtual void particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength); + virtual void particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality); + virtual void particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve); virtual void particles_collision_set_field_texture(RID p_particles_collision, RID p_texture); //for SDF and vector field, heightfield is dynamic virtual void particles_collision_height_field_update(RID p_particles_collision); //for SDF and vector field virtual void particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution); //for SDF and vector field @@ -2158,9 +2260,17 @@ public: virtual bool particles_collision_is_heightfield(RID p_particles_collision) const; RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const; + virtual RID visibility_notifier_allocate(); + virtual void visibility_notifier_initialize(RID p_notifier); + virtual void visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb); + virtual void visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable); + + virtual AABB visibility_notifier_get_aabb(RID p_notifier) const; + virtual void visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred); + //used from 2D and 3D virtual RID particles_collision_instance_create(RID p_collision); - virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform &p_transform); + virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform); virtual void particles_collision_instance_set_active(RID p_collision_instance, bool p_active); /* GLOBAL VARIABLES API */ @@ -2188,7 +2298,7 @@ public: RID render_target_create(); void render_target_set_position(RID p_render_target, int p_x, int p_y); - void render_target_set_size(RID p_render_target, int p_width, int p_height); + void render_target_set_size(RID p_render_target, int p_width, int p_height, uint32_t p_view_count); RID render_target_get_texture(RID p_render_target); void render_target_set_external_texture(RID p_render_target, unsigned int p_texture_id); void render_target_set_flag(RID p_render_target, RenderTargetFlags p_flag, bool p_value); @@ -2211,6 +2321,8 @@ public: RID render_target_get_sdf_framebuffer(RID p_render_target); void render_target_sdf_process(RID p_render_target); virtual Rect2i render_target_get_sdf_rect(RID p_render_target) const; + void render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled); + bool render_target_is_sdf_enabled(RID p_render_target) const; Size2 render_target_get_size(RID p_render_target); RID render_target_get_rd_framebuffer(RID p_render_target); @@ -2234,13 +2346,16 @@ public: void set_debug_generate_wireframes(bool p_generate) {} - void render_info_begin_capture() {} - void render_info_end_capture() {} - int get_captured_render_info(RS::RenderInfo p_info) { return 0; } + //keep cached since it can be called form any thread + uint64_t texture_mem_cache = 0; + uint64_t buffer_mem_cache = 0; + uint64_t total_mem_cache = 0; + + virtual void update_memory_info(); + virtual uint64_t get_rendering_info(RS::RenderingInfo p_info); - uint64_t get_render_info(RS::RenderInfo p_info) { return 0; } - String get_video_adapter_name() const { return String(); } - String get_video_adapter_vendor() const { return String(); } + String get_video_adapter_name() const; + String get_video_adapter_vendor() const; virtual void capture_timestamps_begin(); virtual void capture_timestamp(const String &p_name); @@ -2254,6 +2369,7 @@ public: static RendererStorageRD *base_singleton; + void init_effects(bool p_prefer_raster_effects); EffectsRD *get_effects(); RendererStorageRD(); diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp index 8135d388e1..bad37f5c25 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -213,7 +213,7 @@ static String _interpstr(SL::DataInterpolation p_interp) { return ""; } -static String _prestr(SL::DataPrecision p_pres) { +static String _prestr(SL::DataPrecision p_pres, bool p_force_highp = false) { switch (p_pres) { case SL::PRECISION_LOWP: return "lowp "; @@ -222,7 +222,7 @@ static String _prestr(SL::DataPrecision p_pres) { case SL::PRECISION_HIGHP: return "highp "; case SL::PRECISION_DEFAULT: - return ""; + return p_force_highp ? "highp " : ""; } return ""; } @@ -369,17 +369,24 @@ void ShaderCompilerRD::_dump_function_deps(const SL::ShaderNode *p_node, const S ERR_FAIL_COND(fidx == -1); + Vector<StringName> uses_functions; + for (Set<StringName>::Element *E = p_node->functions[fidx].uses_function.front(); E; E = E->next()) { - if (added.has(E->get())) { + uses_functions.push_back(E->get()); + } + uses_functions.sort_custom<StringName::AlphCompare>(); //ensure order is deterministic so the same shader is always produced + + for (int k = 0; k < uses_functions.size(); k++) { + if (added.has(uses_functions[k])) { continue; //was added already } - _dump_function_deps(p_node, E->get(), p_func_code, r_to_add, added); + _dump_function_deps(p_node, uses_functions[k], p_func_code, r_to_add, added); SL::FunctionNode *fnode = nullptr; for (int i = 0; i < p_node->functions.size(); i++) { - if (p_node->functions[i].name == E->get()) { + if (p_node->functions[i].name == uses_functions[k]) { fnode = p_node->functions[i].function; break; } @@ -391,10 +398,21 @@ void ShaderCompilerRD::_dump_function_deps(const SL::ShaderNode *p_node, const S String header; if (fnode->return_type == SL::TYPE_STRUCT) { - header = _mkid(fnode->return_struct_name) + " " + _mkid(fnode->name) + "("; + header = _mkid(fnode->return_struct_name); } else { - header = _typestr(fnode->return_type) + " " + _mkid(fnode->name) + "("; + header = _typestr(fnode->return_type); + } + + if (fnode->return_array_size > 0) { + header += "["; + header += itos(fnode->return_array_size); + header += "]"; } + + header += " "; + header += _mkid(fnode->name); + header += "("; + for (int i = 0; i < fnode->arguments.size(); i++) { if (i > 0) { header += ", "; @@ -407,13 +425,18 @@ void ShaderCompilerRD::_dump_function_deps(const SL::ShaderNode *p_node, const S } else { header += _qualstr(fnode->arguments[i].qualifier) + _prestr(fnode->arguments[i].precision) + _typestr(fnode->arguments[i].type) + " " + _mkid(fnode->arguments[i].name); } + if (fnode->arguments[i].array_size > 0) { + header += "["; + header += itos(fnode->arguments[i].array_size); + header += "]"; + } } header += ")\n"; r_to_add += header; - r_to_add += p_func_code[E->get()]; + r_to_add += p_func_code[uses_functions[k]]; - added.insert(E->get()); + added.insert(uses_functions[k]); } } @@ -535,9 +558,9 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge struct_code += "}"; struct_code += ";\n"; - r_gen_code.vertex_global += struct_code; - r_gen_code.fragment_global += struct_code; - r_gen_code.compute_global += struct_code; + for (int j = 0; j < STAGE_MAX; j++) { + r_gen_code.stage_globals[j] += struct_code; + } } int max_texture_uniforms = 0; @@ -548,7 +571,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge max_texture_uniforms++; } else { if (E->get().scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { - continue; //instances are indexed directly, dont need index uniforms + continue; // Instances are indexed directly, don't need index uniforms. } max_uniforms++; @@ -565,64 +588,74 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge uniform_defines.resize(max_uniforms); bool uses_uniforms = false; + Vector<StringName> uniform_names; + for (Map<StringName, SL::ShaderNode::Uniform>::Element *E = pnode->uniforms.front(); E; E = E->next()) { + uniform_names.push_back(E->key()); + } + + uniform_names.sort_custom<StringName::AlphCompare>(); //ensure order is deterministic so the same shader is always produced + + for (int k = 0; k < uniform_names.size(); k++) { + StringName uniform_name = uniform_names[k]; + const SL::ShaderNode::Uniform &uniform = pnode->uniforms[uniform_name]; + String ucode; - if (E->get().scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { + if (uniform.scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { //insert, but don't generate any code. - p_actions.uniforms->insert(E->key(), E->get()); - continue; //instances are indexed directly, dont need index uniforms + p_actions.uniforms->insert(uniform_name, uniform); + continue; // Instances are indexed directly, don't need index uniforms. } - if (SL::is_sampler_type(E->get().type)) { - ucode = "layout(set = " + itos(actions.texture_layout_set) + ", binding = " + itos(actions.base_texture_binding_index + E->get().texture_order) + ") uniform "; + if (SL::is_sampler_type(uniform.type)) { + ucode = "layout(set = " + itos(actions.texture_layout_set) + ", binding = " + itos(actions.base_texture_binding_index + uniform.texture_order) + ") uniform "; } - bool is_buffer_global = !SL::is_sampler_type(E->get().type) && E->get().scope == SL::ShaderNode::Uniform::SCOPE_GLOBAL; + bool is_buffer_global = !SL::is_sampler_type(uniform.type) && uniform.scope == SL::ShaderNode::Uniform::SCOPE_GLOBAL; if (is_buffer_global) { //this is an integer to index the global table ucode += _typestr(ShaderLanguage::TYPE_UINT); } else { - ucode += _prestr(E->get().precision); - ucode += _typestr(E->get().type); + ucode += _prestr(uniform.precision, ShaderLanguage::is_float_type(uniform.type)); + ucode += _typestr(uniform.type); } - ucode += " " + _mkid(E->key()); + ucode += " " + _mkid(uniform_name); ucode += ";\n"; - if (SL::is_sampler_type(E->get().type)) { - r_gen_code.vertex_global += ucode; - r_gen_code.fragment_global += ucode; - r_gen_code.compute_global += ucode; + if (SL::is_sampler_type(uniform.type)) { + for (int j = 0; j < STAGE_MAX; j++) { + r_gen_code.stage_globals[j] += ucode; + } GeneratedCode::Texture texture; - texture.name = E->key(); - texture.hint = E->get().hint; - texture.type = E->get().type; - texture.filter = E->get().filter; - texture.repeat = E->get().repeat; - texture.global = E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL; + texture.name = uniform_name; + texture.hint = uniform.hint; + texture.type = uniform.type; + texture.filter = uniform.filter; + texture.repeat = uniform.repeat; + texture.global = uniform.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL; if (texture.global) { r_gen_code.uses_global_textures = true; } - r_gen_code.texture_uniforms.write[E->get().texture_order] = texture; + r_gen_code.texture_uniforms.write[uniform.texture_order] = texture; } else { if (!uses_uniforms) { - r_gen_code.defines.push_back(String("#define USE_MATERIAL_UNIFORMS\n")); uses_uniforms = true; } - uniform_defines.write[E->get().order] = ucode; + uniform_defines.write[uniform.order] = ucode; if (is_buffer_global) { //globals are indices into the global table - uniform_sizes.write[E->get().order] = _get_datatype_size(ShaderLanguage::TYPE_UINT); - uniform_alignments.write[E->get().order] = _get_datatype_alignment(ShaderLanguage::TYPE_UINT); + uniform_sizes.write[uniform.order] = _get_datatype_size(ShaderLanguage::TYPE_UINT); + uniform_alignments.write[uniform.order] = _get_datatype_alignment(ShaderLanguage::TYPE_UINT); } else { - uniform_sizes.write[E->get().order] = _get_datatype_size(E->get().type); - uniform_alignments.write[E->get().order] = _get_datatype_alignment(E->get().type); + uniform_sizes.write[uniform.order] = _get_datatype_size(uniform.type); + uniform_alignments.write[uniform.order] = _get_datatype_alignment(uniform.type); } } - p_actions.uniforms->insert(E->key(), E->get()); + p_actions.uniforms->insert(uniform_name, uniform); } for (int i = 0; i < max_uniforms; i++) { @@ -689,50 +722,62 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge List<Pair<StringName, SL::ShaderNode::Varying>> var_frag_to_light; + Vector<StringName> varying_names; + for (Map<StringName, SL::ShaderNode::Varying>::Element *E = pnode->varyings.front(); E; E = E->next()) { - if (E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT_TO_LIGHT || E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT) { - var_frag_to_light.push_back(Pair<StringName, SL::ShaderNode::Varying>(E->key(), E->get())); - fragment_varyings.insert(E->key()); + varying_names.push_back(E->key()); + } + + varying_names.sort_custom<StringName::AlphCompare>(); //ensure order is deterministic so the same shader is always produced + + for (int k = 0; k < varying_names.size(); k++) { + StringName varying_name = varying_names[k]; + const SL::ShaderNode::Varying &varying = pnode->varyings[varying_name]; + + if (varying.stage == SL::ShaderNode::Varying::STAGE_FRAGMENT_TO_LIGHT || varying.stage == SL::ShaderNode::Varying::STAGE_FRAGMENT) { + var_frag_to_light.push_back(Pair<StringName, SL::ShaderNode::Varying>(varying_name, varying)); + fragment_varyings.insert(varying_name); continue; } String vcode; - String interp_mode = _interpstr(E->get().interpolation); - vcode += _prestr(E->get().precision); - vcode += _typestr(E->get().type); - vcode += " " + _mkid(E->key()); - if (E->get().array_size > 0) { + String interp_mode = _interpstr(varying.interpolation); + vcode += _prestr(varying.precision, ShaderLanguage::is_float_type(varying.type)); + vcode += _typestr(varying.type); + vcode += " " + _mkid(varying_name); + if (varying.array_size > 0) { vcode += "["; - vcode += itos(E->get().array_size); + vcode += itos(varying.array_size); vcode += "]"; } vcode += ";\n"; - r_gen_code.vertex_global += "layout(location=" + itos(index) + ") " + interp_mode + "out " + vcode; - r_gen_code.fragment_global += "layout(location=" + itos(index) + ") " + interp_mode + "in " + vcode; - r_gen_code.compute_global += "layout(location=" + itos(index) + ") " + interp_mode + "out " + vcode; + + r_gen_code.stage_globals[STAGE_VERTEX] += "layout(location=" + itos(index) + ") " + interp_mode + "out " + vcode; + r_gen_code.stage_globals[STAGE_FRAGMENT] += "layout(location=" + itos(index) + ") " + interp_mode + "in " + vcode; + index++; } if (var_frag_to_light.size() > 0) { String gcode = "\n\nstruct {\n"; - for (List<Pair<StringName, SL::ShaderNode::Varying>>::Element *E = var_frag_to_light.front(); E; E = E->next()) { - gcode += "\t" + _prestr(E->get().second.precision) + _typestr(E->get().second.type) + " " + _mkid(E->get().first); - if (E->get().second.array_size > 0) { + for (const Pair<StringName, SL::ShaderNode::Varying> &E : var_frag_to_light) { + gcode += "\t" + _prestr(E.second.precision) + _typestr(E.second.type) + " " + _mkid(E.first); + if (E.second.array_size > 0) { gcode += "["; - gcode += itos(E->get().second.array_size); + gcode += itos(E.second.array_size); gcode += "]"; } gcode += ";\n"; } gcode += "} frag_to_light;\n"; - r_gen_code.fragment_global += gcode; + r_gen_code.stage_globals[STAGE_FRAGMENT] += gcode; } for (int i = 0; i < pnode->vconstants.size(); i++) { const SL::ShaderNode::Constant &cnode = pnode->vconstants[i]; String gcode; gcode += "const "; - gcode += _prestr(cnode.precision); + gcode += _prestr(cnode.precision, ShaderLanguage::is_float_type(cnode.type)); if (cnode.type == SL::TYPE_STRUCT) { gcode += _mkid(cnode.type_str); } else { @@ -747,9 +792,9 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge gcode += "="; gcode += _dump_node_code(cnode.initializer, p_level, r_gen_code, p_actions, p_default_actions, p_assigning); gcode += ";\n"; - r_gen_code.vertex_global += gcode; - r_gen_code.fragment_global += gcode; - r_gen_code.compute_global += gcode; + for (int j = 0; j < STAGE_MAX; j++) { + r_gen_code.stage_globals[j] += gcode; + } } Map<StringName, String> function_code; @@ -765,9 +810,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge //place functions in actual code - Set<StringName> added_vtx; - Set<StringName> added_fragment; //share for light - Set<StringName> added_compute; //share for light + Set<StringName> added_funcs_per_stage[STAGE_MAX]; for (int i = 0; i < pnode->functions.size(); i++) { SL::FunctionNode *fnode = pnode->functions[i].function; @@ -776,24 +819,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge current_func_name = fnode->name; - if (fnode->name == vertex_name) { - _dump_function_deps(pnode, fnode->name, function_code, r_gen_code.vertex_global, added_vtx); - r_gen_code.vertex = function_code[vertex_name]; - } - - if (fnode->name == fragment_name) { - _dump_function_deps(pnode, fnode->name, function_code, r_gen_code.fragment_global, added_fragment); - r_gen_code.fragment = function_code[fragment_name]; - } - - if (fnode->name == light_name) { - _dump_function_deps(pnode, fnode->name, function_code, r_gen_code.fragment_global, added_fragment); - r_gen_code.light = function_code[light_name]; - } - - if (fnode->name == compute_name) { - _dump_function_deps(pnode, fnode->name, function_code, r_gen_code.compute_global, added_compute); - r_gen_code.compute = function_code[compute_name]; + if (p_actions.entry_point_stages.has(fnode->name)) { + Stage stage = p_actions.entry_point_stages[fnode->name]; + _dump_function_deps(pnode, fnode->name, function_code, r_gen_code.stage_globals[stage], added_funcs_per_stage[stage]); + r_gen_code.code[fnode->name] = function_code[fnode->name]; } function = nullptr; @@ -858,7 +887,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::VariableNode *vnode = (SL::VariableNode *)p_node; bool use_fragment_varying = false; - if (current_func_name != vertex_name) { + if (!vnode->is_local && !(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { if (p_assigning) { if (shader->varyings.has(vnode->name)) { use_fragment_varying = true; @@ -921,10 +950,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } if (vnode->name == time_name) { - if (current_func_name == vertex_name) { + if (p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX) { r_gen_code.uses_vertex_time = true; } - if (current_func_name == fragment_name || current_func_name == light_name) { + if (p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_FRAGMENT) { r_gen_code.uses_fragment_time = true; } } @@ -975,25 +1004,30 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge declaration += itos(adnode->declarations[i].size); } declaration += "]"; - int sz = adnode->declarations[i].initializer.size(); - if (sz > 0) { + if (adnode->declarations[i].single_expression) { declaration += "="; - if (adnode->datatype == SL::TYPE_STRUCT) { - declaration += _mkid(adnode->struct_name); - } else { - declaration += _typestr(adnode->datatype); - } - declaration += "["; - declaration += itos(sz); - declaration += "]"; - declaration += "("; - for (int j = 0; j < sz; j++) { - declaration += _dump_node_code(adnode->declarations[i].initializer[j], p_level, r_gen_code, p_actions, p_default_actions, p_assigning); - if (j != sz - 1) { - declaration += ", "; + declaration += _dump_node_code(adnode->declarations[i].initializer[0], p_level, r_gen_code, p_actions, p_default_actions, p_assigning); + } else { + int sz = adnode->declarations[i].initializer.size(); + if (sz > 0) { + declaration += "="; + if (adnode->datatype == SL::TYPE_STRUCT) { + declaration += _mkid(adnode->struct_name); + } else { + declaration += _typestr(adnode->datatype); + } + declaration += "["; + declaration += itos(sz); + declaration += "]"; + declaration += "("; + for (int j = 0; j < sz; j++) { + declaration += _dump_node_code(adnode->declarations[i].initializer[j], p_level, r_gen_code, p_actions, p_default_actions, p_assigning); + if (j != sz - 1) { + declaration += ", "; + } } + declaration += ")"; } - declaration += ")"; } } @@ -1003,8 +1037,8 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::ArrayNode *anode = (SL::ArrayNode *)p_node; bool use_fragment_varying = false; - if (current_func_name != vertex_name) { - if (anode->assign_expression != nullptr) { + if (!anode->is_local && !(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { + if (anode->assign_expression != nullptr && shader->varyings.has(anode->name)) { use_fragment_varying = true; } else { if (p_assigning) { @@ -1059,10 +1093,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } if (anode->name == time_name) { - if (current_func_name == vertex_name) { + if (p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX) { r_gen_code.uses_vertex_time = true; } - if (current_func_name == fragment_name || current_func_name == light_name) { + if (p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_FRAGMENT) { r_gen_code.uses_fragment_time = true; } } @@ -1296,6 +1330,9 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } else if (mnode->assign_expression != nullptr) { code += "="; code += _dump_node_code(mnode->assign_expression, p_level, r_gen_code, p_actions, p_default_actions, true, false); + } else if (mnode->call_expression != nullptr) { + code += "."; + code += _dump_node_code(mnode->call_expression, p_level, r_gen_code, p_actions, p_default_actions, p_assigning, false); } } break; } @@ -1309,12 +1346,18 @@ ShaderLanguage::DataType ShaderCompilerRD::_get_variable_type(const StringName & } Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, IdentifierActions *p_actions, const String &p_path, GeneratedCode &r_gen_code) { - Error err = parser.compile(p_code, ShaderTypes::get_singleton()->get_functions(p_mode), ShaderTypes::get_singleton()->get_modes(p_mode), ShaderTypes::get_singleton()->get_types(), _get_variable_type); + Error err = parser.compile(p_code, ShaderTypes::get_singleton()->get_functions(p_mode), ShaderTypes::get_singleton()->get_modes(p_mode), ShaderLanguage::VaryingFunctionNames(), ShaderTypes::get_singleton()->get_types(), _get_variable_type); if (err != OK) { Vector<String> shader = p_code.split("\n"); for (int i = 0; i < shader.size(); i++) { - print_line(itos(i + 1) + " " + shader[i]); + if (i + 1 == parser.get_error_line()) { + // Mark the error line to be visible without having to look at + // the trace at the end. + print_line(vformat("E%4d-> %s", i + 1, shader[i])); + } else { + print_line(vformat("%5d | %s", i + 1, shader[i])); + } } _err_print_error(nullptr, p_path.utf8().get_data(), parser.get_error_line(), parser.get_error_text().utf8().get_data(), ERR_HANDLER_SHADER); @@ -1322,13 +1365,10 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide } r_gen_code.defines.clear(); - r_gen_code.vertex = String(); - r_gen_code.vertex_global = String(); - r_gen_code.fragment = String(); - r_gen_code.fragment_global = String(); - r_gen_code.compute = String(); - r_gen_code.compute_global = String(); - r_gen_code.light = String(); + r_gen_code.code.clear(); + for (int i = 0; i < STAGE_MAX; i++) { + r_gen_code.stage_globals[i] = String(); + } r_gen_code.uses_fragment_time = false; r_gen_code.uses_vertex_time = false; r_gen_code.uses_global_textures = false; @@ -1348,18 +1388,14 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide void ShaderCompilerRD::initialize(DefaultIdentifierActions p_actions) { actions = p_actions; - vertex_name = "vertex"; - fragment_name = "fragment"; - compute_name = "compute"; - light_name = "light"; time_name = "TIME"; List<String> func_list; ShaderLanguage::get_builtin_funcs(&func_list); - for (List<String>::Element *E = func_list.front(); E; E = E->next()) { - internal_functions.insert(E->get()); + for (const String &E : func_list) { + internal_functions.insert(E); } texture_functions.insert("texture"); texture_functions.insert("textureProj"); @@ -1474,7 +1510,6 @@ ShaderCompilerRD::ShaderCompilerRD() { actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; } - actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_oren_nayar"] = "#define DIFFUSE_OREN_NAYAR\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.h b/servers/rendering/renderer_rd/shader_compiler_rd.h index 6575829e73..2da127ffa3 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.h +++ b/servers/rendering/renderer_rd/shader_compiler_rd.h @@ -38,7 +38,16 @@ class ShaderCompilerRD { public: + enum Stage { + STAGE_VERTEX, + STAGE_FRAGMENT, + STAGE_COMPUTE, + STAGE_MAX + }; + struct IdentifierActions { + Map<StringName, Stage> entry_point_stages; + Map<StringName, Pair<int *, int>> render_mode_values; Map<StringName, bool *> render_mode_flags; Map<StringName, bool *> usage_flag_pointers; @@ -63,13 +72,9 @@ public: Vector<uint32_t> uniform_offsets; uint32_t uniform_total_size; String uniforms; - String vertex_global; - String vertex; - String fragment_global; - String fragment; - String light; - String compute_global; - String compute; + String stage_globals[STAGE_MAX]; + + Map<String, String> code; bool uses_global_textures; bool uses_fragment_time; @@ -103,10 +108,6 @@ private: const ShaderLanguage::ShaderNode *shader; const ShaderLanguage::FunctionNode *function; StringName current_func_name; - StringName vertex_name; - StringName fragment_name; - StringName light_name; - StringName compute_name; StringName time_name; Set<StringName> texture_functions; diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index e4a39ff813..82efa1318c 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -30,148 +30,104 @@ #include "shader_rd.h" -#include "core/string/string_builder.h" +#include "core/io/compression.h" +#include "core/io/dir_access.h" +#include "core/io/file_access.h" #include "renderer_compositor_rd.h" #include "servers/rendering/rendering_device.h" - -void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name) { - name = p_name; - //split vertex and shader code (thank you, shader compiler programmers from you know what company). - if (p_vertex_code) { - String defines_tag = "\nVERSION_DEFINES"; - String globals_tag = "\nVERTEX_SHADER_GLOBALS"; - String material_tag = "\nMATERIAL_UNIFORMS"; - String code_tag = "\nVERTEX_SHADER_CODE"; - String code = p_vertex_code; - - int cpos = code.find(defines_tag); - if (cpos != -1) { - vertex_codev = code.substr(0, cpos).ascii(); - code = code.substr(cpos + defines_tag.length(), code.length()); - } - - cpos = code.find(material_tag); - - if (cpos == -1) { - vertex_code0 = code.ascii(); - } else { - vertex_code0 = code.substr(0, cpos).ascii(); - code = code.substr(cpos + material_tag.length(), code.length()); - - cpos = code.find(globals_tag); - - if (cpos == -1) { - vertex_code1 = code.ascii(); - } else { - vertex_code1 = code.substr(0, cpos).ascii(); - String code2 = code.substr(cpos + globals_tag.length(), code.length()); - - cpos = code2.find(code_tag); - if (cpos == -1) { - vertex_code2 = code2.ascii(); - } else { - vertex_code2 = code2.substr(0, cpos).ascii(); - vertex_code3 = code2.substr(cpos + code_tag.length(), code2.length()).ascii(); +#include "thirdparty/misc/smolv.h" + +void ShaderRD::_add_stage(const char *p_code, StageType p_stage_type) { + Vector<String> lines = String(p_code).split("\n"); + + String text; + + for (int i = 0; i < lines.size(); i++) { + String l = lines[i]; + bool push_chunk = false; + + StageTemplate::Chunk chunk; + + if (l.begins_with("#VERSION_DEFINES")) { + chunk.type = StageTemplate::Chunk::TYPE_VERSION_DEFINES; + push_chunk = true; + } else if (l.begins_with("#GLOBALS")) { + switch (p_stage_type) { + case STAGE_TYPE_VERTEX: + chunk.type = StageTemplate::Chunk::TYPE_VERTEX_GLOBALS; + break; + case STAGE_TYPE_FRAGMENT: + chunk.type = StageTemplate::Chunk::TYPE_FRAGMENT_GLOBALS; + break; + case STAGE_TYPE_COMPUTE: + chunk.type = StageTemplate::Chunk::TYPE_COMPUTE_GLOBALS; + break; + default: { } } - } - } - if (p_fragment_code) { - String defines_tag = "\nVERSION_DEFINES"; - String globals_tag = "\nFRAGMENT_SHADER_GLOBALS"; - String material_tag = "\nMATERIAL_UNIFORMS"; - String code_tag = "\nFRAGMENT_SHADER_CODE"; - String light_code_tag = "\nLIGHT_SHADER_CODE"; - String code = p_fragment_code; - - int cpos = code.find(defines_tag); - if (cpos != -1) { - fragment_codev = code.substr(0, cpos).ascii(); - code = code.substr(cpos + defines_tag.length(), code.length()); + push_chunk = true; + } else if (l.begins_with("#MATERIAL_UNIFORMS")) { + chunk.type = StageTemplate::Chunk::TYPE_MATERIAL_UNIFORMS; + push_chunk = true; + } else if (l.begins_with("#CODE")) { + chunk.type = StageTemplate::Chunk::TYPE_CODE; + push_chunk = true; + chunk.code = l.replace_first("#CODE", String()).replace(":", "").strip_edges().to_upper(); + } else { + text += l + "\n"; } - cpos = code.find(material_tag); - if (cpos == -1) { - fragment_code0 = code.ascii(); - } else { - fragment_code0 = code.substr(0, cpos).ascii(); - //print_line("CODE0:\n"+String(fragment_code0.get_data())); - code = code.substr(cpos + material_tag.length(), code.length()); - cpos = code.find(globals_tag); - - if (cpos == -1) { - fragment_code1 = code.ascii(); - } else { - fragment_code1 = code.substr(0, cpos).ascii(); - //print_line("CODE1:\n"+String(fragment_code1.get_data())); - - String code2 = code.substr(cpos + globals_tag.length(), code.length()); - cpos = code2.find(light_code_tag); - - if (cpos == -1) { - fragment_code2 = code2.ascii(); - } else { - fragment_code2 = code2.substr(0, cpos).ascii(); - //print_line("CODE2:\n"+String(fragment_code2.get_data())); - - String code3 = code2.substr(cpos + light_code_tag.length(), code2.length()); - - cpos = code3.find(code_tag); - if (cpos == -1) { - fragment_code3 = code3.ascii(); - } else { - fragment_code3 = code3.substr(0, cpos).ascii(); - //print_line("CODE3:\n"+String(fragment_code3.get_data())); - fragment_code4 = code3.substr(cpos + code_tag.length(), code3.length()).ascii(); - //print_line("CODE4:\n"+String(fragment_code4.get_data())); - } - } + if (push_chunk) { + if (text != String()) { + StageTemplate::Chunk text_chunk; + text_chunk.type = StageTemplate::Chunk::TYPE_TEXT; + text_chunk.text = text.utf8(); + stage_templates[p_stage_type].chunks.push_back(text_chunk); + text = String(); } + stage_templates[p_stage_type].chunks.push_back(chunk); } } - if (p_compute_code) { - is_compute = true; + if (text != String()) { + StageTemplate::Chunk text_chunk; + text_chunk.type = StageTemplate::Chunk::TYPE_TEXT; + text_chunk.text = text.utf8(); + stage_templates[p_stage_type].chunks.push_back(text_chunk); + text = String(); + } +} - String defines_tag = "\nVERSION_DEFINES"; - String globals_tag = "\nCOMPUTE_SHADER_GLOBALS"; - String material_tag = "\nMATERIAL_UNIFORMS"; - String code_tag = "\nCOMPUTE_SHADER_CODE"; - String code = p_compute_code; +void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name) { + name = p_name; - int cpos = code.find(defines_tag); - if (cpos != -1) { - compute_codev = code.substr(0, cpos).ascii(); - code = code.substr(cpos + defines_tag.length(), code.length()); + if (p_compute_code) { + _add_stage(p_compute_code, STAGE_TYPE_COMPUTE); + is_compute = true; + } else { + is_compute = false; + if (p_vertex_code) { + _add_stage(p_vertex_code, STAGE_TYPE_VERTEX); } - - cpos = code.find(material_tag); - - if (cpos == -1) { - compute_code0 = code.ascii(); - } else { - compute_code0 = code.substr(0, cpos).ascii(); - code = code.substr(cpos + material_tag.length(), code.length()); - - cpos = code.find(globals_tag); - - if (cpos == -1) { - compute_code1 = code.ascii(); - } else { - compute_code1 = code.substr(0, cpos).ascii(); - String code2 = code.substr(cpos + globals_tag.length(), code.length()); - - cpos = code2.find(code_tag); - if (cpos == -1) { - compute_code2 = code2.ascii(); - } else { - compute_code2 = code2.substr(0, cpos).ascii(); - compute_code3 = code2.substr(cpos + code_tag.length(), code2.length()).ascii(); - } - } + if (p_fragment_code) { + _add_stage(p_fragment_code, STAGE_TYPE_FRAGMENT); } } + + StringBuilder tohash; + tohash.append("[SpirvCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_spirv_cache_key()); + tohash.append("[BinaryCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_binary_cache_key()); + tohash.append("[Vertex]"); + tohash.append(p_vertex_code ? p_vertex_code : ""); + tohash.append("[Fragment]"); + tohash.append(p_fragment_code ? p_fragment_code : ""); + tohash.append("[Compute]"); + tohash.append(p_compute_code ? p_compute_code : ""); + + base_sha256 = tohash.as_string().sha256_text(); } RID ShaderRD::version_create() { @@ -190,20 +146,68 @@ void ShaderRD::_clear_version(Version *p_version) { //clear versions if they exist if (p_version->variants) { for (int i = 0; i < variant_defines.size(); i++) { - RD::get_singleton()->free(p_version->variants[i]); + if (variants_enabled[i]) { + RD::get_singleton()->free(p_version->variants[i]); + } } memdelete_arr(p_version->variants); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); + } p_version->variants = nullptr; } } +void ShaderRD::_build_variant_code(StringBuilder &builder, uint32_t p_variant, const Version *p_version, const StageTemplate &p_template) { + for (uint32_t i = 0; i < p_template.chunks.size(); i++) { + const StageTemplate::Chunk &chunk = p_template.chunks[i]; + switch (chunk.type) { + case StageTemplate::Chunk::TYPE_VERSION_DEFINES: { + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[p_variant].get_data()); + for (int j = 0; j < p_version->custom_defines.size(); j++) { + builder.append(p_version->custom_defines[j].get_data()); + } + builder.append("\n"); //make sure defines begin at newline + if (p_version->uniforms.size()) { + builder.append("#define MATERIAL_UNIFORMS_USED\n"); + } + for (Map<StringName, CharString>::Element *E = p_version->code_sections.front(); E; E = E->next()) { + builder.append(String("#define ") + String(E->key()) + "_CODE_USED\n"); + } + } break; + case StageTemplate::Chunk::TYPE_MATERIAL_UNIFORMS: { + builder.append(p_version->uniforms.get_data()); //uniforms (same for vertex and fragment) + } break; + case StageTemplate::Chunk::TYPE_VERTEX_GLOBALS: { + builder.append(p_version->vertex_globals.get_data()); // vertex globals + } break; + case StageTemplate::Chunk::TYPE_FRAGMENT_GLOBALS: { + builder.append(p_version->fragment_globals.get_data()); // fragment globals + } break; + case StageTemplate::Chunk::TYPE_COMPUTE_GLOBALS: { + builder.append(p_version->compute_globals.get_data()); // compute globals + } break; + case StageTemplate::Chunk::TYPE_CODE: { + if (p_version->code_sections.has(chunk.code)) { + builder.append(p_version->code_sections[chunk.code].get_data()); + } + } break; + case StageTemplate::Chunk::TYPE_TEXT: { + builder.append(chunk.text.get_data()); + } break; + } + } +} + void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { if (!variants_enabled[p_variant]) { return; //variant is disabled, return } - Vector<RD::ShaderStageData> stages; + Vector<RD::ShaderStageSPIRVData> stages; String error; String current_source; @@ -214,33 +218,11 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { //vertex stage StringBuilder builder; - - builder.append(vertex_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - builder.append(general_defines.get_data()); - builder.append(variant_defines[p_variant].get_data()); - - for (int j = 0; j < p_version->custom_defines.size(); j++) { - builder.append(p_version->custom_defines[j].get_data()); - } - - builder.append(vertex_code0.get_data()); //first part of vertex - - builder.append(p_version->uniforms.get_data()); //uniforms (same for vertex and fragment) - - builder.append(vertex_code1.get_data()); //second part of vertex - - builder.append(p_version->vertex_globals.get_data()); // vertex globals - - builder.append(vertex_code2.get_data()); //third part of vertex - - builder.append(p_version->vertex_code.get_data()); // code - - builder.append(vertex_code3.get_data()); //fourth of vertex + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_VERTEX]); current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -254,37 +236,11 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { current_stage = RD::SHADER_STAGE_FRAGMENT; StringBuilder builder; - - builder.append(fragment_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - - builder.append(general_defines.get_data()); - builder.append(variant_defines[p_variant].get_data()); - for (int j = 0; j < p_version->custom_defines.size(); j++) { - builder.append(p_version->custom_defines[j].get_data()); - } - - builder.append(fragment_code0.get_data()); //first part of fragment - - builder.append(p_version->uniforms.get_data()); //uniforms (same for fragment and fragment) - - builder.append(fragment_code1.get_data()); //first part of fragment - - builder.append(p_version->fragment_globals.get_data()); // fragment globals - - builder.append(fragment_code2.get_data()); //third part of fragment - - builder.append(p_version->fragment_light.get_data()); // fragment light - - builder.append(fragment_code3.get_data()); //fourth part of fragment - - builder.append(p_version->fragment_code.get_data()); // fragment code - - builder.append(fragment_code4.get_data()); //fourth part of fragment + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_FRAGMENT]); current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -298,34 +254,12 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { current_stage = RD::SHADER_STAGE_COMPUTE; StringBuilder builder; - - builder.append(compute_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - builder.append(base_compute_defines.get_data()); - builder.append(general_defines.get_data()); - builder.append(variant_defines[p_variant].get_data()); - - for (int j = 0; j < p_version->custom_defines.size(); j++) { - builder.append(p_version->custom_defines[j].get_data()); - } - - builder.append(compute_code0.get_data()); //first part of compute - - builder.append(p_version->uniforms.get_data()); //uniforms (same for compute and fragment) - - builder.append(compute_code1.get_data()); //second part of compute - - builder.append(p_version->compute_globals.get_data()); // compute globals - - builder.append(compute_code2.get_data()); //third part of compute - - builder.append(p_version->compute_code.get_data()); // code - - builder.append(compute_code3.get_data()); //fourth of compute + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_COMPUTE]); current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -345,10 +279,15 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { return; } - RID shader = RD::get_singleton()->shader_create(stages); + Vector<uint8_t> shader_data = RD::get_singleton()->shader_compile_binary_from_spirv(stages, name + ":" + itos(p_variant)); + + ERR_FAIL_COND(shader_data.size() == 0); + + RID shader = RD::get_singleton()->shader_create_from_bytecode(shader_data); { MutexLock lock(variant_set_mutex); p_version->variants[p_variant] = shader; + p_version->variant_data[p_variant] = shader_data; } } @@ -364,29 +303,7 @@ RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_versio //vertex stage StringBuilder builder; - - builder.append(vertex_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - builder.append(general_defines.get_data()); - builder.append(variant_defines[i].get_data()); - - for (int j = 0; j < version->custom_defines.size(); j++) { - builder.append(version->custom_defines[j].get_data()); - } - - builder.append(vertex_code0.get_data()); //first part of vertex - - builder.append(version->uniforms.get_data()); //uniforms (same for vertex and fragment) - - builder.append(vertex_code1.get_data()); //second part of vertex - - builder.append(version->vertex_globals.get_data()); // vertex globals - - builder.append(vertex_code2.get_data()); //third part of vertex - - builder.append(version->vertex_code.get_data()); // code - - builder.append(vertex_code3.get_data()); //fourth of vertex + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_VERTEX]); RS::ShaderNativeSourceCode::Version::Stage stage; stage.name = "vertex"; @@ -399,78 +316,145 @@ RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_versio //fragment stage StringBuilder builder; + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_FRAGMENT]); - builder.append(fragment_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - builder.append(general_defines.get_data()); - builder.append(variant_defines[i].get_data()); - for (int j = 0; j < version->custom_defines.size(); j++) { - builder.append(version->custom_defines[j].get_data()); - } + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "fragment"; + stage.code = builder.as_string(); - builder.append(fragment_code0.get_data()); //first part of fragment + source_code.versions.write[i].stages.push_back(stage); + } - builder.append(version->uniforms.get_data()); //uniforms (same for fragment and fragment) + if (is_compute) { + //compute stage - builder.append(fragment_code1.get_data()); //first part of fragment + StringBuilder builder; + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_COMPUTE]); - builder.append(version->fragment_globals.get_data()); // fragment globals + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "compute"; + stage.code = builder.as_string(); - builder.append(fragment_code2.get_data()); //third part of fragment + source_code.versions.write[i].stages.push_back(stage); + } + } - builder.append(version->fragment_light.get_data()); // fragment light + return source_code; +} - builder.append(fragment_code3.get_data()); //fourth part of fragment +String ShaderRD::_version_get_sha1(Version *p_version) const { + StringBuilder hash_build; + + hash_build.append("[uniforms]"); + hash_build.append(p_version->uniforms.get_data()); + hash_build.append("[vertex_globals]"); + hash_build.append(p_version->vertex_globals.get_data()); + hash_build.append("[fragment_globals]"); + hash_build.append(p_version->fragment_globals.get_data()); + hash_build.append("[compute_globals]"); + hash_build.append(p_version->compute_globals.get_data()); + + Vector<StringName> code_sections; + for (Map<StringName, CharString>::Element *E = p_version->code_sections.front(); E; E = E->next()) { + code_sections.push_back(E->key()); + } + code_sections.sort_custom<StringName::AlphCompare>(); - builder.append(version->fragment_code.get_data()); // fragment code + for (int i = 0; i < code_sections.size(); i++) { + hash_build.append(String("[code:") + String(code_sections[i]) + "]"); + hash_build.append(p_version->code_sections[code_sections[i]].get_data()); + } + for (int i = 0; i < p_version->custom_defines.size(); i++) { + hash_build.append("[custom_defines:" + itos(i) + "]"); + hash_build.append(p_version->custom_defines[i].get_data()); + } - builder.append(fragment_code4.get_data()); //fourth part of fragment + return hash_build.as_string().sha1_text(); +} - RS::ShaderNativeSourceCode::Version::Stage stage; - stage.name = "fragment"; - stage.code = builder.as_string(); +static const char *shader_file_header = "GDSC"; +static const uint32_t cache_file_version = 2; - source_code.versions.write[i].stages.push_back(stage); - } +bool ShaderRD::_load_from_cache(Version *p_version) { + String sha1 = _version_get_sha1(p_version); + String path = shader_cache_dir.plus_file(name).plus_file(base_sha256).plus_file(sha1) + ".cache"; - if (is_compute) { - //compute stage + FileAccessRef f = FileAccess::open(path, FileAccess::READ); + if (!f) { + return false; + } - StringBuilder builder; + char header[5] = { 0, 0, 0, 0, 0 }; + f->get_buffer((uint8_t *)header, 4); + ERR_FAIL_COND_V(header != String(shader_file_header), false); - builder.append(compute_codev.get_data()); // version info (if exists) - builder.append("\n"); //make sure defines begin at newline - builder.append(base_compute_defines.get_data()); - builder.append(general_defines.get_data()); - builder.append(variant_defines[i].get_data()); + uint32_t file_version = f->get_32(); + if (file_version != cache_file_version) { + return false; // wrong version + } - for (int j = 0; j < version->custom_defines.size(); j++) { - builder.append(version->custom_defines[j].get_data()); - } + uint32_t variant_count = f->get_32(); - builder.append(compute_code0.get_data()); //first part of compute + ERR_FAIL_COND_V(variant_count != (uint32_t)variant_defines.size(), false); //should not happen but check - builder.append(version->uniforms.get_data()); //uniforms (same for compute and fragment) + for (uint32_t i = 0; i < variant_count; i++) { + uint32_t variant_size = f->get_32(); + ERR_FAIL_COND_V(variant_size == 0 && variants_enabled[i], false); + if (!variants_enabled[i]) { + continue; + } + Vector<uint8_t> variant_bytes; + variant_bytes.resize(variant_size); - builder.append(compute_code1.get_data()); //second part of compute + uint32_t br = f->get_buffer(variant_bytes.ptrw(), variant_size); - builder.append(version->compute_globals.get_data()); // compute globals + ERR_FAIL_COND_V(br != variant_size, false); - builder.append(compute_code2.get_data()); //third part of compute + p_version->variant_data[i] = variant_bytes; + } + + for (uint32_t i = 0; i < variant_count; i++) { + if (!variants_enabled[i]) { + MutexLock lock(variant_set_mutex); + p_version->variants[i] = RID(); + continue; + } + RID shader = RD::get_singleton()->shader_create_from_bytecode(p_version->variant_data[i]); + if (shader.is_null()) { + for (uint32_t j = 0; j < i; j++) { + RD::get_singleton()->free(p_version->variants[i]); + } + ERR_FAIL_COND_V(shader.is_null(), false); + } + { + MutexLock lock(variant_set_mutex); + p_version->variants[i] = shader; + } + } - builder.append(version->compute_code.get_data()); // code + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; + p_version->valid = true; + return true; +} - builder.append(compute_code3.get_data()); //fourth of compute +void ShaderRD::_save_to_cache(Version *p_version) { + String sha1 = _version_get_sha1(p_version); + String path = shader_cache_dir.plus_file(name).plus_file(base_sha256).plus_file(sha1) + ".cache"; - RS::ShaderNativeSourceCode::Version::Stage stage; - stage.name = "compute"; - stage.code = builder.as_string(); + FileAccessRef f = FileAccess::open(path, FileAccess::WRITE); + ERR_FAIL_COND(!f); + f->store_buffer((const uint8_t *)shader_file_header, 4); + f->store_32(cache_file_version); //file version + uint32_t variant_count = variant_defines.size(); + f->store_32(variant_count); //variant count - source_code.versions.write[i].stages.push_back(stage); - } + for (uint32_t i = 0; i < variant_count; i++) { + f->store_32(p_version->variant_data[i].size()); //stage count + f->store_buffer(p_version->variant_data[i].ptr(), p_version->variant_data[i].size()); } - return source_code; + f->close(); } void ShaderRD::_compile_version(Version *p_version) { @@ -480,6 +464,15 @@ void ShaderRD::_compile_version(Version *p_version) { p_version->dirty = false; p_version->variants = memnew_arr(RID, variant_defines.size()); + typedef Vector<uint8_t> ShaderStageData; + p_version->variant_data = memnew_arr(ShaderStageData, variant_defines.size()); + + if (shader_cache_dir_valid) { + if (_load_from_cache(p_version)) { + return; + } + } + #if 1 RendererThreadPool::singleton->thread_work_pool.do_work(variant_defines.size(), this, &ShaderRD::_compile_variant, p_version); @@ -511,24 +504,35 @@ void ShaderRD::_compile_version(Version *p_version) { } } memdelete_arr(p_version->variants); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); + } p_version->variants = nullptr; + p_version->variant_data = nullptr; return; + } else if (shader_cache_dir_valid) { + //save shader cache + _save_to_cache(p_version); } + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; + p_version->valid = true; } -void ShaderRD::version_set_code(RID p_version, const String &p_uniforms, const String &p_vertex_globals, const String &p_vertex_code, const String &p_fragment_globals, const String &p_fragment_light, const String &p_fragment_code, const Vector<String> &p_custom_defines) { +void ShaderRD::version_set_code(RID p_version, const Map<String, String> &p_code, const String &p_uniforms, const String &p_vertex_globals, const String &p_fragment_globals, const Vector<String> &p_custom_defines) { ERR_FAIL_COND(is_compute); Version *version = version_owner.getornull(p_version); ERR_FAIL_COND(!version); version->vertex_globals = p_vertex_globals.utf8(); - version->vertex_code = p_vertex_code.utf8(); - version->fragment_light = p_fragment_light.utf8(); version->fragment_globals = p_fragment_globals.utf8(); - version->fragment_code = p_fragment_code.utf8(); version->uniforms = p_uniforms.utf8(); + version->code_sections.clear(); + for (Map<String, String>::Element *E = p_code.front(); E; E = E->next()) { + version->code_sections[StringName(E->key().to_upper())] = E->get().utf8(); + } version->custom_defines.clear(); for (int i = 0; i < p_custom_defines.size(); i++) { @@ -542,15 +546,20 @@ void ShaderRD::version_set_code(RID p_version, const String &p_uniforms, const S } } -void ShaderRD::version_set_compute_code(RID p_version, const String &p_uniforms, const String &p_compute_globals, const String &p_compute_code, const Vector<String> &p_custom_defines) { +void ShaderRD::version_set_compute_code(RID p_version, const Map<String, String> &p_code, const String &p_uniforms, const String &p_compute_globals, const Vector<String> &p_custom_defines) { ERR_FAIL_COND(!is_compute); Version *version = version_owner.getornull(p_version); ERR_FAIL_COND(!version); + version->compute_globals = p_compute_globals.utf8(); - version->compute_code = p_compute_code.utf8(); version->uniforms = p_uniforms.utf8(); + version->code_sections.clear(); + for (Map<String, String>::Element *E = p_code.front(); E; E = E->next()) { + version->code_sections[StringName(E->key().to_upper())] = E->get().utf8(); + } + version->custom_defines.clear(); for (int i = 0; i < p_custom_defines.size(); i++) { version->custom_defines.push_back(p_custom_defines[i].utf8()); @@ -597,6 +606,8 @@ bool ShaderRD::is_variant_enabled(int p_variant) const { return variants_enabled[p_variant]; } +bool ShaderRD::shader_cache_cleanup_on_start = false; + ShaderRD::ShaderRD() { // Do not feel forced to use this, in most cases it makes little to no difference. bool use_32_threads = false; @@ -623,8 +634,64 @@ void ShaderRD::initialize(const Vector<String> &p_variant_defines, const String variant_defines.push_back(p_variant_defines[i].utf8()); variants_enabled.push_back(true); } + + if (shader_cache_dir != String()) { + StringBuilder hash_build; + + hash_build.append("[base_hash]"); + hash_build.append(base_sha256); + hash_build.append("[general_defines]"); + hash_build.append(general_defines.get_data()); + for (int i = 0; i < variant_defines.size(); i++) { + hash_build.append("[variant_defines:" + itos(i) + "]"); + hash_build.append(variant_defines[i].get_data()); + } + + base_sha256 = hash_build.as_string().sha256_text(); + + DirAccessRef d = DirAccess::open(shader_cache_dir); + ERR_FAIL_COND(!d); + if (d->change_dir(name) != OK) { + Error err = d->make_dir(name); + ERR_FAIL_COND(err != OK); + d->change_dir(name); + } + + //erase other versions? + if (shader_cache_cleanup_on_start) { + } + // + if (d->change_dir(base_sha256) != OK) { + Error err = d->make_dir(base_sha256); + ERR_FAIL_COND(err != OK); + } + shader_cache_dir_valid = true; + + print_verbose("Shader '" + name + "' SHA256: " + base_sha256); + } +} + +void ShaderRD::set_shader_cache_dir(const String &p_dir) { + shader_cache_dir = p_dir; +} + +void ShaderRD::set_shader_cache_save_compressed(bool p_enable) { + shader_cache_save_compressed = p_enable; } +void ShaderRD::set_shader_cache_save_compressed_zstd(bool p_enable) { + shader_cache_save_compressed_zstd = p_enable; +} + +void ShaderRD::set_shader_cache_save_debug(bool p_enable) { + shader_cache_save_debug = p_enable; +} + +String ShaderRD::shader_cache_dir; +bool ShaderRD::shader_cache_save_compressed = true; +bool ShaderRD::shader_cache_save_compressed_zstd = true; +bool ShaderRD::shader_cache_save_debug = true; + ShaderRD::~ShaderRD() { List<RID> remaining; version_owner.get_owned_list(&remaining); diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index e0f4dcf2d0..529328f0ed 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -32,7 +32,9 @@ #define SHADER_RD_H #include "core/os/mutex.h" +#include "core/string/string_builder.h" #include "core/templates/hash_map.h" +#include "core/templates/local_vector.h" #include "core/templates/map.h" #include "core/templates/rid_owner.h" #include "core/variant/variant.h" @@ -52,15 +54,13 @@ class ShaderRD { struct Version { CharString uniforms; CharString vertex_globals; - CharString vertex_code; CharString compute_globals; - CharString compute_code; - CharString fragment_light; CharString fragment_globals; - CharString fragment_code; + Map<StringName, CharString> code_sections; Vector<CharString> custom_defines; - RID *variants; //same size as version defines + Vector<uint8_t> *variant_data = nullptr; + RID *variants = nullptr; //same size as version defines bool valid; bool dirty; @@ -76,31 +76,57 @@ class ShaderRD { RID_Owner<Version> version_owner; - CharString fragment_codev; //for version and extensions - CharString fragment_code0; - CharString fragment_code1; - CharString fragment_code2; - CharString fragment_code3; - CharString fragment_code4; - - CharString vertex_codev; //for version and extensions - CharString vertex_code0; - CharString vertex_code1; - CharString vertex_code2; - CharString vertex_code3; + struct StageTemplate { + struct Chunk { + enum Type { + TYPE_VERSION_DEFINES, + TYPE_MATERIAL_UNIFORMS, + TYPE_VERTEX_GLOBALS, + TYPE_FRAGMENT_GLOBALS, + TYPE_COMPUTE_GLOBALS, + TYPE_CODE, + TYPE_TEXT + }; + + Type type; + StringName code; + CharString text; + }; + LocalVector<Chunk> chunks; + }; bool is_compute = false; - CharString compute_codev; //for version and extensions - CharString compute_code0; - CharString compute_code1; - CharString compute_code2; - CharString compute_code3; - - const char *name; + String name; CharString base_compute_defines; + String base_sha256; + + static String shader_cache_dir; + static bool shader_cache_cleanup_on_start; + static bool shader_cache_save_compressed; + static bool shader_cache_save_compressed_zstd; + static bool shader_cache_save_debug; + bool shader_cache_dir_valid = false; + + enum StageType { + STAGE_TYPE_VERTEX, + STAGE_TYPE_FRAGMENT, + STAGE_TYPE_COMPUTE, + STAGE_TYPE_MAX, + }; + + StageTemplate stage_templates[STAGE_TYPE_MAX]; + + void _build_variant_code(StringBuilder &p_builder, uint32_t p_variant, const Version *p_version, const StageTemplate &p_template); + + void _add_stage(const char *p_code, StageType p_stage_type); + + String _version_get_sha1(Version *p_version) const; + bool _load_from_cache(Version *p_version); + void _save_to_cache(Version *p_version); + protected: ShaderRD(); void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name); @@ -108,8 +134,8 @@ protected: public: RID version_create(); - void version_set_code(RID p_version, const String &p_uniforms, const String &p_vertex_globals, const String &p_vertex_code, const String &p_fragment_globals, const String &p_fragment_light, const String &p_fragment_code, const Vector<String> &p_custom_defines); - void version_set_compute_code(RID p_version, const String &p_uniforms, const String &p_compute_globals, const String &p_compute_code, const Vector<String> &p_custom_defines); + void version_set_code(RID p_version, const Map<String, String> &p_code, const String &p_uniforms, const String &p_vertex_globals, const String &p_fragment_globals, const Vector<String> &p_custom_defines); + void version_set_compute_code(RID p_version, const Map<String, String> &p_code, const String &p_uniforms, const String &p_compute_globals, const Vector<String> &p_custom_defines); _FORCE_INLINE_ RID version_get_shader(RID p_version, int p_variant) { ERR_FAIL_INDEX_V(p_variant, variant_defines.size(), RID()); @@ -136,6 +162,11 @@ public: void set_variant_enabled(int p_variant, bool p_enabled); bool is_variant_enabled(int p_variant) const; + static void set_shader_cache_dir(const String &p_dir); + static void set_shader_cache_save_compressed(bool p_enable); + static void set_shader_cache_save_compressed_zstd(bool p_enable); + static void set_shader_cache_save_debug(bool p_enable); + RS::ShaderNativeSourceCode version_get_native_source_code(RID p_version); void initialize(const Vector<String> &p_variant_defines, const String &p_general_defines = ""); diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl new file mode 100644 index 0000000000..967da1e6e4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blit.glsl @@ -0,0 +1,95 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 0, std140) uniform Pos { + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) out vec2 uv; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv = base_arr[gl_VertexIndex]; + vec2 vtx = data.dst_rect.xy + uv * data.dst_rect.zw; + gl_Position = vec4(vtx * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 0, std140) uniform Pos { + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) in vec2 uv; + +layout(location = 0) out vec4 color; + +#ifdef USE_LAYER +layout(binding = 0) uniform sampler2DArray src_rt; +#else +layout(binding = 0) uniform sampler2D src_rt; +#endif + +void main() { +#ifdef APPLY_LENS_DISTORTION + vec2 coords = uv * 2.0 - 1.0; + vec2 offset = coords - data.eye_center; + + // take aspect ratio into account + offset.y /= data.aspect_ratio; + + // distort + vec2 offset_sq = offset * offset; + float radius_sq = offset_sq.x + offset_sq.y; + float radius_s4 = radius_sq * radius_sq; + float distortion_scale = 1.0 + (data.k1 * radius_sq) + (data.k2 * radius_s4); + offset *= distortion_scale; + + // reapply aspect ratio + offset.y *= data.aspect_ratio; + + // add our eye center back in + coords = offset + data.eye_center; + coords /= data.upscale; + + // and check our color + if (coords.x < -1.0 || coords.y < -1.0 || coords.x > 1.0 || coords.y > 1.0) { + color = vec4(0.0, 0.0, 0.0, 1.0); + } else { + // layer is always used here + coords = (coords + vec2(1.0)) / vec2(2.0); + color = texture(src_rt, vec3(coords, data.layer)); + } +#elif defined(USE_LAYER) + color = texture(src_rt, vec3(uv, data.layer)); +#else + color = texture(src_rt, uv); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/blur_raster.glsl new file mode 100644 index 0000000000..0789a4b396 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster.glsl @@ -0,0 +1,136 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { +#ifdef MODE_MIPMAP + + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size); + frag_color = color / 4.0; + +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + //Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607; + color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303; + color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774; + color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136; + color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136; + frag_color = color; + } +#endif + +#ifdef MODE_GAUSSIAN_GLOW + + //Glow uses larger sigma 1 for a more rounded blur effect + +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + vec2 ofs = uv_interp + m_ofs * pix_size; \ + vec4 c = texture(source_color, ofs) * m_mult; \ + if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \ + c *= 0.0; \ + } \ + color += c; \ + } + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938; + GLOW_ADD(vec2(1.0, 0.0), 0.165569); + GLOW_ADD(vec2(2.0, 0.0), 0.140367); + GLOW_ADD(vec2(3.0, 0.0), 0.106595); + GLOW_ADD(vec2(-1.0, 0.0), 0.165569); + GLOW_ADD(vec2(-2.0, 0.0), 0.140367); + GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + color *= blur.glow_strength; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713; + GLOW_ADD(vec2(0.0, 1.0), 0.233062); + GLOW_ADD(vec2(0.0, 2.0), 0.122581); + GLOW_ADD(vec2(0.0, -1.0), 0.233062); + GLOW_ADD(vec2(0.0, -2.0), 0.122581); + color *= blur.glow_strength; + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { +#ifdef GLOW_USE_AUTO_EXPOSURE + + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; +#endif + frag_color *= blur.glow_exposure; + + float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); + float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); + + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)); + } + +#endif + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl new file mode 100644 index 0000000000..52bf2886b5 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl @@ -0,0 +1,21 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, binding = 1, std430) uniform Blur { + vec2 pixel_size; + uint flags; + uint pad; + + // Glow. + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_grey; +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl index 63f086a83d..0438671dd2 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 @@ -25,34 +25,7 @@ layout(set = 1, binding = 0) uniform sampler2D source_bokeh; // based on https://www.shadertoy.com/view/Xd3GDl -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 size; - float z_far; - float z_near; - - bool orthogonal; - float blur_size; - float blur_scale; - int blur_steps; - - bool blur_near_active; - float blur_near_begin; - float blur_near_end; - bool blur_far_active; - - float blur_far_begin; - float blur_far_end; - bool second_pass; - bool half_size; - - bool use_jitter; - float jitter_seed; - uint pad[2]; -} -params; - -//used to work around downsampling filter -#define DEPTH_GAP 0.0 +#include "bokeh_dof_inc.glsl" #ifdef MODE_GEN_BLUR_SIZE @@ -80,15 +53,6 @@ float get_blur_size(float depth) { #endif -const float GOLDEN_ANGLE = 2.39996323; - -//note: uniform pdf rand [0;1[ -float hash12n(vec2 p) { - p = fract(p * vec2(5.3987, 5.4421)); - p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); - return fract(p.x * p.y * 95.4307); -} - #if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl new file mode 100644 index 0000000000..fadea1631c --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl @@ -0,0 +1,37 @@ +layout(push_constant, binding = 1, std430) uniform Params { + ivec2 size; + float z_far; + float z_near; + + bool orthogonal; + float blur_size; + float blur_scale; + int blur_steps; + + bool blur_near_active; + float blur_near_begin; + float blur_near_end; + bool blur_far_active; + + float blur_far_begin; + float blur_far_end; + bool second_pass; + bool half_size; + + bool use_jitter; + float jitter_seed; + uint pad[2]; +} +params; + +//used to work around downsampling filter +#define DEPTH_GAP 0.0 + +const float GOLDEN_ANGLE = 2.39996323; + +//note: uniform pdf rand [0;1[ +float hash12n(vec2 p) { + p = fract(p * vec2(5.3987, 5.4421)); + p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); + return fract(p.x * p.y * 95.4307); +} diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..43a2a29616 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl @@ -0,0 +1,253 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +#ifdef MODE_GEN_BLUR_SIZE +layout(location = 0) out float weight; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; +#else +layout(location = 0) out vec4 frag_color; +#ifdef OUTPUT_WEIGHT +layout(location = 1) out float weight; +#endif + +layout(set = 0, binding = 0) uniform sampler2D source_color; +layout(set = 1, binding = 0) uniform sampler2D source_weight; +#ifdef MODE_COMPOSITE_BOKEH +layout(set = 2, binding = 0) uniform sampler2D original_weight; +#endif +#endif + +//DOF +// Bokeh single pass implementation based on http://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(source_color, uv); + color.a = texture(source_weight, uv).r; + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(source_color, suv); + sample_color.a = texture(source_weight, suv).r; + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = uv_interp; + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + float center_depth = get_depth_at_pos(uv); + weight = get_blur_size(center_depth); +#endif + +#ifdef MODE_BOKEH_BOX + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(source_color, uv); + float alpha = color.a; // retain this + color.a = texture(source_weight, uv).r; + + vec4 color_accum = color; + float accum = 1.0; + + float radius = params.blur_scale; + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 uv_adj = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + + vec4 sample_color = texture(source_color, uv_adj); + sample_color.a = texture(source_weight, uv_adj).r; + + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + color_accum += mix(color_accum / accum, sample_color, m); + accum += 1.0; + + radius += params.blur_scale / radius; + } + + color_accum = color_accum / accum; + + frag_color.rgb = color_accum.rgb; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color_accum.a; +#endif + +#endif + +#ifdef MODE_COMPOSITE_BOKEH + frag_color.rgb = texture(source_color, uv).rgb; + + float center_weigth = texture(source_weight, uv).r; + float sample_weight = texture(original_weight, uv).r; + + float mix_amount; + if (sample_weight < center_weigth) { + mix_amount = min(1.0, max(0.0, max(abs(center_weigth), abs(sample_weight)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(center_weigth) - DEPTH_GAP)); + } + + // let alpha blending take care of mixing + frag_color.a = mix_amount; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 3b39edc70e..a443bcdcb8 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef USE_ATTRIBUTES layout(location = 0) in vec2 vertex_attrib; @@ -26,17 +26,15 @@ layout(location = 3) out vec2 pixel_size_interp; #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif -/* clang-format off */ -VERTEX_SHADER_GLOBALS -/* clang-format on */ +#GLOBALS void main() { vec4 instance_custom = vec4(0.0); @@ -67,7 +65,7 @@ void main() { #elif defined(USE_ATTRIBUTES) vec2 vertex = vertex_attrib; - vec4 color = color_attrib; + vec4 color = color_attrib * draw_data.modulation; vec2 uv = uv_attrib; uvec4 bones = bone_attrib; @@ -86,40 +84,82 @@ void main() { mat4 world_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); -#if 0 - if (draw_data.flags & FLAGS_INSTANCING_ENABLED) { - uint offset = draw_data.flags & FLAGS_INSTANCING_STRIDE_MASK; - offset *= gl_InstanceIndex; - mat4 instance_xform = mat4( - vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), 0.0, texelFetch(instancing_buffer, offset + 3)), - vec4(texelFetch(instancing_buffer, offset + 4), texelFetch(instancing_buffer, offset + 5), 0.0, texelFetch(instancing_buffer, offset + 7)), - vec4(0.0, 0.0, 1.0, 0.0), - vec4(0.0, 0.0, 0.0, 1.0)); - offset += 8; - if (draw_data.flags & FLAGS_INSTANCING_HAS_COLORS) { - vec4 instance_color; - if (draw_data.flags & FLAGS_INSTANCING_COLOR_8_BIT) { - uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset)); - instance_color = unpackUnorm4x8(bits); - offset += 1; - } else { - instance_color = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3)); - offset += 4; - } +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) + + uint instancing = draw_data.flags & FLAGS_INSTANCING_MASK; + +#ifdef USE_ATTRIBUTES + + if (instancing > 1) { + // trails - color *= instance_color; + uint stride = 2 + 1 + 1; //particles always uses this format + + uint trail_size = instancing; + + uint offset = trail_size * stride * gl_InstanceIndex; + + vec4 pcolor; + vec2 new_vertex; + { + uint boffset = offset + bone_attrib.x * stride; + new_vertex = (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.x; + pcolor = transforms.data[boffset + 2] * weight_attrib.x; } - if (draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA) { - if (draw_data.flags & FLAGS_INSTANCING_CUSTOM_DATA_8_BIT) { - uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset)); - instance_custom = unpackUnorm4x8(bits); - } else { - instance_custom = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3)); + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.y; + pcolor += transforms.data[boffset + 2] * weight_attrib.y; + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.z; + pcolor += transforms.data[boffset + 2] * weight_attrib.z; + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.w; + pcolor += transforms.data[boffset + 2] * weight_attrib.w; + } + + instance_custom = transforms.data[offset + 3]; + + vertex = new_vertex; + color *= pcolor; + + } else +#endif // USE_ATTRIBUTES + + if (instancing == 1) { + uint stride = 2; + { + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + stride += 1; + } + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + stride += 1; } } - } -#endif + uint offset = stride * gl_InstanceIndex; + + mat4 matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + offset += 2; + + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + color *= transforms.data[offset]; + offset += 1; + } + + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + + matrix = transpose(matrix); + world_matrix = world_matrix * matrix; + } #if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) if (bool(draw_data.flags & FLAGS_USING_PARTICLES)) { @@ -132,9 +172,7 @@ void main() { float point_size = 1.0; #endif { - /* clang-format off */ -VERTEX_SHADER_CODE - /* clang-format on */ +#CODE : VERTEX } #ifdef USE_NINEPATCH @@ -212,7 +250,7 @@ VERTEX_SHADER_CODE #version 450 -VERSION_DEFINES +#VERSION_DEFINES #include "canvas_uniforms_inc.glsl" @@ -228,11 +266,11 @@ layout(location = 3) in vec2 pixel_size_interp; layout(location = 0) out vec4 frag_color; -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif @@ -243,7 +281,7 @@ vec2 screen_uv_to_sdf(vec2 p_uv) { float texture_sdf(vec2 p_sdf) { vec2 uv = p_sdf * canvas_data.sdf_to_tex.xy + canvas_data.sdf_to_tex.zw; float d = texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv).r; - d = d * SDF_MAX_LENGTH - 1.0; + d *= SDF_MAX_LENGTH; return d * canvas_data.tex_to_sdf; } @@ -260,11 +298,9 @@ vec2 sdf_to_screen_uv(vec2 p_sdf) { return p_sdf * canvas_data.sdf_to_screen; } -/* clang-format off */ -FRAGMENT_SHADER_GLOBALS -/* clang-format on */ +#GLOBALS -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED vec4 light_compute( vec3 light_vertex, @@ -278,9 +314,9 @@ vec4 light_compute( vec2 uv, vec4 color, bool is_directional) { vec4 light = vec4(0.0); - /* clang-format off */ -LIGHT_SHADER_CODE - /* clang-format on */ + +#CODE : LIGHT + return light; } @@ -356,7 +392,7 @@ vec3 light_normal_compute(vec3 light_vec, vec3 normal, vec3 base_color, vec3 lig //float distance = length(shadow_pos); vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , vec3 shadow_modulate #endif @@ -395,7 +431,7 @@ vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv } vec4 shadow_color = unpackUnorm4x8(light_array.data[light_base].shadow_color); -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED shadow_color.rgb *= shadow_modulate; #endif @@ -504,11 +540,7 @@ void main() { normal_used = true; #endif - /* clang-format off */ - -FRAGMENT_SHADER_CODE - - /* clang-format on */ +#CODE : FRAGMENT #if defined(NORMAL_MAP_USED) normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_map_depth); @@ -543,7 +575,7 @@ FRAGMENT_SHADER_CODE vec2 direction = light_array.data[light_base].position; vec4 light_color = light_array.data[light_base].color; -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED vec4 shadow_modulate = vec4(1.0); light_color = light_compute(light_vertex, vec3(direction, light_array.data[light_base].height), normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, true); @@ -561,7 +593,7 @@ FRAGMENT_SHADER_CODE vec4 shadow_uv = vec4(shadow_pos.x, light_array.data[light_base].shadow_y_ofs, shadow_pos.y * light_array.data[light_base].shadow_zfar_inv, 1.0); light_color = light_shadow_compute(light_base, light_color, shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , shadow_modulate.rgb #endif @@ -599,7 +631,7 @@ FRAGMENT_SHADER_CODE vec4 light_color = textureLod(sampler2D(atlas_texture, texture_sampler), tex_uv_atlas, 0.0); vec4 light_base_color = light_array.data[light_base].color; -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED vec4 shadow_modulate = vec4(1.0); vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height); @@ -657,7 +689,7 @@ FRAGMENT_SHADER_CODE vec4 shadow_uv = vec4(tex_ofs, light_array.data[light_base].shadow_y_ofs, distance, 1.0); light_color = light_shadow_compute(light_base, light_color, shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , shadow_modulate.rgb #endif diff --git a/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl index 5c25235c58..9f89f4b3b7 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in highp vec3 vertex; @@ -32,7 +32,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(push_constant, binding = 0, std430) uniform Constants { mat4 projection; diff --git a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl index 302ad03b41..2bdfbabfcf 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl @@ -2,12 +2,12 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; layout(r8, set = 0, binding = 1) uniform restrict readonly image2D src_pixels; -layout(r16, set = 0, binding = 2) uniform restrict writeonly image2D dst_sdf; +layout(r16_snorm, set = 0, binding = 2) uniform restrict writeonly image2D dst_sdf; layout(rg16i, set = 0, binding = 3) uniform restrict readonly iimage2D src_process; layout(rg16i, set = 0, binding = 4) uniform restrict writeonly iimage2D dst_process; @@ -32,7 +32,7 @@ void main() { #ifdef MODE_LOAD bool solid = imageLoad(src_pixels, pos).r > 0.5; - imageStore(dst_process, pos, solid ? ivec4(pos, 0, 0) : ivec4(ivec2(32767), 0, 0)); + imageStore(dst_process, pos, solid ? ivec4(ivec2(-32767), 0, 0) : ivec4(ivec2(32767), 0, 0)); #endif #ifdef MODE_LOAD_SHRINK @@ -43,6 +43,8 @@ void main() { ivec2 rel = ivec2(32767); float d = 1e20; + int found = 0; + int solid_found = 0; for (int i = 0; i < s; i++) { for (int j = 0; j < s; j++) { ivec2 src_pos = base + ivec2(i, j); @@ -56,10 +58,17 @@ void main() { d = dist; rel = src_pos; } + solid_found++; } + found++; } } + if (solid_found == found) { + //mark solid only if all are solid + rel = ivec2(-32767); + } + imageStore(dst_process, pos, ivec4(rel, 0, 0)); #endif @@ -70,6 +79,12 @@ void main() { ivec2 rel = imageLoad(src_process, pos).xy; + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + if (center != rel) { //only process if it does not point to itself const int ofs_table_size = 8; @@ -92,6 +107,15 @@ void main() { continue; } ivec2 src_rel = imageLoad(src_process, src_pos).xy; + bool src_solid = src_rel.x < 0; + if (src_solid) { + src_rel = -src_rel - ivec2(1); + } + + if (src_solid != solid) { + src_rel = ivec2(src_pos << params.shift); //point to itself if of different type + } + float src_dist = length(vec2(src_rel - center)); if (src_dist < dist) { dist = src_dist; @@ -100,18 +124,31 @@ void main() { } } + if (solid) { + rel = -rel - ivec2(1); + } + imageStore(dst_process, pos, ivec4(rel, 0, 0)); #endif #ifdef MODE_STORE ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + float d = length(vec2(rel - pos)); - if (d > 0.01) { - d += 1.0; //make it signed + + if (solid) { + d = -d; } + d /= SDF_MAX_LENGTH; - d = clamp(d, 0.0, 1.0); + d = clamp(d, -1.0, 1.0); imageStore(dst_sdf, pos, vec4(d)); #endif @@ -122,13 +159,20 @@ void main() { ivec2 center = base + ivec2(params.shift); ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + float d = length(vec2(rel - center)); - if (d > 0.01) { - d += 1.0; //make it signed + if (solid) { + d = -d; } d /= SDF_MAX_LENGTH; - d = clamp(d, 0.0, 1.0); + d = clamp(d, -1.0, 1.0); imageStore(dst_sdf, pos, vec4(d)); #endif diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl index cf7678ea31..451f9b0089 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -5,12 +5,10 @@ #define SDF_MAX_LENGTH 16384.0 -#define FLAGS_INSTANCING_STRIDE_MASK 0xF -#define FLAGS_INSTANCING_ENABLED (1 << 4) -#define FLAGS_INSTANCING_HAS_COLORS (1 << 5) -#define FLAGS_INSTANCING_COLOR_8BIT (1 << 6) -#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 7) -#define FLAGS_INSTANCING_CUSTOM_DATA_8_BIT (1 << 8) +//1 means enabled, 2+ means trails in use +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) #define FLAGS_CLIP_RECT_UV (1 << 9) #define FLAGS_TRANSPOSE_RECT (1 << 10) diff --git a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl index 3a4bf4da07..8e616ebe1f 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl @@ -1,105 +1,3 @@ - #define CLUSTER_COUNTER_SHIFT 20 #define CLUSTER_POINTER_MASK ((1 << CLUSTER_COUNTER_SHIFT) - 1) #define CLUSTER_COUNTER_MASK 0xfff - -struct LightData { //this structure needs to be as packed as possible - vec3 position; - float inv_radius; - - vec3 direction; - float size; - - vec3 color; - float attenuation; - - float cone_attenuation; - float cone_angle; - float specular_amount; - bool shadow_enabled; - - vec4 atlas_rect; // rect in the shadow atlas - mat4 shadow_matrix; - float shadow_bias; - float shadow_normal_bias; - float transmittance_bias; - float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle - float soft_shadow_scale; // scales the shadow kernel for blurrier shadows - uint mask; - float shadow_volumetric_fog_fade; - uint pad; - vec4 projector_rect; //projector rect in srgb decal atlas -}; - -#define REFLECTION_AMBIENT_DISABLED 0 -#define REFLECTION_AMBIENT_ENVIRONMENT 1 -#define REFLECTION_AMBIENT_COLOR 2 - -struct ReflectionData { - vec3 box_extents; - float index; - vec3 box_offset; - uint mask; - vec3 ambient; // ambient color - float intensity; - bool exterior; - bool box_project; - uint ambient_mode; - uint pad; - //0-8 is intensity,8-9 is ambient, mode - mat4 local_matrix; // up to here for spot and omni, rest is for directional - // notes: for ambientblend, use distance to edge to blend between already existing global environment -}; - -struct DirectionalLightData { - vec3 direction; - float energy; - vec3 color; - float size; - float specular; - uint mask; - float softshadow_angle; - float soft_shadow_scale; - bool blend_splits; - bool shadow_enabled; - float fade_from; - float fade_to; - uvec3 pad; - float shadow_volumetric_fog_fade; - vec4 shadow_bias; - vec4 shadow_normal_bias; - vec4 shadow_transmittance_bias; - vec4 shadow_z_range; - vec4 shadow_range_begin; - vec4 shadow_split_offsets; - mat4 shadow_matrix1; - mat4 shadow_matrix2; - mat4 shadow_matrix3; - mat4 shadow_matrix4; - vec4 shadow_color1; - vec4 shadow_color2; - vec4 shadow_color3; - vec4 shadow_color4; - vec2 uv_scale1; - vec2 uv_scale2; - vec2 uv_scale3; - vec2 uv_scale4; -}; - -struct DecalData { - mat4 xform; //to decal transform - vec3 inv_extents; - float albedo_mix; - vec4 albedo_rect; - vec4 normal_rect; - vec4 orm_rect; - vec4 emission_rect; - vec4 modulate; - float emission_energy; - uint mask; - float upper_fade; - float lower_fade; - mat3x4 normal_xform; - vec3 normal; - float normal_fade; -}; diff --git a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl index 70a875192c..40da2c6e5c 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index ca92d2104e..da7d189281 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec3 vertex_attrib; @@ -63,7 +63,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) && defined(has_GL_KHR_shader_subgroup_vote) diff --git a/servers/rendering/renderer_rd/shaders/cluster_store.glsl b/servers/rendering/renderer_rd/shaders/cluster_store.glsl index 5be0893c4f..b0606efa94 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_store.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_store.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/copy.glsl b/servers/rendering/renderer_rd/shaders/copy.glsl index cdd35dfb3f..4110a95ddb 100644 --- a/servers/rendering/renderer_rd/shaders/copy.glsl +++ b/servers/rendering/renderer_rd/shaders/copy.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl index 9751e13b4e..8c68e2dc2f 100644 --- a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl +++ b/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) out vec2 uv_interp; @@ -37,7 +37,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(push_constant, binding = 1, std430) uniform Params { vec4 section; diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index c3ac0bee57..69b895ed29 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -2,13 +2,12 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(push_constant, binding = 1, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; @@ -26,7 +25,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec2 uv_interp; @@ -35,22 +34,23 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; void main() { vec2 uv = uv_interp; + vec2 texel_size = abs(params.texel_size); - vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + uv = clamp(uv * (1.0 + 2.0 * texel_size) - texel_size, vec2(0.0), vec2(1.0)); - normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 * (1.0 - dot(normal.xy, normal.xy)); // z = 1/2 - 1/2 * (x^2 + y^2) normal = normalize(normal); normal.y = -normal.y; //needs to be flipped to match projection matrix - if (!params.z_flip) { + if (params.texel_size.x >= 0.0) { // Sign is used to encode Z flip normal.z = -normal.z; } diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl index 7f269b7af3..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl @@ -22,7 +22,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 @@ -32,53 +32,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cubemap; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_size; -} -params; - -#define M_PI 3.14159265359 - -void get_dir_0(out vec3 dir, in float u, in float v) { - dir[0] = 1.0; - dir[1] = v; - dir[2] = -u; -} - -void get_dir_1(out vec3 dir, in float u, in float v) { - dir[0] = -1.0; - dir[1] = v; - dir[2] = u; -} - -void get_dir_2(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = 1.0; - dir[2] = -v; -} - -void get_dir_3(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = -1.0; - dir[2] = v; -} - -void get_dir_4(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = v; - dir[2] = 1.0; -} - -void get_dir_5(out vec3 dir, in float u, in float v) { - dir[0] = -u; - dir[1] = v; - dir[2] = -1.0; -} - -float calcWeight(float u, float v) { - float val = u * u + v * v + 1.0; - return val * sqrt(val); -} +#include "cubemap_downsampler_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl new file mode 100644 index 0000000000..b329e67293 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl @@ -0,0 +1,48 @@ +layout(push_constant, binding = 1, std430) uniform Params { + uint face_size; + uint face_id; // only used in raster shader +} +params; + +#define M_PI 3.14159265359 + +void get_dir_0(out vec3 dir, in float u, in float v) { + dir[0] = 1.0; + dir[1] = v; + dir[2] = -u; +} + +void get_dir_1(out vec3 dir, in float u, in float v) { + dir[0] = -1.0; + dir[1] = v; + dir[2] = u; +} + +void get_dir_2(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = 1.0; + dir[2] = -v; +} + +void get_dir_3(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = -1.0; + dir[2] = v; +} + +void get_dir_4(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = v; + dir[2] = 1.0; +} + +void get_dir_5(out vec3 dir, in float u, in float v) { + dir[0] = -u; + dir[1] = v; + dir[2] = -1.0; +} + +float calcWeight(float u, float v) { + float val = u * u + v * v + 1.0; + return val * sqrt(val); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl new file mode 100644 index 0000000000..0828ffd921 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl @@ -0,0 +1,163 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex] * float(params.face_size); + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + // Converted from compute shader which uses absolute coordinates. + // Could possibly simplify this + float face_size = float(params.face_size); + + if (uv_interp.x < face_size && uv_interp.y < face_size) { + float inv_face_size = 1.0 / face_size; + + float u0 = (uv_interp.x * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0; + float u1 = (uv_interp.x * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0; + + float v0 = (uv_interp.y * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0; + float v1 = (uv_interp.y * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0; + + float weights[4]; + weights[0] = calcWeight(u0, v0); + weights[1] = calcWeight(u1, v0); + weights[2] = calcWeight(u0, v1); + weights[3] = calcWeight(u1, v1); + + const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]); + for (int i = 0; i < 4; i++) { + weights[i] = weights[i] * wsum + .125; + } + + vec3 dir; + vec4 color; + switch (params.face_id) { + case 0: + get_dir_0(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_0(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_0(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_0(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 1: + get_dir_1(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_1(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_1(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_1(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 2: + get_dir_2(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_2(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_2(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_2(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 3: + get_dir_3(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_3(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_3(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_3(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 4: + get_dir_4(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_4(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_4(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_4(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + default: + get_dir_5(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_5(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_5(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_5(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + } + frag_color = color; + } +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl index 987545fb76..2a774b0eb4 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl @@ -22,7 +22,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define GROUP_SIZE 64 diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl new file mode 100644 index 0000000000..324d306218 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl @@ -0,0 +1,256 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 1, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 1, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; + +/* clang-format on */ + +#ifdef USE_HIGH_QUALITY +#define NUM_TAPS 32 +#else +#define NUM_TAPS 8 +#endif + +#define BASE_RESOLUTION 128 + +#ifdef USE_HIGH_QUALITY +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][3][24] coeffs; +} +data; +#else +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][6] coeffs; +} +data; +#endif + +void get_dir(out vec3 dir, in vec2 uv, in uint face) { + switch (face) { + case 0: + dir = vec3(1.0, uv[1], -uv[0]); + break; + case 1: + dir = vec3(-1.0, uv[1], uv[0]); + break; + case 2: + dir = vec3(uv[0], 1.0, -uv[1]); + break; + case 3: + dir = vec3(uv[0], -1.0, uv[1]); + break; + case 4: + dir = vec3(uv[0], uv[1], 1.0); + break; + default: + dir = vec3(-uv[0], uv[1], -1.0); + break; + } +} + +void main() { + // determine dir / pos for the texel + vec3 dir, adir, frameZ; + { + vec2 uv; + uv.x = uv_interp.x; + uv.y = 1.0 - uv_interp.y; + uv = uv * 2.0 - 1.0; + + get_dir(dir, uv, params.face_id); + frameZ = normalize(dir); + + adir = abs(dir); + } + + // determine which texel this is + // NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name. + int mip_level = 0; + + if (params.mip_level < 0) { + // return as is + frag_color.rgb = textureLod(source_cubemap, frameZ, 0.0).rgb; + frag_color.a = 1.0; + return; + } else if (params.mip_level > 6) { + // maximum level + mip_level = 6; + } else { + mip_level = params.mip_level; + } + + // GGX gather colors + vec4 color = vec4(0.0); + for (int axis = 0; axis < 3; axis++) { + const int otherAxis0 = 1 - (axis & 1) - (axis >> 1); + const int otherAxis1 = 2 - (axis >> 1); + + float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25; + if (frameweight > 0.0) { + // determine frame + vec3 UpVector; + switch (axis) { + case 0: + UpVector = vec3(1, 0, 0); + break; + case 1: + UpVector = vec3(0, 1, 0); + break; + default: + UpVector = vec3(0, 0, 1); + break; + } + + vec3 frameX = normalize(cross(UpVector, frameZ)); + vec3 frameY = cross(frameZ, frameX); + + // calculate parametrization for polynomial + float Nx = dir[otherAxis0]; + float Ny = dir[otherAxis1]; + float Nz = adir[axis]; + + float NmaxXY = max(abs(Ny), abs(Nx)); + Nx /= NmaxXY; + Ny /= NmaxXY; + + float theta; + if (Ny < Nx) { + if (Ny <= -0.999) + theta = Nx; + else + theta = Ny; + } else { + if (Ny >= 0.999) + theta = -Nx; + else + theta = -Ny; + } + + float phi; + if (Nz <= -0.999) + phi = -NmaxXY; + else if (Nz >= 0.999) + phi = NmaxXY; + else + phi = Nz; + + float theta2 = theta * theta; + float phi2 = phi * phi; + + // sample + for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) { + const int index = (NUM_TAPS / 4) * axis + iSuperTap; + +#ifdef USE_HIGH_QUALITY + vec4 coeffsDir0[3]; + vec4 coeffsDir1[3]; + vec4 coeffsDir2[3]; + vec4 coeffsLevel[3]; + vec4 coeffsWeight[3]; + + for (int iCoeff = 0; iCoeff < 3; iCoeff++) { + coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index]; + coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index]; + coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index]; + coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index]; + coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index]; + } + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2); + + float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2; + + float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2; +#else + vec4 coeffsDir0 = data.coeffs[mip_level][0][index]; + vec4 coeffsDir1 = data.coeffs[mip_level][1][index]; + vec4 coeffsDir2 = data.coeffs[mip_level][2][index]; + vec4 coeffsLevel = data.coeffs[mip_level][3][index]; + vec4 coeffsWeight = data.coeffs[mip_level][4][index]; + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap]; + + float sample_level = coeffsLevel[iSubTap]; + + float sample_weight = coeffsWeight[iSubTap]; +#endif + + sample_weight *= frameweight; + + // adjust for jacobian + sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2]))); + sample_level += 0.75 * log2(dot(sample_dir, sample_dir)); + // sample cubemap + color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight; + color.w += sample_weight; + } + } + } + } + color /= color.w; + + // write color + color.xyz = max(vec3(0.0), color.xyz); + color.w = 1.0; + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl index 5cbb00baa4..28f4dc59ec 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define GROUP_SIZE 8 @@ -12,100 +12,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_id; - uint sample_count; - float roughness; - bool use_direct_write; - float face_size; -} -params; - -#define M_PI 3.14159265359 - -vec3 texelCoordToVec(vec2 uv, uint faceID) { - mat3 faceUvVectors[6]; - - // -x - faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z - faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face - - // +x - faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z - faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face - - // -y - faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z - faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face - - // +y - faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z - faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face - - // -z - faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x - faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face - - // +z - faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face - - // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. - vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; - return normalize(result); -} - -vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) { - float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph] - - // Compute distribution direction - float Phi = 2.0 * M_PI * Xi.x; - float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y)); - float SinTheta = sqrt(1.0 - CosTheta * CosTheta); - - // Convert to spherical direction - vec3 H; - H.x = SinTheta * cos(Phi); - H.y = SinTheta * sin(Phi); - H.z = CosTheta; - - vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); - vec3 TangentX = normalize(cross(UpVector, N)); - vec3 TangentY = cross(N, TangentX); - - // Tangent to world space - return TangentX * H.x + TangentY * H.y + N * H.z; -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float GGX(float NdotV, float a) { - float k = a / 2.0; - return NdotV / (NdotV * (1.0 - k) + k); -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float G_Smith(float a, float nDotV, float nDotL) { - return GGX(nDotL, a * a) * GGX(nDotV, a * a); -} - -float radicalInverse_VdC(uint bits) { - bits = (bits << 16u) | (bits >> 16u); - bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); - bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); - bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); - bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); - return float(bits) * 2.3283064365386963e-10; // / 0x100000000 -} - -vec2 Hammersley(uint i, uint N) { - return vec2(float(i) / float(N), radicalInverse_VdC(i)); -} +#include "cubemap_roughness_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl new file mode 100644 index 0000000000..80c0ac4fb4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl @@ -0,0 +1,94 @@ +#define M_PI 3.14159265359 + +layout(push_constant, binding = 1, std430) uniform Params { + uint face_id; + uint sample_count; + float roughness; + bool use_direct_write; + float face_size; +} +params; + +vec3 texelCoordToVec(vec2 uv, uint faceID) { + mat3 faceUvVectors[6]; + + // -x + faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z + faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face + + // +x + faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z + faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face + + // -y + faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z + faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face + + // +y + faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z + faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face + + // -z + faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x + faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face + + // +z + faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face + + // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. + vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; + return normalize(result); +} + +vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) { + float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph] + + // Compute distribution direction + float Phi = 2.0 * M_PI * Xi.x; + float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y)); + float SinTheta = sqrt(1.0 - CosTheta * CosTheta); + + // Convert to spherical direction + vec3 H; + H.x = SinTheta * cos(Phi); + H.y = SinTheta * sin(Phi); + H.z = CosTheta; + + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + vec3 TangentX = normalize(cross(UpVector, N)); + vec3 TangentY = cross(N, TangentX); + + // Tangent to world space + return TangentX * H.x + TangentY * H.y + N * H.z; +} + +// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float GGX(float NdotV, float a) { + float k = a / 2.0; + return NdotV / (NdotV * (1.0 - k) + k); +} + +// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float G_Smith(float a, float nDotV, float nDotL) { + return GGX(nDotL, a * a) * GGX(nDotV, a * a); +} + +float radicalInverse_VdC(uint bits) { + bits = (bits << 16u) | (bits >> 16u); + bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); + bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); + bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); + bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); + return float(bits) * 2.3283064365386963e-10; // / 0x100000000 +} + +vec2 Hammersley(uint i, uint N) { + return vec2(float(i) / float(N), radicalInverse_VdC(i)); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl new file mode 100644 index 0000000000..2570308816 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl @@ -0,0 +1,63 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) in vec2 uv_interp; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + vec3 N = texelCoordToVec(uv_interp * 2.0 - 1.0, params.face_id); + + //vec4 color = color_interp; + + if (params.use_direct_write) { + frag_color = vec4(texture(source_cube, N).rgb, 1.0); + } else { + vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { + vec2 xi = Hammersley(sampleNum, params.sample_count); + + vec3 H = ImportanceSampleGGX(xi, params.roughness, N); + vec3 V = N; + vec3 L = (2.0 * dot(V, H) * H - V); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + sum.rgb += textureLod(source_cube, L, 0.0).rgb * ndotl; + sum.a += ndotl; + } + } + sum /= sum.a; + + frag_color = vec4(sum.rgb, 1.0); + } +} diff --git a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl new file mode 100644 index 0000000000..158096d3c7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl @@ -0,0 +1,18 @@ + +struct DecalData { + highp mat4 xform; //to decal transform + highp vec3 inv_extents; + mediump float albedo_mix; + highp vec4 albedo_rect; + highp vec4 normal_rect; + highp vec4 orm_rect; + highp vec4 emission_rect; + highp vec4 modulate; + mediump float emission_energy; + uint mask; + mediump float upper_fade; + mediump float lower_fade; + mediump mat3x4 normal_xform; + mediump vec3 normal; + mediump float normal_fade; +}; diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index 92a5682572..60c881881d 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -35,7 +35,7 @@ layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture; layout(set = 0, binding = 12) uniform texture2D depth_buffer; layout(set = 0, binding = 13) uniform texture2D normal_roughness_buffer; -layout(set = 0, binding = 14) uniform utexture2D giprobe_buffer; +layout(set = 0, binding = 14) uniform utexture2D voxel_gi_buffer; layout(set = 0, binding = 15, std140) uniform SDFGI { vec3 grid_size; @@ -65,9 +65,9 @@ layout(set = 0, binding = 15, std140) uniform SDFGI { } sdfgi; -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 -struct GIProbeData { +struct VoxelGIData { mat4 xform; vec3 bounds; float dynamic_range; @@ -77,18 +77,18 @@ struct GIProbeData { bool blend_ambient; uint texture_slot; - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; + uint pad0; + uint pad1; + uint pad2; uint mipmaps; }; -layout(set = 0, binding = 16, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; +layout(set = 0, binding = 16, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } -gi_probes; +voxel_gi_instances; -layout(set = 0, binding = 17) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +layout(set = 0, binding = 17) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; layout(push_constant, binding = 0, std430) uniform Params { ivec2 screen_size; @@ -98,7 +98,7 @@ layout(push_constant, binding = 0, std430) uniform Params { vec4 proj_info; vec3 ao_color; - uint max_giprobes; + uint max_voxel_gi_instances; bool high_quality_vct; bool orthogonal; @@ -155,7 +155,7 @@ vec3 reconstruct_position(ivec2 screen_pos) { return pos; } -void sdfgi_probe_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { +void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { cascade_pos += cam_normal * sdfgi.normal_bias; vec3 base_pos = floor(cascade_pos); @@ -293,7 +293,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o float blend; vec3 diffuse, specular; - sdfgi_probe_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular); + sdfvoxel_gi_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular); { //process blend @@ -323,7 +323,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } else { vec3 diffuse2, specular2; cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe; - sdfgi_probe_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2); + sdfvoxel_gi_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2); diffuse = mix(diffuse, diffuse2, blend); specular = mix(specular, specular2, blend); } @@ -494,26 +494,26 @@ vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 return color; } -void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) { - position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz; - ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz); - normal = normalize((gi_probes.data[index].xform * vec4(normal, 0.0)).xyz); +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); - position += normal * gi_probes.data[index].normal_bias; + position += normal * voxel_gi_instances.data[index].normal_bias; //this causes corrupted pixels, i have no idea why.. - if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, gi_probes.data[index].bounds))))) { + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { return; } - mat3 dir_xform = mat3(gi_probes.data[index].xform) * normal_xform; + mat3 dir_xform = mat3(voxel_gi_instances.data[index].xform) * normal_xform; - vec3 blendv = abs(position / gi_probes.data[index].bounds * 2.0 - 1.0); + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); //float blend=1.0; - float max_distance = length(gi_probes.data[index].bounds); - vec3 cell_size = 1.0 / gi_probes.data[index].bounds; + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; //irradiance @@ -534,7 +534,7 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 for (uint i = 0; i < cone_dir_count; i++) { vec3 dir = normalize(dir_xform * cone_dirs[i]); - light += cone_weights[i] * voxel_cone_trace(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias); + light += cone_weights[i] * voxel_cone_trace(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); } } else { const uint cone_dir_count = 4; @@ -547,42 +547,21 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 float cone_weights[cone_dir_count] = float[](0.25, 0.25, 0.25, 0.25); for (int i = 0; i < cone_dir_count; i++) { vec3 dir = normalize(dir_xform * cone_dirs[i]); - light += cone_weights[i] * voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, max_distance, gi_probes.data[index].bias); + light += cone_weights[i] * voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, max_distance, voxel_gi_instances.data[index].bias); } } - if (gi_probes.data[index].ambient_occlusion > 0.001) { - float size = 1.0 + gi_probes.data[index].ambient_occlusion_size * 7.0; - - float taps, blend; - blend = modf(size, taps); - float ao = 0.0; - for (float i = 1.0; i <= taps; i++) { - vec3 ofs = (position + normal * (i * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, i - 1.0).a * i; - } - - if (blend > 0.001) { - vec3 ofs = (position + normal * ((taps + 1.0) * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, taps).a * (taps + 1.0) * blend; - } - - ao = 1.0 - min(1.0, ao); - - light.rgb = mix(params.ao_color, light.rgb, mix(1.0, ao, gi_probes.data[index].ambient_occlusion)); - } - - light.rgb *= gi_probes.data[index].dynamic_range; - if (!gi_probes.data[index].blend_ambient) { + light.rgb *= voxel_gi_instances.data[index].dynamic_range; + if (!voxel_gi_instances.data[index].blend_ambient) { light.a = 1.0; } out_diff += light * blend; //radiance - vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias); - irr_light.rgb *= gi_probes.data[index].dynamic_range; - if (!gi_probes.data[index].blend_ambient) { + vec4 irr_light = voxel_cone_trace(voxel_gi_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, voxel_gi_instances.data[index].bias); + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range; + if (!voxel_gi_instances.data[index].blend_ambient) { irr_light.a = 1.0; } @@ -614,9 +593,9 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); #endif -#ifdef USE_GIPROBES +#ifdef USE_VOXEL_GI_INSTANCES { - uvec2 giprobe_tex = texelFetch(usampler2D(giprobe_buffer, linear_sampler), pos, 0).rg; + uvec2 voxel_gi_tex = texelFetch(usampler2D(voxel_gi_buffer, linear_sampler), pos, 0).rg; roughness *= roughness; //find arbitrary tangent and bitangent, then build a matrix vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); @@ -628,9 +607,9 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref vec4 spec_accum = vec4(0.0); float blend_accum = 0.0; - for (uint i = 0; i < params.max_giprobes; i++) { - if (any(equal(uvec2(i), giprobe_tex))) { - gi_probe_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum); + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + if (any(equal(uvec2(i), voxel_gi_tex))) { + voxel_gi_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum); } } if (blend_accum > 0.0) { diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 56b3b7ccb4..5dc2d08a3b 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl new file mode 100644 index 0000000000..fdc7729338 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -0,0 +1,87 @@ +#define LIGHT_BAKE_DISABLED 0 +#define LIGHT_BAKE_DYNAMIC 1 +#define LIGHT_BAKE_STATIC 2 + +struct LightData { //this structure needs to be as packed as possible + highp vec3 position; + highp float inv_radius; + + mediump vec3 direction; + highp float size; + + mediump vec3 color; + mediump float attenuation; + + mediump float cone_attenuation; + mediump float cone_angle; + mediump float specular_amount; + bool shadow_enabled; + + highp vec4 atlas_rect; // rect in the shadow atlas + highp mat4 shadow_matrix; + highp float shadow_bias; + highp float shadow_normal_bias; + highp float transmittance_bias; + highp float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle + highp float soft_shadow_scale; // scales the shadow kernel for blurrier shadows + uint mask; + mediump float shadow_volumetric_fog_fade; + uint bake_mode; + highp vec4 projector_rect; //projector rect in srgb decal atlas +}; + +#define REFLECTION_AMBIENT_DISABLED 0 +#define REFLECTION_AMBIENT_ENVIRONMENT 1 +#define REFLECTION_AMBIENT_COLOR 2 + +struct ReflectionData { + highp vec3 box_extents; + mediump float index; + highp vec3 box_offset; + uint mask; + mediump vec3 ambient; // ambient color + mediump float intensity; + bool exterior; + bool box_project; + uint ambient_mode; + uint pad; + //0-8 is intensity,8-9 is ambient, mode + highp mat4 local_matrix; // up to here for spot and omni, rest is for directional + // notes: for ambientblend, use distance to edge to blend between already existing global environment +}; + +struct DirectionalLightData { + mediump vec3 direction; + mediump float energy; + mediump vec3 color; + mediump float size; + mediump float specular; + uint mask; + highp float softshadow_angle; + highp float soft_shadow_scale; + bool blend_splits; + bool shadow_enabled; + highp float fade_from; + highp float fade_to; + uvec2 pad; + uint bake_mode; + mediump float shadow_volumetric_fog_fade; + highp vec4 shadow_bias; + highp vec4 shadow_normal_bias; + highp vec4 shadow_transmittance_bias; + highp vec4 shadow_z_range; + highp vec4 shadow_range_begin; + highp vec4 shadow_split_offsets; + highp mat4 shadow_matrix1; + highp mat4 shadow_matrix2; + highp mat4 shadow_matrix3; + highp mat4 shadow_matrix4; + mediump vec4 shadow_color1; + mediump vec4 shadow_color2; + mediump vec4 shadow_color3; + mediump vec4 shadow_color4; + highp vec2 uv_scale1; + highp vec2 uv_scale2; + highp vec2 uv_scale3; + highp vec2 uv_scale4; +}; diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl index 8a11c35b78..466442b67a 100644 --- a/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl new file mode 100644 index 0000000000..29ebd74a90 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl @@ -0,0 +1,74 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_exposure; + +#ifdef FINAL_PASS +layout(set = 1, binding = 0) uniform sampler2D prev_luminance; +#endif + +layout(location = 0) out highp float luminance; + +void main() { + ivec2 dest_pos = ivec2(uv_interp * settings.dest_size); + ivec2 src_pos = ivec2(uv_interp * settings.source_size); + + ivec2 next_pos = (dest_pos + ivec2(1)) * settings.source_size / settings.dest_size; + next_pos = max(next_pos, src_pos + ivec2(1)); //so it at least reads one pixel + + highp vec3 source_color = vec3(0.0); + for (int i = src_pos.x; i < next_pos.x; i++) { + for (int j = src_pos.y; j < next_pos.y; j++) { + source_color += texelFetch(source_exposure, ivec2(i, j), 0).rgb; + } + } + + source_color /= float((next_pos.x - src_pos.x) * (next_pos.y - src_pos.y)); + +#ifdef FIRST_PASS + luminance = max(source_color.r, max(source_color.g, source_color.b)); + + // This formula should be more "accurate" but gave an overexposed result when testing. + // Leaving it here so we can revisit it if we want. + // luminance = source_color.r * 0.21 + source_color.g * 0.71 + source_color.b * 0.07; +#else + luminance = source_color.r; +#endif + +#ifdef FINAL_PASS + // Obtain our target luminance + luminance = clamp(luminance, settings.min_luminance, settings.max_luminance); + + // Now smooth to our transition + highp float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous luminance + luminance = prev_lum + (luminance - prev_lum) * clamp(settings.exposure_adjust, 0.0, 1.0); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl new file mode 100644 index 0000000000..ed389ffe56 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl @@ -0,0 +1,11 @@ + +layout(push_constant, binding = 1, std430) uniform PushConstant { + ivec2 source_size; + ivec2 dest_size; + + float exposure_adjust; + float min_luminance; + float max_luminance; + float pad; +} +settings; diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index cb6d8dc7f6..9f8410fd8a 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -19,6 +19,8 @@ layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 +#define SDF_MAX_LENGTH 16384.0 + /* SET 0: GLOBAL DATA */ layout(set = 0, binding = 1) uniform sampler material_samplers[12]; @@ -54,6 +56,7 @@ struct Attractor { #define COLLIDER_TYPE_BOX 1 #define COLLIDER_TYPE_SDF 2 #define COLLIDER_TYPE_HEIGHT_FIELD 3 +#define COLLIDER_TYPE_2D_SDF 4 struct Collider { mat4 transform; @@ -76,6 +79,11 @@ struct FrameParams { float time; float delta; + uint frame; + uint pad0; + uint pad1; + uint pad2; + uint random_seed; uint attractor_count; uint collider_count; @@ -92,10 +100,16 @@ layout(set = 1, binding = 0, std430) restrict buffer FrameHistory { } frame_history; +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) +#define PARTICLE_FRAME_MASK uint(0xFFFF) +#define PARTICLE_FRAME_SHIFT uint(16) + struct ParticleData { mat4 xform; vec3 velocity; - bool is_active; + uint flags; vec4 color; vec4 custom; }; @@ -146,11 +160,11 @@ layout(set = 2, binding = 1) uniform texture2D height_field_texture; /* SET 3: MATERIAL */ -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 3, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif @@ -162,7 +176,7 @@ layout(push_constant, binding = 0, std430) uniform Params { bool use_fractional_delta; bool sub_emitter_mode; bool can_emit; - uint pad; + bool trail_pass; } params; @@ -196,15 +210,19 @@ bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom return true; } -/* clang-format off */ - -COMPUTE_SHADER_GLOBALS - -/* clang-format on */ +#GLOBALS void main() { uint particle = gl_GlobalInvocationID.x; + if (params.trail_size > 1) { + if (params.trail_pass) { + particle += (particle / (params.trail_size - 1)) + 1; + } else { + particle *= params.trail_size; + } + } + if (particle >= params.total_particles * params.trail_size) { return; //discard } @@ -233,7 +251,7 @@ void main() { PARTICLE.color = vec4(1.0); PARTICLE.custom = vec4(0.0); PARTICLE.velocity = vec3(0.0); - PARTICLE.is_active = false; + PARTICLE.flags = 0; PARTICLE.xform = mat4( vec4(1.0, 0.0, 0.0, 0.0), vec4(0.0, 1.0, 0.0, 0.0), @@ -241,6 +259,29 @@ void main() { vec4(0.0, 0.0, 0.0, 1.0)); } + //clear started flag if set + + if (params.trail_pass) { + //trail started + uint src_idx = index * params.trail_size; + if (bool(particles.data[src_idx].flags & PARTICLE_FLAG_STARTED)) { + //save start conditions for trails + PARTICLE.color = particles.data[src_idx].color; + PARTICLE.custom = particles.data[src_idx].custom; + PARTICLE.velocity = particles.data[src_idx].velocity; + PARTICLE.flags = PARTICLE_FLAG_TRAILED | ((frame_history.data[0].frame & PARTICLE_FRAME_MASK) << PARTICLE_FRAME_SHIFT); //mark it as trailed, save in which frame it will start + PARTICLE.xform = particles.data[src_idx].xform; + } + + if (bool(PARTICLE.flags & PARTICLE_FLAG_TRAILED) && ((PARTICLE.flags >> PARTICLE_FRAME_SHIFT) == (FRAME.frame & PARTICLE_FRAME_MASK))) { //check this is trailed and see if it should start now + // we just assume that this is the first frame of the particle, the rest is deterministic + PARTICLE.flags = PARTICLE_FLAG_ACTIVE | (particles.data[src_idx].flags & (PARTICLE_FRAME_MASK << PARTICLE_FRAME_SHIFT)); + return; //- this appears like it should be correct, but it seems not to be.. wonder why. + } + } else { + PARTICLE.flags &= ~PARTICLE_FLAG_STARTED; + } + bool collided = false; vec3 collision_normal = vec3(0.0); float collision_depth = 0.0; @@ -249,197 +290,17 @@ void main() { #if !defined(DISABLE_VELOCITY) - if (PARTICLE.is_active) { + if (bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { PARTICLE.xform[3].xyz += PARTICLE.velocity * local_delta; } #endif - /* Process physics if active */ - - if (PARTICLE.is_active) { - for (uint i = 0; i < FRAME.attractor_count; i++) { - vec3 dir; - float amount; - vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.attractors[i].transform[3].xyz; - vec3 local_pos = rel_vec * mat3(FRAME.attractors[i].transform); - - switch (FRAME.attractors[i].type) { - case ATTRACTOR_TYPE_SPHERE: { - dir = normalize(rel_vec); - float d = length(local_pos) / FRAME.attractors[i].extents.x; - if (d > 1.0) { - continue; - } - amount = max(0.0, 1.0 - d); - } break; - case ATTRACTOR_TYPE_BOX: { - dir = normalize(rel_vec); - - vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); - float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); - if (d > 1.0) { - continue; - } - amount = max(0.0, 1.0 - d); - - } break; - case ATTRACTOR_TYPE_VECTOR_FIELD: { - vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents) * 2.0 - 1.0; - if (any(lessThan(uvw_pos, vec3(0.0))) || any(greaterThan(uvw_pos, vec3(1.0)))) { - continue; - } - vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz; - dir = mat3(FRAME.attractors[i].transform) * normalize(s); //revert direction - amount = length(s); - - } break; - } - amount = pow(amount, FRAME.attractors[i].attenuation); - dir = normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); - attractor_force -= amount * dir * FRAME.attractors[i].strength; - } - - float particle_size = FRAME.particle_size; - -#ifdef USE_COLLISON_SCALE - - particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); - -#endif - - for (uint i = 0; i < FRAME.collider_count; i++) { - vec3 normal; - float depth; - bool col = false; - - vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; - vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); - - switch (FRAME.colliders[i].type) { - case COLLIDER_TYPE_SPHERE: { - float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); - - if (d < 0.0) { - col = true; - depth = -d; - normal = normalize(rel_vec); - } - - } break; - case COLLIDER_TYPE_BOX: { - vec3 abs_pos = abs(local_pos); - vec3 sgn_pos = sign(local_pos); - - if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { - //point outside box - - vec3 closest = min(abs_pos, FRAME.colliders[i].extents); - vec3 rel = abs_pos - closest; - depth = length(rel) - particle_size; - if (depth < 0.0) { - col = true; - normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); - depth = -depth; - } - } else { - //point inside box - vec3 axis_len = FRAME.colliders[i].extents - abs_pos; - // there has to be a faster way to do this? - if (all(lessThan(axis_len.xx, axis_len.yz))) { - normal = vec3(1, 0, 0); - } else if (all(lessThan(axis_len.yy, axis_len.xz))) { - normal = vec3(0, 1, 0); - } else { - normal = vec3(0, 0, 1); - } - - col = true; - depth = dot(normal * axis_len, vec3(1)) + particle_size; - normal = mat3(FRAME.colliders[i].transform) * (normal * sgn_pos); - } - - } break; - case COLLIDER_TYPE_SDF: { - vec3 apos = abs(local_pos); - float extra_dist = 0.0; - if (any(greaterThan(apos, FRAME.colliders[i].extents))) { //outside - vec3 mpos = min(apos, FRAME.colliders[i].extents); - extra_dist = distance(mpos, apos); - } - - if (extra_dist > particle_size) { - continue; - } - - vec3 uvw_pos = (local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5; - float s = texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).r; - s *= FRAME.colliders[i].scale; - s += extra_dist; - if (s < particle_size) { - col = true; - depth = particle_size - s; - const float EPSILON = 0.001; - normal = mat3(FRAME.colliders[i].transform) * - normalize( - vec3( - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, 0.0, EPSILON)).r)); - } - - } break; - case COLLIDER_TYPE_HEIGHT_FIELD: { - vec3 local_pos_bottom = local_pos; - local_pos_bottom.y -= particle_size; - - if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { - continue; - } - - const float DELTA = 1.0 / 8192.0; - - vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; - - float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; - - if (y > uvw_pos.y) { - //inside heightfield - - vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; - vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; - vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * FRAME.colliders[i].extents; - - normal = normalize(cross(pos1 - pos2, pos1 - pos3)); - float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; - - col = true; - depth = dot(normal, pos1) - dot(normal, local_pos_bottom); - } - - } break; - } - - if (col) { - if (!collided) { - collided = true; - collision_normal = normal; - collision_depth = depth; - } else { - vec3 c = collision_normal * collision_depth; - c += normal * max(0.0, depth - dot(normal, c)); - collision_normal = normalize(c); - collision_depth = length(c); - } - } - } - } - - if (params.sub_emitter_mode) { - if (!PARTICLE.is_active) { + if (!params.trail_pass && params.sub_emitter_mode) { + if (!bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { int src_index = atomicAdd(src_particles.particle_count, -1) - 1; if (src_index >= 0) { - PARTICLE.is_active = true; + PARTICLE.flags = (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)); restart = true; if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_POSITION)) { @@ -521,16 +382,12 @@ void main() { } } - uint current_cycle = FRAME.cycle; - - if (FRAME.system_phase < restart_phase) { - current_cycle -= uint(1); + if (params.trail_pass) { + restart = false; } - uint particle_number = current_cycle * uint(params.total_particles) + particle; - if (restart) { - PARTICLE.is_active = FRAME.emitting; + PARTICLE.flags = FRAME.emitting ? (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)) : 0; restart_position = true; restart_rotation_scale = true; restart_velocity = true; @@ -539,11 +396,237 @@ void main() { } } - if (PARTICLE.is_active) { - /* clang-format off */ + bool particle_active = bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE); + + uint particle_number = (PARTICLE.flags >> PARTICLE_FRAME_SHIFT) * uint(params.total_particles) + index; + + if (restart && particle_active) { +#CODE : START + } + + if (particle_active) { + for (uint i = 0; i < FRAME.attractor_count; i++) { + vec3 dir; + float amount; + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.attractors[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.attractors[i].transform); + + switch (FRAME.attractors[i].type) { + case ATTRACTOR_TYPE_SPHERE: { + dir = normalize(rel_vec); + float d = length(local_pos) / FRAME.attractors[i].extents.x; + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + } break; + case ATTRACTOR_TYPE_BOX: { + dir = normalize(rel_vec); + + vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); + float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + + } break; + case ATTRACTOR_TYPE_VECTOR_FIELD: { + vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents) * 2.0 - 1.0; + if (any(lessThan(uvw_pos, vec3(0.0))) || any(greaterThan(uvw_pos, vec3(1.0)))) { + continue; + } + vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz; + dir = mat3(FRAME.attractors[i].transform) * normalize(s); //revert direction + amount = length(s); + + } break; + } + amount = pow(amount, FRAME.attractors[i].attenuation); + dir = normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); + attractor_force -= amount * dir * FRAME.attractors[i].strength; + } + + float particle_size = FRAME.particle_size; + +#ifdef USE_COLLISON_SCALE + + particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); + +#endif + + if (FRAME.collider_count == 1 && FRAME.colliders[0].type == COLLIDER_TYPE_2D_SDF) { + //2D collision + + vec2 pos = PARTICLE.xform[3].xy; + vec4 to_sdf_x = FRAME.colliders[0].transform[0]; + vec4 to_sdf_y = FRAME.colliders[0].transform[1]; + vec2 sdf_pos = vec2(dot(vec4(pos, 0, 1), to_sdf_x), dot(vec4(pos, 0, 1), to_sdf_y)); + + vec4 sdf_to_screen = vec4(FRAME.colliders[0].extents, FRAME.colliders[0].scale); + + vec2 uv_pos = sdf_pos * sdf_to_screen.xy + sdf_to_screen.zw; + + if (all(greaterThan(uv_pos, vec2(0.0))) && all(lessThan(uv_pos, vec2(1.0)))) { + vec2 pos2 = pos + vec2(0, particle_size); + vec2 sdf_pos2 = vec2(dot(vec4(pos2, 0, 1), to_sdf_x), dot(vec4(pos2, 0, 1), to_sdf_y)); + float sdf_particle_size = distance(sdf_pos, sdf_pos2); + + float d = texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos).r * SDF_MAX_LENGTH; + + d -= sdf_particle_size; + + if (d < 0.0) { + const float EPSILON = 0.001; + vec2 n = normalize(vec2( + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(EPSILON, 0.0)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(EPSILON, 0.0)).r, + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(0.0, EPSILON)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(0.0, EPSILON)).r)); + + collided = true; + sdf_pos2 = sdf_pos + n * d; + pos2 = vec2(dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[2]), dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[3])); + + n = pos - pos2; + + collision_normal = normalize(vec3(n, 0.0)); + collision_depth = length(n); + } + } + + } else { + for (uint i = 0; i < FRAME.collider_count; i++) { + vec3 normal; + float depth; + bool col = false; + + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); + + switch (FRAME.colliders[i].type) { + case COLLIDER_TYPE_SPHERE: { + float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); + + if (d < 0.0) { + col = true; + depth = -d; + normal = normalize(rel_vec); + } + + } break; + case COLLIDER_TYPE_BOX: { + vec3 abs_pos = abs(local_pos); + vec3 sgn_pos = sign(local_pos); + + if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { + //point outside box + + vec3 closest = min(abs_pos, FRAME.colliders[i].extents); + vec3 rel = abs_pos - closest; + depth = length(rel) - particle_size; + if (depth < 0.0) { + col = true; + normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); + depth = -depth; + } + } else { + //point inside box + vec3 axis_len = FRAME.colliders[i].extents - abs_pos; + // there has to be a faster way to do this? + if (all(lessThan(axis_len.xx, axis_len.yz))) { + normal = vec3(1, 0, 0); + } else if (all(lessThan(axis_len.yy, axis_len.xz))) { + normal = vec3(0, 1, 0); + } else { + normal = vec3(0, 0, 1); + } + + col = true; + depth = dot(normal * axis_len, vec3(1)) + particle_size; + normal = mat3(FRAME.colliders[i].transform) * (normal * sgn_pos); + } -COMPUTE_SHADER_CODE + } break; + case COLLIDER_TYPE_SDF: { + vec3 apos = abs(local_pos); + float extra_dist = 0.0; + if (any(greaterThan(apos, FRAME.colliders[i].extents))) { //outside + vec3 mpos = min(apos, FRAME.colliders[i].extents); + extra_dist = distance(mpos, apos); + } + + if (extra_dist > particle_size) { + continue; + } + + vec3 uvw_pos = (local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5; + float s = texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).r; + s *= FRAME.colliders[i].scale; + s += extra_dist; + if (s < particle_size) { + col = true; + depth = particle_size - s; + const float EPSILON = 0.001; + normal = mat3(FRAME.colliders[i].transform) * + normalize( + vec3( + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, 0.0, EPSILON)).r)); + } + + } break; + case COLLIDER_TYPE_HEIGHT_FIELD: { + vec3 local_pos_bottom = local_pos; + local_pos_bottom.y -= particle_size; + + if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { + continue; + } + const float DELTA = 1.0 / 8192.0; + + vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; + + float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; + + if (y > uvw_pos.y) { + //inside heightfield + + vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * FRAME.colliders[i].extents; + + normal = normalize(cross(pos1 - pos2, pos1 - pos3)); + float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; + + col = true; + depth = dot(normal, pos1) - dot(normal, local_pos_bottom); + } + + } break; + } + + if (col) { + if (!collided) { + collided = true; + collision_normal = normal; + collision_depth = depth; + } else { + vec3 c = collision_normal * collision_depth; + c += normal * max(0.0, depth - dot(normal, c)); + collision_normal = normalize(c); + collision_depth = length(c); + } + } + } + } + } + + if (particle_active) { +#CODE : PROCESS + } - /* clang-format on */ + PARTICLE.flags &= ~PARTICLE_FLAG_ACTIVE; + if (particle_active) { + PARTICLE.flags |= PARTICLE_FLAG_ACTIVE; } } diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index 6c782b6045..e88e68b511 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -2,14 +2,18 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) + struct ParticleData { mat4 xform; vec3 velocity; - bool is_active; + uint flags; vec4 color; vec4 custom; }; @@ -33,12 +37,35 @@ sort_buffer; #endif // USE_SORT_BUFFER +layout(set = 2, binding = 0, std430) restrict readonly buffer TrailBindPoses { + mat4 data[]; +} +trail_bind_poses; + layout(push_constant, binding = 0, std430) uniform Params { vec3 sort_direction; uint total_particles; + + uint trail_size; + uint trail_total; + float frame_delta; + float frame_remainder; + + vec3 align_up; + uint align_mode; + + bool order_by_lifetime; + uint lifetime_split; + bool lifetime_reverse; + uint pad; } params; +#define TRANSFORM_ALIGN_DISABLED 0 +#define TRANSFORM_ALIGN_Z_BILLBOARD 1 +#define TRANSFORM_ALIGN_Y_TO_VELOCITY 2 +#define TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY 3 + void main() { #ifdef MODE_FILL_SORT_BUFFER @@ -47,36 +74,152 @@ void main() { return; //discard } - sort_buffer.data[particle].x = dot(params.sort_direction, particles.data[particle].xform[3].xyz); + uint src_particle = particle; + if (params.trail_size > 1) { + src_particle = src_particle * params.trail_size + params.trail_size / 2; //use trail center for sorting + } + sort_buffer.data[particle].x = dot(params.sort_direction, particles.data[src_particle].xform[3].xyz); sort_buffer.data[particle].y = float(particle); #endif #ifdef MODE_FILL_INSTANCES uint particle = gl_GlobalInvocationID.x; - uint write_offset = gl_GlobalInvocationID.x * (3 + 1 + 1); //xform + color + custom if (particle >= params.total_particles) { return; //discard } #ifdef USE_SORT_BUFFER - particle = uint(sort_buffer.data[particle].y); //use index from sort buffer -#endif + + if (params.trail_size > 1) { + particle = uint(sort_buffer.data[particle / params.trail_size].y) + (particle % params.trail_size); + } else { + particle = uint(sort_buffer.data[particle].y); //use index from sort buffer + } +#else + if (params.order_by_lifetime) { + if (params.trail_size > 1) { + uint limit = (params.total_particles / params.trail_size) - params.lifetime_split; + + uint base_index = particle / params.trail_size; + uint base_offset = particle % params.trail_size; + + if (params.lifetime_reverse) { + base_index = (params.total_particles / params.trail_size) - base_index - 1; + } + + if (base_index < limit) { + base_index = params.lifetime_split + base_index; + } else { + base_index -= limit; + } + + particle = base_index * params.trail_size + base_offset; + + } else { + uint limit = params.total_particles - params.lifetime_split; + + if (params.lifetime_reverse) { + particle = params.total_particles - particle - 1; + } + + if (particle < limit) { + particle = params.lifetime_split + particle; + } else { + particle -= limit; + } + } + } +#endif // USE_SORT_BUFFER mat4 txform; - if (particles.data[particle].is_active) { - txform = transpose(particles.data[particle].xform); + if (bool(particles.data[particle].flags & PARTICLE_FLAG_ACTIVE) || bool(particles.data[particle].flags & PARTICLE_FLAG_TRAILED)) { + txform = particles.data[particle].xform; + if (params.trail_size > 1) { + // Since the steps don't fit precisely in the history frames, must do a tiny bit of + // interpolation to get them close to their intended location. + uint part_ofs = particle % params.trail_size; + float natural_ofs = fract((float(part_ofs) / float(params.trail_size)) * float(params.trail_total)) * params.frame_delta; + + txform[3].xyz -= particles.data[particle].velocity * natural_ofs; + } + + switch (params.align_mode) { + case TRANSFORM_ALIGN_DISABLED: { + } break; //nothing + case TRANSFORM_ALIGN_Z_BILLBOARD: { + mat3 local = mat3(normalize(cross(params.align_up, params.sort_direction)), params.align_up, params.sort_direction); + local = local * mat3(txform); + txform[0].xyz = local[0]; + txform[1].xyz = local[1]; + txform[2].xyz = local[2]; + + } break; + case TRANSFORM_ALIGN_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + if (length(v) > 0.0) { + txform[1].xyz = normalize(v); + } else { + txform[1].xyz = normalize(txform[1].xyz); + } + + txform[0].xyz = normalize(cross(txform[1].xyz, txform[2].xyz)); + txform[2].xyz = vec3(0.0, 0.0, 1.0) * s; + txform[0].xyz *= s; + txform[1].xyz *= s; + } break; + case TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + vec3 sv = v - params.sort_direction * dot(params.sort_direction, v); //screen velocity + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + + if (length(sv) == 0) { + sv = params.align_up; + } + + sv = normalize(sv); + + txform[0].xyz = normalize(cross(sv, params.sort_direction)) * s; + txform[1].xyz = sv * s; + txform[2].xyz = params.sort_direction * s; + + } break; + } + + txform[3].xyz += particles.data[particle].velocity * params.frame_remainder; + + if (params.trail_size > 1) { + uint part_ofs = particle % params.trail_size; + txform = txform * trail_bind_poses.data[part_ofs]; + } + + txform = transpose(txform); } else { txform = mat4(vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); //zero scale, becomes invisible } +#ifdef MODE_2D + + uint write_offset = gl_GlobalInvocationID.x * (2 + 1 + 1); //xform + color + custom + + instances.data[write_offset + 0] = txform[0]; + instances.data[write_offset + 1] = txform[1]; + instances.data[write_offset + 2] = particles.data[particle].color; + instances.data[write_offset + 3] = particles.data[particle].custom; + +#else + + uint write_offset = gl_GlobalInvocationID.x * (3 + 1 + 1); //xform + color + custom + instances.data[write_offset + 0] = txform[0]; instances.data[write_offset + 1] = txform[1]; instances.data[write_offset + 2] = txform[2]; instances.data[write_offset + 3] = particles.data[particle].color; instances.data[write_offset + 4] = particles.data[particle].custom; +#endif //MODE_2D #endif } diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index e83c4ca93b..fecf812a8c 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -2,10 +2,15 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#ifdef MODE_RESOLVE_DEPTH +layout(set = 0, binding = 0) uniform sampler2DMS source_depth; +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; +#endif + #ifdef MODE_RESOLVE_GI layout(set = 0, binding = 0) uniform sampler2DMS source_depth; layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; @@ -13,9 +18,9 @@ layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; layout(rgba8, set = 1, binding = 1) uniform restrict writeonly image2D dest_normal_roughness; -#ifdef GIPROBE_RESOLVE -layout(set = 2, binding = 0) uniform usampler2DMS source_giprobe; -layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_giprobe; +#ifdef VOXEL_GI_RESOLVE +layout(set = 2, binding = 0) uniform usampler2DMS source_voxel_gi; +layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_voxel_gi; #endif #endif @@ -34,12 +39,23 @@ void main() { return; } +#ifdef MODE_RESOLVE_DEPTH + + float depth_avg = 0.0; + for (int i = 0; i < params.sample_count; i++) { + depth_avg += texelFetch(source_depth, pos, i).r; + } + depth_avg /= float(params.sample_count); + imageStore(dest_depth, pos, vec4(depth_avg)); + +#endif + #ifdef MODE_RESOLVE_GI float best_depth = 1e20; vec4 best_normal_roughness = vec4(0.0); -#ifdef GIPROBE_RESOLVE - uvec2 best_giprobe; +#ifdef VOXEL_GI_RESOLVE + uvec2 best_voxel_gi; #endif #if 0 @@ -50,8 +66,8 @@ void main() { best_depth = depth; best_normal_roughness = texelFetch(source_normal_roughness,pos,i); -#ifdef GIPROBE_RESOLVE - best_giprobe = texelFetch(source_giprobe,pos,i).rg; +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi,pos,i).rg; #endif } } @@ -204,16 +220,16 @@ void main() { #endif best_depth = texelFetch(source_depth, pos, best_index).r; best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); -#ifdef GIPROBE_RESOLVE - best_giprobe = texelFetch(source_giprobe, pos, best_index).rg; +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi, pos, best_index).rg; #endif #endif imageStore(dest_depth, pos, vec4(best_depth)); imageStore(dest_normal_roughness, pos, vec4(best_normal_roughness)); -#ifdef GIPROBE_RESOLVE - imageStore(dest_giprobe, pos, uvec4(best_giprobe, 0, 0)); +#ifdef VOXEL_GI_RESOLVE + imageStore(dest_voxel_gi, pos, uvec4(best_voxel_gi, 0, 0)); #endif #endif diff --git a/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl index 464895928a..7b964675ca 100644 --- a/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl +++ b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl new file mode 100644 index 0000000000..99714b4504 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -0,0 +1,58 @@ +#ifdef ALPHA_HASH_USED + +float hash_2d(vec2 p) { + return fract(1.0e4 * sin(17.0 * p.x + 0.1 * p.y) * + (0.1 + abs(sin(13.0 * p.y + p.x)))); +} + +float hash_3d(vec3 p) { + return hash_2d(vec2(hash_2d(p.xy), p.z)); +} + +float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { + vec3 dx = dFdx(pos); + vec3 dy = dFdx(pos); + float delta_max_sqr = max(length(dx), length(dy)); + float pix_scale = 1.0 / (hash_scale * delta_max_sqr); + + vec2 pix_scales = + vec2(exp2(floor(log2(pix_scale))), exp2(ceil(log2(pix_scale)))); + + vec2 a_thresh = vec2(hash_3d(floor(pix_scales.x * pos.xyz)), + hash_3d(floor(pix_scales.y * pos.xyz))); + + float lerp_factor = fract(log2(pix_scale)); + + float a_interp = (1.0 - lerp_factor) * a_thresh.x + lerp_factor * a_thresh.y; + + float min_lerp = min(lerp_factor, 1.0 - lerp_factor); + + vec3 cases = vec3(a_interp * a_interp / (2.0 * min_lerp * (1.0 - min_lerp)), + (a_interp - 0.5 * min_lerp) / (1.0 - min_lerp), + 1.0 - ((1.0 - a_interp) * (1.0 - a_interp) / + (2.0 * min_lerp * (1.0 - min_lerp)))); + + float alpha_hash_threshold = + (lerp_factor < (1.0 - min_lerp)) ? ((lerp_factor < min_lerp) ? cases.x : cases.y) : cases.z; + + return clamp(alpha_hash_threshold, 0.0, 1.0); +} + +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + +float calc_mip_level(vec2 texture_coord) { + vec2 dx = dFdx(texture_coord); + vec2 dy = dFdy(texture_coord); + float delta_max_sqr = max(dot(dx, dx), dot(dy, dy)); + return max(0.0, 0.5 * log2(delta_max_sqr)); +} + +float compute_alpha_antialiasing_edge(float input_alpha, vec2 texture_coord, float alpha_edge) { + input_alpha *= 1.0 + max(0, calc_mip_level(texture_coord)) * 0.25; // 0.25 mip scale, magic number + input_alpha = (input_alpha - alpha_edge) / max(fwidth(input_alpha), 0.0001) + 0.5; + return clamp(input_alpha, 0.0, 1.0); +} + +#endif // ALPHA_ANTIALIASING_USED diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 7b86dac143..adf9f20618 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #include "scene_forward_clustered_inc.glsl" @@ -48,11 +48,11 @@ layout(location = 8) in vec4 custom2_attrib; layout(location = 9) in vec4 custom3_attrib; #endif -#if defined(BONES_USED) +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) layout(location = 10) in uvec4 bone_attrib; #endif -#if defined(WEIGHTS_USED) +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) layout(location = 11) in vec4 weight_attrib; #endif @@ -81,16 +81,14 @@ layout(location = 5) out vec3 tangent_interp; layout(location = 6) out vec3 binormal_interp; #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif -invariant gl_Position; - #ifdef MODE_DUAL_PARABOLOID layout(location = 8) out float dp_clip; @@ -99,11 +97,9 @@ layout(location = 8) out float dp_clip; layout(location = 9) out flat uint instance_index; -/* clang-format off */ - -VERTEX_SHADER_GLOBALS +invariant gl_Position; -/* clang-format on */ +#GLOBALS void main() { vec4 instance_custom = vec4(0.0); @@ -129,10 +125,72 @@ void main() { if (is_multimesh) { //multimesh, instances are for it - uint offset = (instances.data[instance_index].flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; - offset *= gl_InstanceIndex; mat4 matrix; + +#ifdef USE_PARTICLE_TRAILS + uint trail_size = (instances.data[instance_index].flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; + uint stride = 3 + 1 + 1; //particles always uses this format + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); offset += 2; @@ -152,6 +210,7 @@ void main() { instance_custom = transforms.data[offset]; } +#endif //transpose matrix = transpose(matrix); world_matrix = world_matrix * matrix; @@ -169,32 +228,6 @@ void main() { vec3 binormal = normalize(cross(normal, tangent) * binormalf); #endif -#if 0 - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_SKELETON)) { - //multimesh, instances are for it - - uvec2 bones_01 = uvec2(bone_attrib.x & 0xFFFF, bone_attrib.x >> 16) * 3; - uvec2 bones_23 = uvec2(bone_attrib.y & 0xFFFF, bone_attrib.y >> 16) * 3; - vec2 weights_01 = unpackUnorm2x16(bone_attrib.z); - vec2 weights_23 = unpackUnorm2x16(bone_attrib.w); - - mat4 m = mat4(transforms.data[bones_01.x], transforms.data[bones_01.x + 1], transforms.data[bones_01.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; - m += mat4(transforms.data[bones_01.y], transforms.data[bones_01.y + 1], transforms.data[bones_01.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; - m += mat4(transforms.data[bones_23.x], transforms.data[bones_23.x + 1], transforms.data[bones_23.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; - m += mat4(transforms.data[bones_23.y], transforms.data[bones_23.y + 1], transforms.data[bones_23.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; - - //reverse order because its transposed - vertex = (vec4(vertex, 1.0) * m).xyz; - normal = (vec4(normal, 0.0) * m).xyz; - -#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - - tangent = (vec4(tangent, 0.0) * m).xyz; - binormal = (vec4(binormal, 0.0) * m).xyz; -#endif - } -#endif - #ifdef UV_USED uv_interp = uv_attrib; #endif @@ -230,11 +263,7 @@ void main() { mat3 modelview_normal = mat3(scene_data.inv_camera_matrix) * world_normal_matrix; { - /* clang-format off */ - -VERTEX_SHADER_CODE - - /* clang-format on */ +#CODE : VERTEX } // using local coordinates (default) @@ -325,7 +354,25 @@ VERTEX_SHADER_CODE #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_use_forward_gi = false; +layout(constant_id = 1) const bool sc_use_light_projector = false; +layout(constant_id = 2) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 3) const bool sc_use_directional_soft_shadows = false; + +/* Specialization Constants (Values) */ + +layout(constant_id = 6) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 7) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 9) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 10) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 11) const bool sc_projector_use_mipmaps = true; #include "scene_forward_clustered_inc.glsl" @@ -372,19 +419,15 @@ layout(location = 9) in flat uint instance_index; #define LIGHT_TRANSMITTANCE_USED #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ -} material; -#endif -/* clang-format off */ +#MATERIAL_UNIFORMS -FRAGMENT_SHADER_GLOBALS +} material; +#endif -/* clang-format on */ +#GLOBALS #ifdef MODE_RENDER_DEPTH @@ -396,13 +439,13 @@ layout(location = 2) out vec4 orm_output_buffer; layout(location = 3) out vec4 emission_output_buffer; layout(location = 4) out float depth_output_buffer; -#endif +#endif // MODE_RENDER_MATERIAL #ifdef MODE_RENDER_NORMAL_ROUGHNESS layout(location = 0) out vec4 normal_roughness_output_buffer; -#ifdef MODE_RENDER_GIPROBE -layout(location = 1) out uvec2 giprobe_buffer; +#ifdef MODE_RENDER_VOXEL_GI +layout(location = 1) out uvec2 voxel_gi_buffer; #endif #endif //MODE_RENDER_NORMAL @@ -415,1328 +458,27 @@ layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface sc #else layout(location = 0) out vec4 frag_color; -#endif +#endif // MODE_MULTIPLE_RENDER_TARGETS #endif // RENDER DEPTH -#ifdef ALPHA_HASH_USED - -float hash_2d(vec2 p) { - return fract(1.0e4 * sin(17.0 * p.x + 0.1 * p.y) * - (0.1 + abs(sin(13.0 * p.y + p.x)))); -} - -float hash_3d(vec3 p) { - return hash_2d(vec2(hash_2d(p.xy), p.z)); -} - -float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { - vec3 dx = dFdx(pos); - vec3 dy = dFdx(pos); - float delta_max_sqr = max(length(dx), length(dy)); - float pix_scale = 1.0 / (hash_scale * delta_max_sqr); - - vec2 pix_scales = - vec2(exp2(floor(log2(pix_scale))), exp2(ceil(log2(pix_scale)))); - - vec2 a_thresh = vec2(hash_3d(floor(pix_scales.x * pos.xyz)), - hash_3d(floor(pix_scales.y * pos.xyz))); - - float lerp_factor = fract(log2(pix_scale)); - - float a_interp = (1.0 - lerp_factor) * a_thresh.x + lerp_factor * a_thresh.y; - - float min_lerp = min(lerp_factor, 1.0 - lerp_factor); - - vec3 cases = vec3(a_interp * a_interp / (2.0 * min_lerp * (1.0 - min_lerp)), - (a_interp - 0.5 * min_lerp) / (1.0 - min_lerp), - 1.0 - ((1.0 - a_interp) * (1.0 - a_interp) / - (2.0 * min_lerp * (1.0 - min_lerp)))); - - float alpha_hash_threshold = - (lerp_factor < (1.0 - min_lerp)) ? ((lerp_factor < min_lerp) ? cases.x : cases.y) : cases.z; - - return clamp(alpha_hash_threshold, 0.0, 1.0); -} - -#endif // ALPHA_HASH_USED - -#ifdef ALPHA_ANTIALIASING_EDGE_USED - -float calc_mip_level(vec2 texture_coord) { - vec2 dx = dFdx(texture_coord); - vec2 dy = dFdy(texture_coord); - float delta_max_sqr = max(dot(dx, dx), dot(dy, dy)); - return max(0.0, 0.5 * log2(delta_max_sqr)); -} - -float compute_alpha_antialiasing_edge(float input_alpha, vec2 texture_coord, float alpha_edge) { - input_alpha *= 1.0 + max(0, calc_mip_level(texture_coord)) * 0.25; // 0.25 mip scale, magic number - input_alpha = (input_alpha - alpha_edge) / max(fwidth(input_alpha), 0.0001) + 0.5; - return clamp(input_alpha, 0.0, 1.0); -} - -#endif // ALPHA_ANTIALIASING_USED - -// This returns the G_GGX function divided by 2 cos_theta_m, where in practice cos_theta_m is either N.L or N.V. -// We're dividing this factor off because the overall term we'll end up looks like -// (see, for example, the first unnumbered equation in B. Burley, "Physically Based Shading at Disney", SIGGRAPH 2012): -// -// F(L.V) D(N.H) G(N.L) G(N.V) / (4 N.L N.V) -// -// We're basically regouping this as -// -// F(L.V) D(N.H) [G(N.L)/(2 N.L)] [G(N.V) / (2 N.V)] -// -// and thus, this function implements the [G(N.m)/(2 N.m)] part with m = L or V. -// -// The contents of the D and G (G1) functions (GGX) are taken from -// E. Heitz, "Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs", J. Comp. Graph. Tech. 3 (2) (2014). -// Eqns 71-72 and 85-86 (see also Eqns 43 and 80). +#include "scene_forward_aa_inc.glsl" #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -float G_GGX_2cos(float cos_theta_m, float alpha) { - // Schlick's approximation - // C. Schlick, "An Inexpensive BRDF Model for Physically-based Rendering", Computer Graphics Forum. 13 (3): 233 (1994) - // Eq. (19), although see Heitz (2014) the about the problems with his derivation. - // It nevertheless approximates GGX well with k = alpha/2. - float k = 0.5 * alpha; - return 0.5 / (cos_theta_m * (1.0 - k) + k); - - // float cos2 = cos_theta_m * cos_theta_m; - // float sin2 = (1.0 - cos2); - // return 1.0 / (cos_theta_m + sqrt(cos2 + alpha * alpha * sin2)); -} - -float D_GGX(float cos_theta_m, float alpha) { - float alpha2 = alpha * alpha; - float d = 1.0 + (alpha2 - 1.0) * cos_theta_m * cos_theta_m; - return alpha2 / (M_PI * d * d); -} - -float G_GGX_anisotropic_2cos(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float s_x = alpha_x * cos_phi; - float s_y = alpha_y * sin_phi; - return 1.0 / max(cos_theta_m + sqrt(cos2 + (s_x * s_x + s_y * s_y) * sin2), 0.001); -} - -float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float r_x = cos_phi / alpha_x; - float r_y = sin_phi / alpha_y; - float d = cos2 + sin2 * (r_x * r_x + r_y * r_y); - return 1.0 / max(M_PI * alpha_x * alpha_y * d * d, 0.001); -} - -float SchlickFresnel(float u) { - float m = 1.0 - u; - float m2 = m * m; - return m2 * m2 * m; // pow(m,5) -} - -float GTR1(float NdotH, float a) { - if (a >= 1.0) - return 1.0 / M_PI; - float a2 = a * a; - float t = 1.0 + (a2 - 1.0) * NdotH * NdotH; - return (a2 - 1.0) / (M_PI * log(a2) * t); -} - -vec3 F0(float metallic, float specular, vec3 albedo) { - float dielectric = 0.16 * specular * specular; - // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; - // see https://google.github.io/filament/Filament.md.html - return mix(vec3(dielectric), albedo, vec3(metallic)); -} - -void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, - float transmittance_z, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 B, vec3 T, float anisotropy, -#endif -#ifdef USE_SOFT_SHADOWS - float A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - -#if defined(USE_LIGHT_SHADER_CODE) - // light is written by the light shader - - vec3 normal = N; - vec3 light = L; - vec3 view = V; - - /* clang-format off */ - -LIGHT_SHADER_CODE - - /* clang-format on */ - -#else - -#ifdef USE_SOFT_SHADOWS - float NdotL = min(A + dot(N, L), 1.0); -#else - float NdotL = dot(N, L); -#endif - float cNdotL = max(NdotL, 0.0); // clamped NdotL - float NdotV = dot(N, V); - float cNdotV = max(NdotV, 0.0); - -#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) - vec3 H = normalize(V + L); -#endif - -#if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS - float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); -#else - float cNdotH = clamp(dot(N, H), 0.0, 1.0); -#endif -#endif - -#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS - float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); -#else - float cLdotH = clamp(dot(L, H), 0.0, 1.0); -#endif -#endif - - float metallic = unpackUnorm4x8(orms).z; - if (metallic < 1.0) { - float roughness = unpackUnorm4x8(orms).y; - -#if defined(DIFFUSE_OREN_NAYAR) - vec3 diffuse_brdf_NL; -#else - float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance -#endif - -#if defined(DIFFUSE_LAMBERT_WRAP) - // energy conserving lambert wrap shader - diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); -#elif defined(DIFFUSE_TOON) - - diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); - -#elif defined(DIFFUSE_BURLEY) - - { - float FD90_minus_1 = 2.0 * cLdotH * cLdotH * roughness - 0.5; - float FdV = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotV); - float FdL = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotL); - diffuse_brdf_NL = (1.0 / M_PI) * FdV * FdL * cNdotL; - /* - float energyBias = mix(roughness, 0.0, 0.5); - float energyFactor = mix(roughness, 1.0, 1.0 / 1.51); - float fd90 = energyBias + 2.0 * VoH * VoH * roughness; - float f0 = 1.0; - float lightScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotL, 5.0); - float viewScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotV, 5.0); - - diffuse_brdf_NL = lightScatter * viewScatter * energyFactor; - */ - } -#else - // lambert - diffuse_brdf_NL = cNdotL * (1.0 / M_PI); -#endif - - diffuse_light += light_color * diffuse_brdf_NL * attenuation; - -#if defined(LIGHT_BACKLIGHT_USED) - diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; -#endif - -#if defined(LIGHT_RIM_USED) - float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); - diffuse_light += rim_light * rim * mix(vec3(1.0), rim_color, rim_tint) * light_color; -#endif - -#ifdef LIGHT_TRANSMITTANCE_USED - -#ifdef SSS_MODE_SKIN - - { - float scale = 8.25 / transmittance_depth; - float d = scale * abs(transmittance_z); - float dd = -d * d; - vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + - vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + - vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + - vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + - vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + - vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); - - diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); - } -#else - - if (transmittance_depth > 0.0) { - float fade = clamp(abs(transmittance_z / transmittance_depth), 0.0, 1.0); - - fade = pow(max(0.0, 1.0 - fade), transmittance_curve); - fade *= clamp(transmittance_boost - NdotL, 0.0, 1.0); - - diffuse_light += transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade; - } - -#endif //SSS_MODE_SKIN - -#endif //LIGHT_TRANSMITTANCE_USED - } - - float roughness = unpackUnorm4x8(orms).y; - if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely - - // D - -#if defined(SPECULAR_BLINN) - - //normalized blinn - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float blinn = pow(cNdotH, shininess) * cNdotL; - blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = blinn; - - specular_light += light_color * intensity * attenuation * specular_amount; - -#elif defined(SPECULAR_PHONG) - - vec3 R = normalize(-reflect(L, N)); - float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float phong = pow(cRdotV, shininess); - phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); - - specular_light += light_color * intensity * attenuation * specular_amount; - -#elif defined(SPECULAR_TOON) - - vec3 R = normalize(-reflect(L, N)); - float RdotV = dot(R, V); - float mid = 1.0 - roughness; - mid *= mid; - float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; - diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection - -#elif defined(SPECULAR_DISABLED) - // none.. - -#elif defined(SPECULAR_SCHLICK_GGX) - // shlick+ggx as default - -#if defined(LIGHT_ANISOTROPY_USED) - - float alpha_ggx = roughness * roughness; - float aspect = sqrt(1.0 - anisotropy * 0.9); - float ax = alpha_ggx / aspect; - float ay = alpha_ggx * aspect; - float XdotH = dot(T, H); - float YdotH = dot(B, H); - float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); - float G = G_GGX_anisotropic_2cos(cNdotL, ax, ay, XdotH, YdotH) * G_GGX_anisotropic_2cos(cNdotV, ax, ay, XdotH, YdotH); - -#else - float alpha_ggx = roughness * roughness; - float D = D_GGX(cNdotH, alpha_ggx); - float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); -#endif - // F - float cLdotH5 = SchlickFresnel(cLdotH); - vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); - - vec3 specular_brdf_NL = cNdotL * D * F * G; - - specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; -#endif - -#if defined(LIGHT_CLEARCOAT_USED) - -#if !defined(SPECULAR_SCHLICK_GGX) - float cLdotH5 = SchlickFresnel(cLdotH); -#endif - float Dr = GTR1(cNdotH, mix(.1, .001, clearcoat_gloss)); - float Fr = mix(.04, 1.0, cLdotH5); - float Gr = G_GGX_2cos(cNdotL, .25) * G_GGX_2cos(cNdotV, .25); - - float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; - - specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; -#endif - } - -#ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); -#endif - -#endif //defined(USE_LIGHT_SHADER_CODE) -} - -#ifndef USE_NO_SHADOWS - -// Interleaved Gradient Noise -// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare -float quick_hash(vec2 pos) { - const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); - return fract(magic.z * fract(dot(pos, magic.xy))); -} - -float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - vec2 pos = coord.xy; - float depth = coord.z; - - //if only one sample is taken, take it from the center - if (scene_data.directional_soft_shadow_samples == 1) { - return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); - } - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float avg = 0.0; - - for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); - } - - return avg * (1.0 / float(scene_data.directional_soft_shadow_samples)); -} - -float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - vec2 pos = coord.xy; - float depth = coord.z; - - //if only one sample is taken, take it from the center - if (scene_data.soft_shadow_samples == 1) { - return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); - } - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float avg = 0.0; - - for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); - } - - return avg * (1.0 / float(scene_data.soft_shadow_samples)); -} - -float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { - //find blocker - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; - float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; - if (d < pssm_coord.z) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (pssm_coord.z - blocker_average) / blocker_average; - tex_scale *= penumbra; - - float s = 0.0; - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; - s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); - } - - return s / float(scene_data.directional_penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - return 1.0; - } -} - -#endif //USE_NO_SHADOWS - -float get_omni_attenuation(float distance, float inv_range, float decay) { - float nd = distance * inv_range; - nd *= nd; - nd *= nd; // nd^4 - nd = max(1.0 - nd, 0.0); - nd *= nd; // nd^2 - return nd * pow(max(distance, 0.0001), -decay); -} - -float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { -#ifndef USE_NO_SHADOWS - if (omni_lights.data[idx].shadow_enabled) { - // there is a shadowmap - - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - - vec4 v = vec4(vertex, 1.0); - - vec4 splane = (omni_lights.data[idx].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here - - { - vec3 nofs = normal_interp * omni_lights.data[idx].shadow_normal_bias / omni_lights.data[idx].inv_radius; - nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); - v.xyz += nofs; - splane = (omni_lights.data[idx].shadow_matrix * v); - } - - float shadow; - -#ifdef USE_SOFT_SHADOWS - if (omni_lights.data[idx].soft_shadow_size > 0.0) { - //soft shadow - - //find blocker - - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - vec3 normal = normalize(splane.xyz); - vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); - vec3 tangent = normalize(cross(v0, normal)); - vec3 bitangent = normalize(cross(tangent, normal)); - float z_norm = shadow_len * omni_lights.data[idx].inv_radius; - - tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; - bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; - - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; - - pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; - - if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; - } - - pos.xy /= pos.z; - - pos.xy = pos.xy * 0.5 + 0.5; - pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; - - float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; - if (d < z_norm) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (z_norm - blocker_average) / blocker_average; - tangent *= penumbra; - bitangent *= penumbra; - - z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; - - shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; - - pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; - - if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; - } - - pos.xy /= pos.z; - - pos.xy = pos.xy * 0.5 + 0.5; - pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; - shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); - } - - shadow /= float(scene_data.penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - shadow = 1.0; - } - } else { -#endif - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - - if (splane.z >= 0.0) { - splane.z += 1.0; - - clamp_rect.y += clamp_rect.w; - - } else { - splane.z = 1.0 - splane.z; - } - - splane.xy /= splane.z; - - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = (shadow_len - omni_lights.data[idx].shadow_bias) * omni_lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - shadow = sample_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); -#ifdef USE_SOFT_SHADOWS - } -#endif - - return shadow; - } -#endif - - return 1.0; -} - -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); - float light_attenuation = omni_attenuation; - vec3 color = omni_lights.data[idx].color; - -#ifdef USE_SOFT_SHADOWS - float size_A = 0.0; - - if (omni_lights.data[idx].size > 0.0) { - float t = omni_lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -#endif - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; //no transmittance by default - transmittance_color.a *= light_attenuation; - { - vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - - //redo shadowmapping, but shrink the model a bit to avoid arctifacts - vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); - - shadow_len = length(splane.xyz); - splane = normalize(splane.xyz); - - if (splane.z >= 0.0) { - splane.z += 1.0; - - } else { - splane.z = 1.0 - splane.z; - } - - splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * omni_lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; - } -#endif - -#if 0 - - if (omni_lights.data[idx].projector_rect != vec4(0.0)) { - vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - local_v = normalize(local_v); - - vec4 atlas_rect = omni_lights.data[idx].projector_rect; - - if (local_v.z >= 0.0) { - local_v.z += 1.0; - atlas_rect.y += atlas_rect.w; - - } else { - local_v.z = 1.0 - local_v.z; - } - - local_v.xy /= local_v.z; - local_v.xy = local_v.xy * 0.5 + 0.5; - vec2 proj_uv = local_v.xy * atlas_rect.zw; - - vec2 proj_uv_ddx; - vec2 proj_uv_ddy; - { - vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; - local_v_ddx = normalize(local_v_ddx); - - if (local_v_ddx.z >= 0.0) { - local_v_ddx.z += 1.0; - } else { - local_v_ddx.z = 1.0 - local_v_ddx.z; - } - - local_v_ddx.xy /= local_v_ddx.z; - local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; - - proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; - - vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; - local_v_ddy = normalize(local_v_ddy); - - if (local_v_ddy.z >= 0.0) { - local_v_ddy.z += 1.0; - } else { - local_v_ddy.z = 1.0 - local_v_ddy.z; - } - - local_v_ddy.xy /= local_v_ddy.z; - local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; - - proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; - } - - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); - } -#endif - - light_attenuation *= shadow; - - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, -#endif -#ifdef LIGHT_RIM_USED - rim * omni_attenuation, rim_tint, rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOWS - size_A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, - specular_light); -} - -float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { -#ifndef USE_NO_SHADOWS - if (spot_lights.data[idx].shadow_enabled) { - vec3 light_rel_vec = spot_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - vec3 spot_dir = spot_lights.data[idx].direction; - //there is a shadowmap - vec4 v = vec4(vertex, 1.0); - - v.xyz -= spot_dir * spot_lights.data[idx].shadow_bias; - - float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; - - float depth_bias_scale = 1.0 / (max(0.0001, z_norm)); //the closer to the light origin, the more you have to offset to reach 1px in the map - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * spot_lights.data[idx].shadow_normal_bias * depth_bias_scale; - normal_bias -= spot_dir * dot(spot_dir, normal_bias); //only XY, no Z - v.xyz += normal_bias; - - //adjust with bias - z_norm = dot(spot_dir, v.xyz - spot_lights.data[idx].position) * spot_lights.data[idx].inv_radius; - - float shadow; - - vec4 splane = (spot_lights.data[idx].shadow_matrix * v); - splane /= splane.w; - -#ifdef USE_SOFT_SHADOWS - if (spot_lights.data[idx].soft_shadow_size > 0.0) { - //soft shadow - - //find blocker - - vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; - - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; - vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); - float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; - if (d < z_norm) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (z_norm - blocker_average) / blocker_average; - uv_size *= penumbra; - - shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); - shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); - } - - shadow /= float(scene_data.penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - shadow = 1.0; - } - - } else { -#endif - //hard shadow - vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); - - shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); -#ifdef USE_SOFT_SHADOWS - } -#endif - - return shadow; - } - -#endif //USE_NO_SHADOWS - - return 1.0; -} - -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, - inout vec3 specular_light) { - vec3 light_rel_vec = spot_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); - vec3 spot_dir = spot_lights.data[idx].direction; - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); - spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); - float light_attenuation = spot_attenuation; - vec3 color = spot_lights.data[idx].color; - float specular_amount = spot_lights.data[idx].specular_amount; - -#ifdef USE_SOFT_SHADOWS - float size_A = 0.0; - - if (spot_lights.data[idx].size > 0.0) { - float t = spot_lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -#endif - - /* - if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; - transmittance_color.a *= light_attenuation; - { - splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); - splane /= splane.w; - splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - //reconstruct depth - shadow_z /= spot_lights.data[idx].inv_radius; - //distance to light plane - float z = dot(spot_dir, -light_rel_vec); - transmittance_z = z - shadow_z; - } -#endif //LIGHT_TRANSMITTANCE_USED - - light_attenuation *= shadow; - - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX #endif -#ifdef LIGHT_RIM_USED - rim * spot_attenuation, rim_tint, rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOW - size_A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, specular_light); -} - -void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { - vec3 box_extents = reflections.data[ref_index].box_extents; - vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; - - if (any(greaterThan(abs(local_pos), box_extents))) { //out of the reflection box - return; - } - - vec3 ref_vec = normalize(reflect(vertex, normal)); - - vec3 inner_pos = abs(local_pos / box_extents); - float blend = max(inner_pos.x, max(inner_pos.y, inner_pos.z)); - //make blend more rounded - blend = mix(length(inner_pos), blend, blend); - blend *= blend; - blend = max(0.0, 1.0 - blend); - - if (reflections.data[ref_index].intensity > 0.0) { // compute reflection - - vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; - - if (reflections.data[ref_index].box_project) { //box project - - vec3 nrdir = normalize(local_ref_vec); - vec3 rbmax = (box_extents - local_pos) / nrdir; - vec3 rbmin = (-box_extents - local_pos) / nrdir; - - vec3 rbminmax = mix(rbmin, rbmax, greaterThan(nrdir, vec3(0.0, 0.0, 0.0))); - - float fa = min(min(rbminmax.x, rbminmax.y), rbminmax.z); - vec3 posonbox = local_pos + nrdir * fa; - local_ref_vec = posonbox - reflections.data[ref_index].box_offset; - } - - vec4 reflection; - - reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb; - - if (reflections.data[ref_index].exterior) { - reflection.rgb = mix(specular_light, reflection.rgb, blend); - } - - reflection.rgb *= reflections.data[ref_index].intensity; //intensity - reflection.a = blend; - reflection.rgb *= reflection.a; - - reflection_accum += reflection; - } - - switch (reflections.data[ref_index].ambient_mode) { - case REFLECTION_AMBIENT_DISABLED: { - //do nothing - } break; - case REFLECTION_AMBIENT_ENVIRONMENT: { - //do nothing - vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz; - - vec4 ambient_out; - - ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb; - ambient_out.a = blend; - if (reflections.data[ref_index].exterior) { - ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); - } - - ambient_out.rgb *= ambient_out.a; - ambient_accum += ambient_out; - } break; - case REFLECTION_AMBIENT_COLOR: { - vec4 ambient_out; - ambient_out.a = blend; - ambient_out.rgb = reflections.data[ref_index].ambient; - if (reflections.data[ref_index].exterior) { - ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); - } - ambient_out.rgb *= ambient_out.a; - ambient_accum += ambient_out; - } break; - } -} - -#ifdef USE_FORWARD_GI - -//standard voxel cone trace -vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { - float dist = p_bias; - vec4 color = vec4(0.0); - - while (dist < max_distance && color.a < 0.95) { - float diameter = max(1.0, 2.0 * tan_half_angle * dist); - vec3 uvw_pos = (pos + dist * direction) * cell_size; - float half_diameter = diameter * 0.5; - //check if outside, then break - if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { - break; - } - vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2(diameter)); - float a = (1.0 - color.a); - color += a * scolor; - dist += half_diameter; - } - - return color; -} - -vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { - float dist = p_bias; - vec4 color = vec4(0.0); - float radius = max(0.5, tan_half_angle * dist); - float lod_level = log2(radius * 2.0); - - while (dist < max_distance && color.a < 0.95) { - vec3 uvw_pos = (pos + dist * direction) * cell_size; - - //check if outside, then break - if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { - break; - } - vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level); - lod_level += 1.0; - - float a = (1.0 - color.a); - scolor *= a; - color += scolor; - dist += radius; - radius = max(0.5, tan_half_angle * dist); - } - - return color; -} - -void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) { - position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz; - ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz); - normal = normalize((gi_probes.data[index].xform * vec4(normal, 0.0)).xyz); - - position += normal * gi_probes.data[index].normal_bias; - - //this causes corrupted pixels, i have no idea why.. - if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, gi_probes.data[index].bounds))))) { - return; - } - - vec3 blendv = abs(position / gi_probes.data[index].bounds * 2.0 - 1.0); - float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); - //float blend=1.0; - - float max_distance = length(gi_probes.data[index].bounds); - vec3 cell_size = 1.0 / gi_probes.data[index].bounds; - - //radiance - -#define MAX_CONE_DIRS 4 - - vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( - vec3(0.707107, 0.0, 0.707107), - vec3(0.0, 0.707107, 0.707107), - vec3(-0.707107, 0.0, 0.707107), - vec3(0.0, -0.707107, 0.707107)); - - float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25); - float cone_angle_tan = 0.98269; - - vec3 light = vec3(0.0); - - for (int i = 0; i < MAX_CONE_DIRS; i++) { - vec3 dir = normalize((gi_probes.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz); - - vec4 cone_light = voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias); - - if (gi_probes.data[index].blend_ambient) { - cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95)); - } - - light += cone_weights[i] * cone_light.rgb; - } - - light *= gi_probes.data[index].dynamic_range; - out_diff += vec4(light * blend, blend); - - //irradiance - vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias); - if (gi_probes.data[index].blend_ambient) { - irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95)); - } - irr_light.rgb *= gi_probes.data[index].dynamic_range; - //irr_light=vec3(0.0); - - out_spec += vec4(irr_light.rgb * blend, blend); -} - -vec2 octahedron_wrap(vec2 v) { - vec2 signVal; - signVal.x = v.x >= 0.0 ? 1.0 : -1.0; - signVal.y = v.y >= 0.0 ? 1.0 : -1.0; - return (1.0 - abs(v.yx)) * signVal; -} - -vec2 octahedron_encode(vec3 n) { - // https://twitter.com/Stubbesaurus/status/937994790553227264 - n /= (abs(n.x) + abs(n.y) + abs(n.z)); - n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); - n.xy = n.xy * 0.5 + 0.5; - return n.xy; -} - -void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) { - cascade_pos += cam_normal * sdfgi.normal_bias; - - vec3 base_pos = floor(cascade_pos); - //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; - ivec3 probe_base_pos = ivec3(base_pos); - - vec4 diffuse_accum = vec4(0.0); - vec3 specular_accum; - - ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); - tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; - tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); - - vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; - - vec3 specular_posf; - - if (use_specular) { - specular_accum = vec3(0.0); - specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; - } - - vec4 light_accum = vec4(0.0); - float weight_accum = 0.0; - - for (uint j = 0; j < 8; j++) { - ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); - ivec3 probe_posi = probe_base_pos; - probe_posi += offset; - - // Compute weight - vec3 probe_pos = vec3(probe_posi); - vec3 probe_to_pos = cascade_pos - probe_pos; - vec3 probe_dir = normalize(-probe_to_pos); +#include "scene_forward_lights_inc.glsl" - vec3 trilinear = vec3(1.0) - abs(probe_to_pos); - float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); - - // Compute lightprobe occlusion - - if (sdfgi.use_occlusion) { - ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); - vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); - - vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; - occ_pos.z += float(cascade); - if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures - occ_pos.x += 1.0; - } - - occ_pos *= sdfgi.occlusion_renormalize; - float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask); - - weight *= max(occlusion, 0.01); - } - - // Compute lightprobe texture position - - vec3 diffuse; - vec3 pos_uvw = diffuse_posf; - pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; - pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; - diffuse = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb; - - diffuse_accum += vec4(diffuse * weight, weight); - - if (use_specular) { - vec3 specular = vec3(0.0); - vec3 pos_uvw = specular_posf; - pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; - pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; - if (roughness < 0.99) { - specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; - } - if (roughness > 0.5) { - specular = mix(specular, textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb, (roughness - 0.5) * 2.0); - } - - specular_accum += specular * weight; - } - } - - if (diffuse_accum.a > 0.0) { - diffuse_accum.rgb /= diffuse_accum.a; - } - - diffuse_light = diffuse_accum.rgb; - - if (use_specular) { - if (diffuse_accum.a > 0.0) { - specular_accum /= diffuse_accum.a; - } - - specular_light = specular_accum; - } - - { - //process blend - float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - - vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; - - float len = length(inner_pos); - - inner_pos = abs(normalize(inner_pos)); - len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); - - if (len >= blend_from) { - blend = smoothstep(blend_from, blend_to, len); - } else { - blend = 0.0; - } - } -} - -#endif //USE_FORWARD_GI +#include "scene_forward_gi_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH -#ifndef LOW_END_MODE - vec4 volumetric_fog_process(vec2 screen_uv, float z) { vec3 fog_pos = vec3(screen_uv, z * scene_data.volumetric_fog_inv_length); if (fog_pos.z < 0.0) { @@ -1747,7 +489,6 @@ vec4 volumetric_fog_process(vec2 screen_uv, float z) { return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); } -#endif vec4 fog_process(vec3 vertex) { vec3 fog_color = scene_data.fog_light_color; @@ -1811,26 +552,6 @@ uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); } -float blur_shadow(float shadow) { - return shadow; -#if 0 - //disabling for now, will investigate later - float interp_shadow = shadow; - if (gl_HelperInvocation) { - interp_shadow = -4.0; // technically anything below -4 will do but just to make sure - } - - uvec2 fc2 = uvec2(gl_FragCoord.xy); - interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); - interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); - - if (interp_shadow >= 0.0) { - shadow = interp_shadow; - } - return shadow; -#endif -} - #endif //!MODE_RENDER DEPTH void main() { @@ -1845,9 +566,8 @@ void main() { vec3 view = -normalize(vertex_interp); vec3 albedo = vec3(1.0); vec3 backlight = vec3(0.0); - vec4 transmittance_color = vec4(0.0); + vec4 transmittance_color = vec4(0.0, 0.0, 0.0, 1.0); float transmittance_depth = 0.0; - float transmittance_curve = 1.0; float transmittance_boost = 0.0; float metallic = 0.0; float specular = 0.5; @@ -1928,20 +648,12 @@ void main() { #endif // ALPHA_ANTIALIASING_EDGE_USED { - /* clang-format off */ - -FRAGMENT_SHADER_CODE - - /* clang-format on */ +#CODE : FRAGMENT } #ifdef LIGHT_TRANSMITTANCE_USED -#ifdef SSS_MODE_SKIN - transmittance_color.a = sss_strength; -#else transmittance_color.a *= sss_strength; #endif -#endif #ifndef USE_SHADOW_TO_OPACITY @@ -1964,7 +676,7 @@ FRAGMENT_SHADER_CODE #endif #ifdef ALPHA_ANTIALIASING_EDGE_USED -// If alpha scissor is used, we must further the edge threshold, otherwise we wont get any edge feather +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather #ifdef ALPHA_SCISSOR_USED alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); #endif @@ -2019,7 +731,6 @@ FRAGMENT_SHADER_CODE fog = fog_process(vertex); } -#ifndef LOW_END_MODE if (scene_data.volumetric_fog_enabled) { vec4 volumetric_fog = volumetric_fog_process(screen_uv, -vertex.z); if (scene_data.fog_enabled) { @@ -2037,7 +748,6 @@ FRAGMENT_SHADER_CODE fog = volumetric_fog; } } -#endif //!LOW_END_MODE #endif //!CUSTOM_FOG_USED uint fog_rg = packHalf2x16(fog.rg); @@ -2102,25 +812,35 @@ FRAGMENT_SHADER_CODE continue; //out of decal } - //we need ddx/ddy for mipmaps, so simulate them - vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; - vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; - float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); if (decals.data[decal_index].normal_fade > 0.0) { fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); } + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } decal_albedo *= decals.data[decal_index].modulate; decal_albedo.a *= fade; albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); if (decals.data[decal_index].normal_rect != vec4(0.0)) { - vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); //convert to view space, use xzy because y is up @@ -2130,7 +850,12 @@ FRAGMENT_SHADER_CODE } if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } ao = mix(ao, decal_orm.r, decal_albedo.a); roughness = mix(roughness, decal_orm.g, decal_albedo.a); metallic = mix(metallic, decal_orm.b, decal_albedo.a); @@ -2139,7 +864,11 @@ FRAGMENT_SHADER_CODE if (decals.data[decal_index].emission_rect != vec4(0.0)) { //emission is additive, so its independent from albedo - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + } } } } @@ -2154,7 +883,7 @@ FRAGMENT_SHADER_CODE if (scene_data.roughness_limiter_enabled) { //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf float roughness2 = roughness * roughness; - vec3 dndu = dFdx(normal), dndv = dFdx(normal); + vec3 dndu = dFdx(normal), dndv = dFdy(normal); float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); @@ -2183,6 +912,8 @@ FRAGMENT_SHADER_CODE specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -2208,7 +939,7 @@ FRAGMENT_SHADER_CODE } #endif // USE_LIGHTMAP #if defined(CUSTOM_IRRADIANCE_USED) - ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); #endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -2272,9 +1003,9 @@ FRAGMENT_SHADER_CODE ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; } } -#elif defined(USE_FORWARD_GI) +#else - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture //make vertex orientation the world one, but still align to camera vec3 cam_pos = mat3(scene_data.camera_matrix) * vertex; @@ -2346,7 +1077,7 @@ FRAGMENT_SHADER_CODE } } - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); @@ -2358,12 +1089,12 @@ FRAGMENT_SHADER_CODE vec4 amb_accum = vec4(0.0); vec4 spec_accum = vec4(0.0); - gi_probe_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); uint index2 = instances.data[instance_index].gi_offset >> 16; if (index2 != 0xFFFF) { - gi_probe_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); } if (amb_accum.a > 0.0) { @@ -2377,9 +1108,8 @@ FRAGMENT_SHADER_CODE specular_light = spec_accum.rgb; ambient_light = amb_accum.rgb; } -#elif !defined(LOW_END_MODE) - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + if (!sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers vec2 coord; @@ -2410,15 +1140,13 @@ FRAGMENT_SHADER_CODE ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); } -#endif +#endif // !USE_LIGHTMAP -#ifndef LOW_END_MODE if (scene_data.ssao_enabled) { float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; ao = min(ao, ssao); ao_light_affect = mix(ao_light_affect, max(ao_light_affect, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); } -#endif //LOW_END_MODE { // process reflections @@ -2493,7 +1221,7 @@ FRAGMENT_SHADER_CODE specular_light *= specular * metallic * albedo * 2.0; #else - // scales the specular reflections, needs to be be computed before lighting happens, + // scales the specular reflections, needs to be computed before lighting happens, // but after environment, GI, and reflection probes are added // Environment brdf approximation (Lazarov 2013) // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile @@ -2533,16 +1261,19 @@ FRAGMENT_SHADER_CODE continue; //not masked } + if (directional_lights.data[i].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + float shadow = 1.0; -#ifdef USE_SOFT_SHADOWS //version with soft shadows, more expensive if (directional_lights.data[i].shadow_enabled) { - float depth_z = -vertex.z; + if (sc_use_directional_soft_shadows && directional_lights.data[i].softshadow_angle > 0) { + float depth_z = -vertex.z; - vec4 pssm_coord; - vec3 shadow_color = vec3(0.0); - vec3 light_dir = directional_lights.data[i].direction; + vec3 shadow_color = vec3(0.0); + vec3 light_dir = directional_lights.data[i].direction; #define BIAS_FUNC(m_var, m_idx) \ m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ @@ -2550,168 +1281,105 @@ FRAGMENT_SHADER_CODE normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); + uint blend_index = 0; - BIAS_FUNC(v, 0) + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); - pssm_coord /= pssm_coord.w; + BIAS_FUNC(v, 0) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.x; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color1.rgb; - - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); + if (blend_index < 2 && depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) + BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.y; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_index == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow = mix(shadow, s, blend); + } + + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color2.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); + if (blend_index < 2 && depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) + BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.z; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - shadow_color = directional_lights.data[i].shadow_color3.rgb; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 3) - - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; + if (blend_index == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + shadow = mix(shadow, s, blend); + } - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.w; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color4.rgb; - } - - if (directional_lights.data[i].blend_splits) { - vec3 shadow_color_blend = vec3(0.0); - float pssm_blend; - float shadow2; - - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + if (blend_index < 2) { vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.y; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } + BIAS_FUNC(v, 3) - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.z; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 3) - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.w; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + if (blend_index == 0) { + shadow = s; } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + shadow = mix(shadow, s, blend); } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color4.rgb; - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) } - pssm_blend = sqrt(pssm_blend); - - shadow = mix(shadow, shadow2, pssm_blend); - shadow_color = mix(shadow_color, shadow_color_blend, pssm_blend); - } - - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance - #undef BIAS_FUNC - } -#else - // Soft shadow disabled version + } else { //no soft shadows - if (directional_lights.data[i].shadow_enabled) { - float depth_z = -vertex.z; + float depth_z = -vertex.z; - vec4 pssm_coord; - vec3 light_dir = directional_lights.data[i].direction; - vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); #define BIAS_FUNC(m_var, m_idx) \ m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ @@ -2719,122 +1387,70 @@ FRAGMENT_SHADER_CODE normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 0) - - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 1) - - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 2) - - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; - - transmittance_z = z - shadow_z; - } -#endif - - } else { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 3) - - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; - - transmittance_z = z - shadow_z; - } -#endif - } - - pssm_coord /= pssm_coord.w; - - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); - if (directional_lights.data[i].blend_splits) { - float pssm_blend; + BIAS_FUNC(v, 0) - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + + } else { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) } pssm_coord /= pssm_coord.w; - float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - shadow = mix(shadow, shadow2, pssm_blend); - } + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance #undef BIAS_FUNC - } -#endif + } + } // shadows if (i < 4) { shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); @@ -2864,8 +1480,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + shadow_z *= directional_lights.data[i].shadow_z_range.x; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.x; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { @@ -2874,8 +1490,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + shadow_z *= directional_lights.data[i].shadow_z_range.y; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.y; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { @@ -2884,8 +1500,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + shadow_z *= directional_lights.data[i].shadow_z_range.z; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.z; transmittance_z = z - shadow_z; @@ -2895,213 +1511,218 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + shadow_z *= directional_lights.data[i].shadow_z_range.w; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.w; transmittance_z = z - shadow_z; } + } #endif - float shadow = 1.0; + float shadow = 1.0; - if (i < 4) { - shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; - } else { - shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; - } + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } - blur_shadow(shadow); + blur_shadow(shadow); - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, + float size_A = sc_use_light_soft_shadows ? directional_lights.data[i].size : 0.0; + + light_compute(normal, directional_lights.data[i].direction, normalize(view), size_A, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim, rim_tint, albedo, + rim, rim_tint, albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOW - directional_lights.data[i].size, + binormal, tangent, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, - specular_light); - } + diffuse_light, + specular_light); } + } - { //omni lights + { //omni lights - uint cluster_omni_offset = cluster_offset; + uint cluster_omni_offset = cluster_offset; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS - item_from = subgroupBroadcastFirst(subgroupMin(item_from)); - item_to = subgroupBroadcastFirst(subgroupMax(item_to)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_omni_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - float shadow = light_process_omni_shadow(light_index, vertex, view); + if (omni_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + + float shadow = light_process_omni_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } - { //spot lights + { //spot lights - uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; + uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS - item_from = subgroupBroadcastFirst(subgroupMin(item_from)); - item_to = subgroupBroadcastFirst(subgroupMax(item_to)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_spot_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - float shadow = light_process_spot_shadow(light_index, vertex, view); + if (spot_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + + float shadow = light_process_spot_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); #if defined(ALPHA_SCISSOR_USED) - if (alpha < alpha_scissor) { - discard; - } + if (alpha < alpha_scissor) { + discard; + } #endif // ALPHA_SCISSOR_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { - discard; - } + if (alpha < opaque_prepass_threshold) { + discard; + } #endif // USE_OPAQUE_PREPASS @@ -3113,126 +1734,126 @@ FRAGMENT_SHADER_CODE #ifdef MODE_RENDER_SDF - { - vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; - ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); - - uint albedo16 = 0x1; //solid flag - albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; - albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; - albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; - - imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); - - uint facing_bits = 0; - const vec3 aniso_dir[6] = vec3[]( - vec3(1, 0, 0), - vec3(0, 1, 0), - vec3(0, 0, 1), - vec3(-1, 0, 0), - vec3(0, -1, 0), - vec3(0, 0, -1)); - - vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); - - float closest_dist = -1e20; - - for (uint i = 0; i < 6; i++) { - float d = dot(cam_normal, aniso_dir[i]); - if (d > closest_dist) { - closest_dist = d; - facing_bits = (1 << i); - } + { + vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); + + uint albedo16 = 0x1; //solid flag + albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; + albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; + albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; + + imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); + + uint facing_bits = 0; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); + + float closest_dist = -1e20; + + for (uint i = 0; i < 6; i++) { + float d = dot(cam_normal, aniso_dir[i]); + if (d > closest_dist) { + closest_dist = d; + facing_bits = (1 << i); } + } - imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits - if (length(emission) > 0.001) { - float lumas[6]; - vec3 light_total = vec3(0); + if (length(emission) > 0.001) { + float lumas[6]; + vec3 light_total = vec3(0); - for (int i = 0; i < 6; i++) { - float strength = max(0.0, dot(cam_normal, aniso_dir[i])); - vec3 light = emission * strength; - light_total += light; - lumas[i] = max(light.r, max(light.g, light.b)); - } + for (int i = 0; i < 6; i++) { + float strength = max(0.0, dot(cam_normal, aniso_dir[i])); + vec3 light = emission * strength; + light_total += light; + lumas[i] = max(light.r, max(light.g, light.b)); + } - float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); - uint light_aniso = 0; + uint light_aniso = 0; - for (int i = 0; i < 6; i++) { - light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); - } + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } - //compress to RGBE9995 to save space + //compress to RGBE9995 to save space - const float pow2to9 = 512.0f; - const float B = 15.0f; - const float N = 9.0f; - const float LN2 = 0.6931471805599453094172321215; + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; - float cRed = clamp(light_total.r, 0.0, 65408.0); - float cGreen = clamp(light_total.g, 0.0, 65408.0); - float cBlue = clamp(light_total.b, 0.0, 65408.0); + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); - float cMax = max(cRed, max(cGreen, cBlue)); + float cMax = max(cRed, max(cGreen, cBlue)); - float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; - float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); - float exps = expp + 1.0f; + float exps = expp + 1.0f; - if (0.0 <= sMax && sMax < pow2to9) { - exps = expp; - } + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } - float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); - float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); - float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); - //store as 8985 to have 2 extra neighbour bits - uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); + //store as 8985 to have 2 extra neighbour bits + uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); - imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); - imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); - } + imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); + imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); } + } #endif #ifdef MODE_RENDER_MATERIAL - albedo_output_buffer.rgb = albedo; - albedo_output_buffer.a = alpha; + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; - normal_output_buffer.rgb = normal * 0.5 + 0.5; - normal_output_buffer.a = 0.0; - depth_output_buffer.r = -vertex.z; + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; - orm_output_buffer.r = ao; - orm_output_buffer.g = roughness; - orm_output_buffer.b = metallic; - orm_output_buffer.a = sss_strength; + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; - emission_output_buffer.rgb = emission; - emission_output_buffer.a = 0.0; + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; #endif #ifdef MODE_RENDER_NORMAL_ROUGHNESS - normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); - -#ifdef MODE_RENDER_GIPROBE - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - uint index2 = instances.data[instance_index].gi_offset >> 16; - giprobe_buffer.x = index1 & 0xFF; - giprobe_buffer.y = index2 & 0xFF; - } else { - giprobe_buffer.x = 0xFF; - giprobe_buffer.y = 0xFF; - } + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); + +#ifdef MODE_RENDER_VOXEL_GI + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + uint index2 = instances.data[instance_index].gi_offset >> 16; + voxel_gi_buffer.x = index1 & 0xFF; + voxel_gi_buffer.y = index2 & 0xFF; + } else { + voxel_gi_buffer.x = 0xFF; + voxel_gi_buffer.y = 0xFF; + } #endif #endif //MODE_RENDER_NORMAL_ROUGHNESS @@ -3290,4 +1911,4 @@ FRAGMENT_SHADER_CODE #endif //MODE_MULTIPLE_RENDER_TARGETS #endif //MODE_RENDER_DEPTH - } +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index 4ea05c9ccc..b53bf6a6d4 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -1,7 +1,7 @@ #define M_PI 3.14159265359 #define ROUGHNESS_MAX_LOD 5 -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) @@ -13,8 +13,9 @@ #endif #include "cluster_data_inc.glsl" +#include "decal_data_inc.glsl" -#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_GIPROBE) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_VOXEL_GI) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) #ifndef NORMAL_USED #define NORMAL_USED #endif @@ -28,7 +29,11 @@ layout(push_constant, binding = 0, std430) uniform DrawCall { } draw_call; -/* Set 0 Scene data that never changes, ever */ +#define SDFGI_MAX_CASCADES 8 + +/* Set 0: Base Pass (never changes) */ + +#include "light_data_inc.glsl" #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 @@ -43,47 +48,45 @@ draw_call; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 -#define SDFGI_MAX_CASCADES 8 - -/* Set 1: Base Pass (never changes) */ - layout(set = 0, binding = 1) uniform sampler material_samplers[12]; layout(set = 0, binding = 2) uniform sampler shadow_sampler; +layout(set = 0, binding = 3) uniform sampler decal_sampler; + +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) #define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) #define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 9) #define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 10) -#define INSTANCE_FLAGS_USE_GIPROBE (1 << 11) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 11) #define INSTANCE_FLAGS_MULTIMESH (1 << 12) #define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) #define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) #define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) -#define INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT 16 +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 //3 bits of stride -#define INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK 0x7 +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_SKELETON (1 << 19) -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) - -layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 5) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 6, std140) uniform DirectionalLights { +layout(set = 0, binding = 8, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -95,7 +98,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -104,34 +107,32 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 9) uniform texture2D decal_atlas; -layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 11) uniform texture2D decal_atlas; +layout(set = 0, binding = 12) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; -#ifndef LOW_END_MODE - -struct SDFGIProbeCascadeData { +struct SDFVoxelGICascadeData { vec3 position; float to_probe; ivec3 probe_world_offset; float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 13, std140) uniform SDFGI { +layout(set = 0, binding = 15, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -155,13 +156,11 @@ layout(set = 0, binding = 13, std140) uniform SDFGI { vec3 cascade_probe_size; uint pad5; - SDFGIProbeCascadeData cascades[SDFGI_MAX_CASCADES]; + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; } sdfgi; -#endif //LOW_END_MODE - -/* Set 2: Render Pass (changes per render pass) */ +/* Set 1: Render Pass (changes per render pass) */ layout(set = 1, binding = 0, std140) uniform SceneData { mat4 projection_matrix; @@ -178,17 +177,12 @@ layout(set = 1, binding = 0, std140) uniform SceneData { uint cluster_type_size; uint max_cluster_element_count_div_32; - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. vec4 directional_penumbra_shadow_kernel[32]; vec4 directional_soft_shadow_kernel[32]; vec4 penumbra_shadow_kernel[32]; vec4 soft_shadow_kernel[32]; - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - vec4 ambient_light_color_energy; float ambient_color_sky_mix; @@ -245,7 +239,6 @@ layout(set = 1, binding = 0, std140) uniform SceneData { bool pancake_shadows; } - scene_data; struct InstanceData { @@ -280,9 +273,7 @@ layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -#ifndef LOW_END_MOD -layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; -#endif +layout(set = 1, binding = 7) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { uint data[]; @@ -306,8 +297,6 @@ layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; layout(set = 1, binding = 9) uniform texture2D depth_buffer; layout(set = 1, binding = 10) uniform texture2D color_buffer; -#ifndef LOW_END_MODE - layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; layout(set = 1, binding = 12) uniform texture2D ao_buffer; layout(set = 1, binding = 13) uniform texture2D ambient_buffer; @@ -315,7 +304,7 @@ layout(set = 1, binding = 14) uniform texture2D reflection_buffer; layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; -struct GIProbeData { +struct VoxelGIData { mat4 xform; vec3 bounds; float dynamic_range; @@ -331,15 +320,13 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 1, binding = 17, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; +layout(set = 1, binding = 17, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } -gi_probes; +voxel_gi_instances; layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; -#endif // LOW_END_MODE - #endif /* Set 2 Skeleton & Instancing (can change per item) */ diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl new file mode 100644 index 0000000000..c88bd0a14b --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl @@ -0,0 +1,242 @@ +// Functions related to gi/sdfgi for our forward renderer + +//standard voxel cone trace +vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + + while (dist < max_distance && color.a < 0.95) { + float diameter = max(1.0, 2.0 * tan_half_angle * dist); + vec3 uvw_pos = (pos + dist * direction) * cell_size; + float half_diameter = diameter * 0.5; + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2(diameter)); + float a = (1.0 - color.a); + color += a * scolor; + dist += half_diameter; + } + + return color; +} + +vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + float radius = max(0.5, tan_half_angle * dist); + float lod_level = log2(radius * 2.0); + + while (dist < max_distance && color.a < 0.95) { + vec3 uvw_pos = (pos + dist * direction) * cell_size; + + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level); + lod_level += 1.0; + + float a = (1.0 - color.a); + scolor *= a; + color += scolor; + dist += radius; + radius = max(0.5, tan_half_angle * dist); + } + + return color; +} + +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); + + position += normal * voxel_gi_instances.data[index].normal_bias; + + //this causes corrupted pixels, i have no idea why.. + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { + return; + } + + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); + float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); + //float blend=1.0; + + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; + + //radiance + +#define MAX_CONE_DIRS 4 + + vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( + vec3(0.707107, 0.0, 0.707107), + vec3(0.0, 0.707107, 0.707107), + vec3(-0.707107, 0.0, 0.707107), + vec3(0.0, -0.707107, 0.707107)); + + float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25); + float cone_angle_tan = 0.98269; + + vec3 light = vec3(0.0); + + for (int i = 0; i < MAX_CONE_DIRS; i++) { + vec3 dir = normalize((voxel_gi_instances.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz); + + vec4 cone_light = voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); + + if (voxel_gi_instances.data[index].blend_ambient) { + cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95)); + } + + light += cone_weights[i] * cone_light.rgb; + } + + light *= voxel_gi_instances.data[index].dynamic_range; + out_diff += vec4(light * blend, blend); + + //irradiance + vec4 irr_light = voxel_cone_trace(voxel_gi_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, voxel_gi_instances.data[index].bias); + if (voxel_gi_instances.data[index].blend_ambient) { + irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95)); + } + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range; + //irr_light=vec3(0.0); + + out_spec += vec4(irr_light.rgb * blend, blend); +} + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) { + cascade_pos += cam_normal * sdfgi.normal_bias; + + vec3 base_pos = floor(cascade_pos); + //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 diffuse_accum = vec4(0.0); + vec3 specular_accum; + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); + + vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + + vec3 specular_posf; + + if (use_specular) { + specular_accum = vec3(0.0); + specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + } + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute lightprobe texture position + + vec3 diffuse; + vec3 pos_uvw = diffuse_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + diffuse = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb; + + diffuse_accum += vec4(diffuse * weight, weight); + + if (use_specular) { + vec3 specular = vec3(0.0); + vec3 pos_uvw = specular_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + if (roughness < 0.99) { + specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; + } + if (roughness > 0.5) { + specular = mix(specular, textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb, (roughness - 0.5) * 2.0); + } + + specular_accum += specular * weight; + } + } + + if (diffuse_accum.a > 0.0) { + diffuse_accum.rgb /= diffuse_accum.a; + } + + diffuse_light = diffuse_accum.rgb; + + if (use_specular) { + if (diffuse_accum.a > 0.0) { + specular_accum /= diffuse_accum.a; + } + + specular_light = specular_accum; + } + + { + //process blend + float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; + float blend_to = blend_from + 2.0; + + vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; + + float len = length(inner_pos); + + inner_pos = abs(normalize(inner_pos)); + len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + + if (len >= blend_from) { + blend = smoothstep(blend_from, blend_to, len); + } else { + blend = 0.0; + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl new file mode 100644 index 0000000000..ef2fde7516 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -0,0 +1,1035 @@ +// Functions related to lighting + +// This returns the G_GGX function divided by 2 cos_theta_m, where in practice cos_theta_m is either N.L or N.V. +// We're dividing this factor off because the overall term we'll end up looks like +// (see, for example, the first unnumbered equation in B. Burley, "Physically Based Shading at Disney", SIGGRAPH 2012): +// +// F(L.V) D(N.H) G(N.L) G(N.V) / (4 N.L N.V) +// +// We're basically regouping this as +// +// F(L.V) D(N.H) [G(N.L)/(2 N.L)] [G(N.V) / (2 N.V)] +// +// and thus, this function implements the [G(N.m)/(2 N.m)] part with m = L or V. +// +// The contents of the D and G (G1) functions (GGX) are taken from +// E. Heitz, "Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs", J. Comp. Graph. Tech. 3 (2) (2014). +// Eqns 71-72 and 85-86 (see also Eqns 43 and 80). + +float G_GGX_2cos(float cos_theta_m, float alpha) { + // Schlick's approximation + // C. Schlick, "An Inexpensive BRDF Model for Physically-based Rendering", Computer Graphics Forum. 13 (3): 233 (1994) + // Eq. (19), although see Heitz (2014) the about the problems with his derivation. + // It nevertheless approximates GGX well with k = alpha/2. + float k = 0.5 * alpha; + return 0.5 / (cos_theta_m * (1.0 - k) + k); + + // float cos2 = cos_theta_m * cos_theta_m; + // float sin2 = (1.0 - cos2); + // return 1.0 / (cos_theta_m + sqrt(cos2 + alpha * alpha * sin2)); +} + +float D_GGX(float cos_theta_m, float alpha) { + float alpha2 = alpha * alpha; + float d = 1.0 + (alpha2 - 1.0) * cos_theta_m * cos_theta_m; + return alpha2 / (M_PI * d * d); +} + +float G_GGX_anisotropic_2cos(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { + float cos2 = cos_theta_m * cos_theta_m; + float sin2 = (1.0 - cos2); + float s_x = alpha_x * cos_phi; + float s_y = alpha_y * sin_phi; + return 1.0 / max(cos_theta_m + sqrt(cos2 + (s_x * s_x + s_y * s_y) * sin2), 0.001); +} + +float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { + float cos2 = cos_theta_m * cos_theta_m; + float sin2 = (1.0 - cos2); + float r_x = cos_phi / alpha_x; + float r_y = sin_phi / alpha_y; + float d = cos2 + sin2 * (r_x * r_x + r_y * r_y); + return 1.0 / max(M_PI * alpha_x * alpha_y * d * d, 0.001); +} + +float SchlickFresnel(float u) { + float m = 1.0 - u; + float m2 = m * m; + return m2 * m2 * m; // pow(m,5) +} + +float GTR1(float NdotH, float a) { + if (a >= 1.0) + return 1.0 / M_PI; + float a2 = a * a; + float t = 1.0 + (a2 - 1.0) * NdotH * NdotH; + return (a2 - 1.0) / (M_PI * log(a2) * t); +} + +vec3 F0(float metallic, float specular, vec3 albedo) { + float dielectric = 0.16 * specular * specular; + // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; + // see https://google.github.io/filament/Filament.md.html + return mix(vec3(dielectric), albedo, vec3(metallic)); +} + +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, + float transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 B, vec3 T, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + +#if defined(LIGHT_CODE_USED) + // light is written by the light shader + + vec3 normal = N; + vec3 light = L; + vec3 view = V; + +#CODE : LIGHT + +#else + + float NdotL = min(A + dot(N, L), 1.0); + float cNdotL = max(NdotL, 0.0); // clamped NdotL + float NdotV = dot(N, V); + float cNdotV = max(NdotV, 0.0); + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + vec3 H = normalize(V + L); +#endif + +#if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); +#endif + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); +#endif + + float metallic = unpackUnorm4x8(orms).z; + if (metallic < 1.0) { + float roughness = unpackUnorm4x8(orms).y; + float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance + +#if defined(DIFFUSE_LAMBERT_WRAP) + // energy conserving lambert wrap shader + diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); +#elif defined(DIFFUSE_TOON) + + diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); + +#elif defined(DIFFUSE_BURLEY) + + { + float FD90_minus_1 = 2.0 * cLdotH * cLdotH * roughness - 0.5; + float FdV = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotV); + float FdL = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotL); + diffuse_brdf_NL = (1.0 / M_PI) * FdV * FdL * cNdotL; + /* + float energyBias = mix(roughness, 0.0, 0.5); + float energyFactor = mix(roughness, 1.0, 1.0 / 1.51); + float fd90 = energyBias + 2.0 * VoH * VoH * roughness; + float f0 = 1.0; + float lightScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotL, 5.0); + float viewScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotV, 5.0); + + diffuse_brdf_NL = lightScatter * viewScatter * energyFactor; + */ + } +#else + // lambert + diffuse_brdf_NL = cNdotL * (1.0 / M_PI); +#endif + + diffuse_light += light_color * diffuse_brdf_NL * attenuation; + +#if defined(LIGHT_BACKLIGHT_USED) + diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; +#endif + +#if defined(LIGHT_RIM_USED) + float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); + diffuse_light += rim_light * rim * mix(vec3(1.0), rim_color, rim_tint) * light_color; +#endif + +#ifdef LIGHT_TRANSMITTANCE_USED + + { +#ifdef SSS_MODE_SKIN + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + + vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + + vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + + vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + + vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + + vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); + + diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#else + + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + diffuse_light += exp(dd) * transmittance_color.rgb * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#endif + } +#else + +#endif //LIGHT_TRANSMITTANCE_USED + } + + float roughness = unpackUnorm4x8(orms).y; + if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely + + // D + +#if defined(SPECULAR_BLINN) + + //normalized blinn + float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; + float blinn = pow(cNdotH, shininess); + blinn *= (shininess + 2.0) * (1.0 / (8.0 * M_PI)); + + specular_light += light_color * attenuation * specular_amount * blinn * f0 * unpackUnorm4x8(orms).w; + +#elif defined(SPECULAR_PHONG) + + vec3 R = normalize(-reflect(L, N)); + float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); + float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; + float phong = pow(cRdotV, shininess); + phong *= (shininess + 1.0) * (1.0 / (8.0 * M_PI)); + + specular_light += light_color * attenuation * specular_amount * phong * f0 * unpackUnorm4x8(orms).w; + +#elif defined(SPECULAR_TOON) + + vec3 R = normalize(-reflect(L, N)); + float RdotV = dot(R, V); + float mid = 1.0 - roughness; + mid *= mid; + float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; + diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection + +#elif defined(SPECULAR_DISABLED) + // none.. + +#elif defined(SPECULAR_SCHLICK_GGX) + // shlick+ggx as default + +#if defined(LIGHT_ANISOTROPY_USED) + + float alpha_ggx = roughness * roughness; + float aspect = sqrt(1.0 - anisotropy * 0.9); + float ax = alpha_ggx / aspect; + float ay = alpha_ggx * aspect; + float XdotH = dot(T, H); + float YdotH = dot(B, H); + float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); + float G = G_GGX_anisotropic_2cos(cNdotL, ax, ay, XdotH, YdotH) * G_GGX_anisotropic_2cos(cNdotV, ax, ay, XdotH, YdotH); + +#else + float alpha_ggx = roughness * roughness; + float D = D_GGX(cNdotH, alpha_ggx); + float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); +#endif + // F + float cLdotH5 = SchlickFresnel(cLdotH); + vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); + + vec3 specular_brdf_NL = cNdotL * D * F * G; + + specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; +#endif + +#if defined(LIGHT_CLEARCOAT_USED) + +#if !defined(SPECULAR_SCHLICK_GGX) + float cLdotH5 = SchlickFresnel(cLdotH); +#endif + float Dr = GTR1(cNdotH, mix(.1, .001, clearcoat_gloss)); + float Fr = mix(.04, 1.0, cLdotH5); + float Gr = G_GGX_2cos(cNdotL, .25) * G_GGX_2cos(cNdotV, .25); + + float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; + + specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; +#endif + } + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(1.0 - attenuation, 0.0, 1.0)); +#endif + +#endif //defined(LIGHT_CODE_USED) +} + +#ifndef USE_NO_SHADOWS + +// Interleaved Gradient Noise +// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +float quick_hash(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + return fract(magic.z * fract(dot(pos, magic.xy))); +} + +float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_directional_soft_shadow_samples == 1) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); +} + +float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec3 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 1) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_omni_pcf_shadow(texture2D shadow, float blur_scale, vec2 coord, vec4 uv_rect, vec2 flip_offset, float depth) { + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 1) { + vec2 pos = coord * 0.5 + 0.5; + pos = uv_rect.xy + pos * uv_rect.zw; + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + vec2 offset_scale = blur_scale * 2.0 * scene_data.shadow_atlas_pixel_size / uv_rect.zw; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + vec2 offset = offset_scale * (disk_rotation * scene_data.soft_shadow_kernel[i].xy); + vec2 sample_coord = coord + offset; + + float sample_coord_length_sqaured = dot(sample_coord, sample_coord); + bool do_flip = sample_coord_length_sqaured > 1.0; + + if (do_flip) { + float len = sqrt(sample_coord_length_sqaured); + sample_coord = sample_coord * (2.0 / len - 1.0); + } + + sample_coord = sample_coord * 0.5 + 0.5; + sample_coord = uv_rect.xy + sample_coord * uv_rect.zw; + + if (do_flip) { + sample_coord += flip_offset; + } + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(sample_coord, depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { + //find blocker + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < pssm_coord.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (pssm_coord.z - blocker_average) / blocker_average; + tex_scale *= penumbra; + + float s = 0.0; + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); + } + + return s / float(sc_directional_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + return 1.0; + } +} + +#endif //USE_NO_SHADOWS + +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef USE_NO_SHADOWS + if (omni_lights.data[idx].shadow_enabled) { + // there is a shadowmap + vec2 texel_size = scene_data.shadow_atlas_pixel_size; + vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; + base_uv_rect.xy += texel_size; + base_uv_rect.zw -= texel_size * 2.0; + + // Omni lights use direction.xy to store to store the offset between the two paraboloid regions + vec2 flip_offset = omni_lights.data[idx].direction.xy; + + vec3 local_vert = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_dir = normalize(local_vert); + + vec3 local_normal = normalize(mat3(omni_lights.data[idx].shadow_matrix) * normal); + vec3 normal_bias = local_normal * omni_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(local_normal, shadow_dir))); + + float shadow; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + vec3 basis_normal = shadow_dir; + vec3 v0 = abs(basis_normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, basis_normal)); + vec3 bitangent = normalize(cross(tangent, basis_normal)); + float z_norm = shadow_len * omni_lights.data[idx].inv_radius; + + tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; + if (d < z_norm) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + tangent *= penumbra; + bitangent *= penumbra; + + z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + pos = normalize(pos + normal_bias); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } else { + vec4 uv_rect = base_uv_rect; + + vec3 shadow_sample = normalize(shadow_dir + normal_bias); + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + flip_offset *= -1.0; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec2 pos = shadow_sample.xy / shadow_sample.z; + float depth = shadow_len - omni_lights.data[idx].shadow_bias; + depth *= omni_lights.data[idx].inv_radius; + shadow = sample_omni_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale / shadow_sample.z, pos, uv_rect, flip_offset, depth); + } + + return shadow; + } +#endif + + return 1.0; +} + +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); + float light_attenuation = omni_attenuation; + vec3 color = omni_lights.data[idx].color; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].size > 0.0) { + float t = omni_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; //no transmittance by default + transmittance_color.a *= light_attenuation; + { + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; + + //redo shadowmapping, but shrink the model a bit to avoid arctifacts + vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); + + float shadow_len = length(splane.xyz); + splane.xyz = normalize(splane.xyz); + + if (splane.z >= 0.0) { + splane.z += 1.0; + clamp_rect.y += clamp_rect.w; + } else { + splane.z = 1.0 - splane.z; + } + + splane.xy /= splane.z; + + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[idx].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data.shadow_atlas_pixel_size ); + splane.w = 1.0; //needed? i think it should be 1 already + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; + } +#endif + + if (sc_use_light_projector && omni_lights.data[idx].projector_rect != vec4(0.0)) { + vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + local_v = normalize(local_v); + + vec4 atlas_rect = omni_lights.data[idx].projector_rect; + + if (local_v.z >= 0.0) { + atlas_rect.y += atlas_rect.w; + } + + local_v.z = 1.0 + abs(local_v.z); + + local_v.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; + + if (sc_projector_use_mipmaps) { + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); + + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } + + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); + + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; + } + + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + } + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * omni_attenuation, rim_tint, rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + alpha, +#endif + diffuse_light, + specular_light); +} + +float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef USE_NO_SHADOWS + if (spot_lights.data[idx].shadow_enabled) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + vec3 spot_dir = spot_lights.data[idx].direction; + + vec3 shadow_dir = light_rel_vec / light_length; + vec3 normal_bias = normal * light_length * spot_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(normal, shadow_dir))); + + //there is a shadowmap + vec4 v = vec4(vertex + normal_bias, 1.0); + + vec4 splane = (spot_lights.data[idx].shadow_matrix * v); + splane.z -= spot_lights.data[idx].shadow_bias / (light_length * spot_lights.data[idx].inv_radius); + splane /= splane.w; + + float shadow; + if (sc_use_light_soft_shadows && spot_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; + + vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; + vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < splane.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + uv_size *= penumbra; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } else { + //hard shadow + vec3 shadow_uv = vec3(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z); + shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); + } + + return shadow; + } + +#endif //USE_NO_SHADOWS + + return 1.0; +} + +vec2 normal_to_panorama(vec3 n) { + n = normalize(n); + vec2 panorama_coords = vec2(atan(n.x, n.z), acos(-n.y)); + + if (panorama_coords.x < 0.0) { + panorama_coords.x += M_PI * 2.0; + } + + panorama_coords /= vec2(M_PI * 2.0, M_PI); + return panorama_coords; +} + +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#endif + inout vec3 diffuse_light, + inout vec3 specular_light) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); + vec3 spot_dir = spot_lights.data[idx].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); + spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); + float light_attenuation = spot_attenuation; + vec3 color = spot_lights.data[idx].color; + float specular_amount = spot_lights.data[idx].specular_amount; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && spot_lights.data[idx].size > 0.0) { + float t = spot_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + transmittance_color.a *= light_attenuation; + { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); + splane /= splane.w; + splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + + shadow_z = shadow_z * 2.0 - 1.0; + float z_far = 1.0 / spot_lights.data[idx].inv_radius; + float z_near = 0.01; + shadow_z = 2.0 * z_near * z_far / (z_far + z_near - shadow_z * (z_far - z_near)); + + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; + } +#endif //LIGHT_TRANSMITTANCE_USED + + if (sc_use_light_projector && spot_lights.data[idx].projector_rect != vec4(0.0)) { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + splane /= splane.w; + + vec2 proj_uv = normal_to_panorama(splane.xyz) * spot_lights.data[idx].projector_rect.zw; + + if (sc_projector_use_mipmaps) { + //ensure we have proper mipmaps + vec4 splane_ddx = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); + splane_ddx /= splane_ddx.w; + vec2 proj_uv_ddx = normal_to_panorama(splane_ddx.xyz) * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 splane_ddy = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); + splane_ddy /= splane_ddy.w; + vec2 proj_uv_ddy = normal_to_panorama(splane_ddy.xyz) * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * spot_attenuation, rim_tint, rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + alpha, +#endif + diffuse_light, specular_light); +} + +void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { + vec3 box_extents = reflections.data[ref_index].box_extents; + vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; + + if (any(greaterThan(abs(local_pos), box_extents))) { //out of the reflection box + return; + } + + vec3 ref_vec = normalize(reflect(vertex, normal)); + + vec3 inner_pos = abs(local_pos / box_extents); + float blend = max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + //make blend more rounded + blend = mix(length(inner_pos), blend, blend); + blend *= blend; + blend = max(0.0, 1.0 - blend); + + if (reflections.data[ref_index].intensity > 0.0) { // compute reflection + + vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; + + if (reflections.data[ref_index].box_project) { //box project + + vec3 nrdir = normalize(local_ref_vec); + vec3 rbmax = (box_extents - local_pos) / nrdir; + vec3 rbmin = (-box_extents - local_pos) / nrdir; + + vec3 rbminmax = mix(rbmin, rbmax, greaterThan(nrdir, vec3(0.0, 0.0, 0.0))); + + float fa = min(min(rbminmax.x, rbminmax.y), rbminmax.z); + vec3 posonbox = local_pos + nrdir * fa; + local_ref_vec = posonbox - reflections.data[ref_index].box_offset; + } + + vec4 reflection; + + reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb; + + if (reflections.data[ref_index].exterior) { + reflection.rgb = mix(specular_light, reflection.rgb, blend); + } + + reflection.rgb *= reflections.data[ref_index].intensity; //intensity + reflection.a = blend; + reflection.rgb *= reflection.a; + + reflection_accum += reflection; + } + + switch (reflections.data[ref_index].ambient_mode) { + case REFLECTION_AMBIENT_DISABLED: { + //do nothing + } break; + case REFLECTION_AMBIENT_ENVIRONMENT: { + //do nothing + vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz; + + vec4 ambient_out; + + ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb; + ambient_out.a = blend; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + case REFLECTION_AMBIENT_COLOR: { + vec4 ambient_out; + ambient_out.a = blend; + ambient_out.rgb = reflections.data[ref_index].ambient; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + } +} + +float blur_shadow(float shadow) { + return shadow; +#if 0 + //disabling for now, will investigate later + float interp_shadow = shadow; + if (gl_HelperInvocation) { + interp_shadow = -4.0; // technically anything below -4 will do but just to make sure + } + + uvec2 fc2 = uvec2(gl_FragCoord.xy); + interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); + interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); + + if (interp_shadow >= 0.0) { + shadow = interp_shadow; + } + return shadow; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl new file mode 100644 index 0000000000..a2a54a0511 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -0,0 +1,1563 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +/* INPUT ATTRIBS */ + +layout(location = 0) in vec3 vertex_attrib; + +//only for pure render depth when normal is not used + +#ifdef NORMAL_USED +layout(location = 1) in vec3 normal_attrib; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 2) in vec4 tangent_attrib; +#endif + +#if defined(COLOR_USED) +layout(location = 3) in vec4 color_attrib; +#endif + +#ifdef UV_USED +layout(location = 4) in vec2 uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) || defined(MODE_RENDER_MATERIAL) +layout(location = 5) in vec2 uv2_attrib; +#endif // MODE_RENDER_MATERIAL + +#if defined(CUSTOM0_USED) +layout(location = 6) in vec4 custom0_attrib; +#endif + +#if defined(CUSTOM1_USED) +layout(location = 7) in vec4 custom1_attrib; +#endif + +#if defined(CUSTOM2_USED) +layout(location = 8) in vec4 custom2_attrib; +#endif + +#if defined(CUSTOM3_USED) +layout(location = 9) in vec4 custom3_attrib; +#endif + +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 10) in uvec4 bone_attrib; +#endif + +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 11) in vec4 weight_attrib; +#endif + +/* Varyings */ + +layout(location = 0) highp out vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump out vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump out vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump out vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump out vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 8) out highp float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +invariant gl_Position; + +#GLOBALS + +void main() { + vec4 instance_custom = vec4(0.0); +#if defined(COLOR_USED) + color_interp = color_attrib; +#endif + + bool is_multimesh = bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH); + + mat4 world_matrix = draw_call.transform; + + mat3 world_normal_matrix; + if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + world_normal_matrix = inverse(mat3(world_matrix)); + } else { + world_normal_matrix = mat3(world_matrix); + } + + if (is_multimesh) { + //multimesh, instances are for it + + mat4 matrix; + +#ifdef USE_PARTICLE_TRAILS + uint trail_size = (draw_call.flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; + uint stride = 3 + 1 + 1; //particles always uses this format + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + offset += 2; + } else { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], transforms.data[offset + 2], vec4(0.0, 0.0, 0.0, 1.0)); + offset += 3; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { +#ifdef COLOR_USED + color_interp *= transforms.data[offset]; +#endif + offset += 1; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + +#endif + //transpose + matrix = transpose(matrix); + world_matrix = world_matrix * matrix; + world_normal_matrix = world_normal_matrix * mat3(matrix); + } + + vec3 vertex = vertex_attrib; +#ifdef NORMAL_USED + vec3 normal = normal_attrib * 2.0 - 1.0; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec3 tangent = tangent_attrib.xyz * 2.0 - 1.0; + float binormalf = tangent_attrib.a * 2.0 - 1.0; + vec3 binormal = normalize(cross(normal, tangent) * binormalf); +#endif + +#ifdef UV_USED + uv_interp = uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + uv2_interp = uv2_attrib; +#endif + +#ifdef OVERRIDE_POSITION + vec4 position; +#endif + +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else + mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (world_matrix * vec4(vertex, 1.0)).xyz; + + normal = world_normal_matrix * normal; + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + tangent = world_normal_matrix * tangent; + binormal = world_normal_matrix * binormal; + +#endif +#endif + + float roughness = 1.0; + + mat4 modelview = scene_data.inv_camera_matrix * world_matrix; + mat3 modelview_normal = mat3(scene_data.inv_camera_matrix) * world_normal_matrix; + + { +#CODE : VERTEX + } + + /* output */ + +// using local coordinates (default) +#if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) + + vertex = (modelview * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = modelview_normal * normal; +#endif + +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + binormal = modelview_normal * binormal; + tangent = modelview_normal * tangent; +#endif + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (scene_data.inv_camera_matrix * vec4(vertex, 1.0)).xyz; + normal = mat3(scene_data.inverse_normal_matrix) * normal; + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + binormal = mat3(scene_data.camera_inverse_binormal_matrix) * binormal; + tangent = mat3(scene_data.camera_inverse_tangent_matrix) * tangent; +#endif +#endif + + vertex_interp = vertex; +#ifdef NORMAL_USED + normal_interp = normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + tangent_interp = tangent; + binormal_interp = binormal; +#endif + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_DUAL_PARABOLOID + + vertex_interp.z *= scene_data.dual_paraboloid_side; + + dp_clip = vertex_interp.z; //this attempts to avoid noise caused by objects sent to the other parabolloid side due to bias + + //for dual paraboloid shadow mapping, this is the fastest but least correct way, as it curves straight edges + + vec3 vtx = vertex_interp; + float distance = length(vtx); + vtx = normalize(vtx); + vtx.xy /= 1.0 - vtx.z; + vtx.z = (distance / scene_data.z_far); + vtx.z = vtx.z * 2.0 - 1.0; + vertex_interp = vtx; + +#endif + +#endif //MODE_RENDER_DEPTH + +#ifdef OVERRIDE_POSITION + gl_Position = position; +#else + gl_Position = projection_matrix * vec4(vertex_interp, 1.0); +#endif // OVERRIDE_POSITION + +#ifdef MODE_RENDER_DEPTH + if (scene_data.pancake_shadows) { + if (gl_Position.z <= 0.00001) { + gl_Position.z = 0.00001; + } + } +#endif // MODE_RENDER_DEPTH +#ifdef MODE_RENDER_MATERIAL + if (scene_data.material_uv2_mode) { + vec2 uv_offset = draw_call.lightmap_uv_scale.xy; // we are abusing lightmap_uv_scale here, we shouldn't have a lightmap during a depth pass... + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; + gl_Position.z = 0.00001; + gl_Position.w = 1.0; + } +#endif // MODE_RENDER_MATERIAL +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +/* Specialization Constants */ + +#if !defined(MODE_RENDER_DEPTH) + +#if !defined(MODE_UNSHADED) + +layout(constant_id = 0) const bool sc_use_light_projector = false; +layout(constant_id = 1) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 2) const bool sc_use_directional_soft_shadows = false; + +layout(constant_id = 3) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 4) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 5) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 6) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const bool sc_projector_use_mipmaps = true; + +layout(constant_id = 9) const bool sc_disable_omni_lights = false; +layout(constant_id = 10) const bool sc_disable_spot_lights = false; +layout(constant_id = 11) const bool sc_disable_reflection_probes = false; +layout(constant_id = 12) const bool sc_disable_directional_lights = false; + +#endif //!MODE_UNSHADED + +layout(constant_id = 7) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 13) const bool sc_disable_decals = false; +layout(constant_id = 14) const bool sc_disable_fog = false; + +#endif //!MODE_RENDER_DEPTH + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +/* Varyings */ + +layout(location = 0) highp in vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump in vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump in vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump in vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump in vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 8) highp in float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +//defines to keep compatibility with vertex + +#define world_matrix draw_call.transform +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#else +#define projection_matrix scene_data.projection_matrix +#endif + +#if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) +//both required for transmittance to be enabled +#define LIGHT_TRANSMITTANCE_USED +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#GLOBALS + +/* clang-format on */ + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + +layout(location = 0) out vec4 albedo_output_buffer; +layout(location = 1) out vec4 normal_output_buffer; +layout(location = 2) out vec4 orm_output_buffer; +layout(location = 3) out vec4 emission_output_buffer; +layout(location = 4) out float depth_output_buffer; + +#endif // MODE_RENDER_MATERIAL + +#else // RENDER DEPTH + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness +layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) +#else + +layout(location = 0) out mediump vec4 frag_color; +#endif // MODE_MULTIPLE_RENDER_TARGETS + +#endif // RENDER DEPTH + +#include "scene_forward_aa_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + +#include "scene_forward_lights_inc.glsl" + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifndef MODE_RENDER_DEPTH + +/* + Only supporting normal fog here. +*/ + +vec4 fog_process(vec3 vertex) { + vec3 fog_color = scene_data.fog_light_color; + + if (scene_data.fog_aerial_perspective > 0.0) { + vec3 sky_fog_color = vec3(0.0); + vec3 cube_view = scene_data.radiance_inverse_xform * vertex; + // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near)); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + float lod, blend; + blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); + sky_fog_color = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod)).rgb; + sky_fog_color = mix(sky_fog_color, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod + 1)).rgb, blend); +#else + sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + fog_color = mix(fog_color, sky_fog_color, scene_data.fog_aerial_perspective); + } + + if (scene_data.fog_sun_scatter > 0.001) { + vec4 sun_scatter = vec4(0.0); + float sun_total = 0.0; + vec3 view = normalize(vertex); + + for (uint i = 0; i < scene_data.directional_light_count; i++) { + vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; + float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); + fog_color += light_color * light_amount * scene_data.fog_sun_scatter; + } + } + + float fog_amount = 1.0 - exp(min(0.0, vertex.z * scene_data.fog_density)); + + if (abs(scene_data.fog_height_density) > 0.001) { + float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; + + float y_dist = scene_data.fog_height - y; + + float vfog_amount = clamp(exp(y_dist * scene_data.fog_height_density), 0.0, 1.0); + + fog_amount = max(vfog_amount, fog_amount); + } + + return vec4(fog_color, fog_amount); +} + +#endif //!MODE_RENDER DEPTH + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + //lay out everything, whathever is unused is optimized away anyway + vec3 vertex = vertex_interp; + vec3 view = -normalize(vertex_interp); + vec3 albedo = vec3(1.0); + vec3 backlight = vec3(0.0); + vec4 transmittance_color = vec4(0.0); + float transmittance_depth = 0.0; + float transmittance_boost = 0.0; + float metallic = 0.0; + float specular = 0.5; + vec3 emission = vec3(0.0); + float roughness = 1.0; + float rim = 0.0; + float rim_tint = 0.0; + float clearcoat = 0.0; + float clearcoat_gloss = 0.0; + float anisotropy = 0.0; + vec2 anisotropy_flow = vec2(1.0, 0.0); + vec4 fog = vec4(0.0); +#if defined(CUSTOM_RADIANCE_USED) + vec4 custom_radiance = vec4(0.0); +#endif +#if defined(CUSTOM_IRRADIANCE_USED) + vec4 custom_irradiance = vec4(0.0); +#endif + + float ao = 1.0; + float ao_light_affect = 0.0; + + float alpha = 1.0; + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec3 binormal = normalize(binormal_interp); + vec3 tangent = normalize(tangent_interp); +#else + vec3 binormal = vec3(0.0); + vec3 tangent = vec3(0.0); +#endif + +#ifdef NORMAL_USED + vec3 normal = normalize(normal_interp); + +#if defined(DO_SIDE_CHECK) + if (!gl_FrontFacing) { + normal = -normal; + } +#endif + +#endif //NORMAL_USED + +#ifdef UV_USED + vec2 uv = uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + vec2 uv2 = uv2_interp; +#endif + +#if defined(COLOR_USED) + vec4 color = color_interp; +#endif + +#if defined(NORMAL_MAP_USED) + + vec3 normal_map = vec3(0.5); +#endif + + float normal_map_depth = 1.0; + + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size + scene_data.screen_pixel_size * 0.5; //account for center + + float sss_strength = 0.0; + +#ifdef ALPHA_SCISSOR_USED + float alpha_scissor_threshold = 1.0; +#endif // ALPHA_SCISSOR_USED + +#ifdef ALPHA_HASH_USED + float alpha_hash_scale = 1.0; +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + float alpha_antialiasing_edge = 0.0; + vec2 alpha_texture_coordinate = vec2(0.0, 0.0); +#endif // ALPHA_ANTIALIASING_EDGE_USED + + { +#CODE : FRAGMENT + } + +#ifdef LIGHT_TRANSMITTANCE_USED +#ifdef SSS_MODE_SKIN + transmittance_color.a = sss_strength; +#else + transmittance_color.a *= sss_strength; +#endif +#endif + +#ifndef USE_SHADOW_TO_OPACITY + +#ifdef ALPHA_SCISSOR_USED + if (alpha < alpha_scissor_threshold) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +// alpha hash can be used in unison with alpha antialiasing +#ifdef ALPHA_HASH_USED + if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + discard; + } +#endif // ALPHA_HASH_USED + +// If we are not edge antialiasing, we need to remove the output alpha channel from scissor and hash +#if (defined(ALPHA_SCISSOR_USED) || defined(ALPHA_HASH_USED)) && !defined(ALPHA_ANTIALIASING_EDGE_USED) + alpha = 1.0; +#endif + +#ifdef ALPHA_ANTIALIASING_EDGE_USED +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather +#ifdef ALPHA_SCISSOR_USED + alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); +#endif + alpha = compute_alpha_antialiasing_edge(alpha, alpha_texture_coordinate, alpha_antialiasing_edge); +#endif // ALPHA_ANTIALIASING_EDGE_USED + +#ifdef USE_OPAQUE_PREPASS + if (alpha < opaque_prepass_threshold) { + discard; + } +#endif // USE_OPAQUE_PREPASS + +#endif // !USE_SHADOW_TO_OPACITY + +#ifdef NORMAL_MAP_USED + + normal_map.xy = normal_map.xy * 2.0 - 1.0; + normal_map.z = sqrt(max(0.0, 1.0 - dot(normal_map.xy, normal_map.xy))); //always ignore Z, as it can be RG packed, Z may be pos/neg, etc. + + normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_map_depth)); + +#endif + +#ifdef LIGHT_ANISOTROPY_USED + + if (anisotropy > 0.01) { + //rotation matrix + mat3 rot = mat3(tangent, binormal, normal); + //make local to space + tangent = normalize(rot * vec3(anisotropy_flow.x, anisotropy_flow.y, 0.0)); + binormal = normalize(rot * vec3(-anisotropy_flow.y, anisotropy_flow.x, 0.0)); + } + +#endif + +#ifdef ENABLE_CLIP_ALPHA + if (albedo.a < 0.99) { + //used for doublepass and shadowmapping + discard; + } +#endif + + /////////////////////// FOG ////////////////////// +#ifndef MODE_RENDER_DEPTH + +#ifndef CUSTOM_FOG_USED + // fog must be processed as early as possible and then packed. + // to maximize VGPR usage + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + + if (!sc_disable_fog && scene_data.fog_enabled) { + fog = fog_process(vertex); + } + +#endif //!CUSTOM_FOG_USED + + uint fog_rg = packHalf2x16(fog.rg); + uint fog_ba = packHalf2x16(fog.ba); + +#endif //!MODE_RENDER_DEPTH + + /////////////////////// DECALS //////////////////////////////// + +#ifndef MODE_RENDER_DEPTH + + vec3 vertex_ddx = dFdx(vertex); + vec3 vertex_ddy = dFdy(vertex); + + if (!sc_disable_decals) { //Decals + // must implement + + uint decal_indices = draw_call.decals.x; + for (uint i = 0; i < 8; i++) { + uint decal_index = decal_indices & 0xFF; + if (i == 4) { + decal_indices = draw_call.decals.y; + } else { + decal_indices = decal_indices >> 8; + } + + if (decal_index == 0xFF) { + break; + } + + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal + } + + float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + + if (decals.data[decal_index].normal_fade > 0.0) { + fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); + } + + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } + decal_albedo *= decals.data[decal_index].modulate; + decal_albedo.a *= fade; + albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); + + if (decals.data[decal_index].normal_rect != vec4(0.0)) { + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } + ao = mix(ao, decal_orm.r, decal_albedo.a); + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + } + } + } + } //Decals +#endif //!MODE_RENDER_DEPTH + + /////////////////////// LIGHTING ////////////////////////////// + +#ifdef NORMAL_USED + if (scene_data.roughness_limiter_enabled) { + //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + float roughness2 = roughness * roughness; + vec3 dndu = dFdx(normal), dndv = dFdy(normal); + float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); + float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect + float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); + roughness = sqrt(filteredRoughness2); + } +#endif // NORMAL_USED + //apply energy conservation + + vec3 specular_light = vec3(0.0, 0.0, 0.0); + vec3 diffuse_light = vec3(0.0, 0.0, 0.0); + vec3 ambient_light = vec3(0.0, 0.0, 0.0); + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (scene_data.use_reflection_cubemap) { + vec3 ref_vec = reflect(-view, normal); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness * MAX_ROUGHNESS_LOD, lod); + specular_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + specular_light = mix(specular_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else // USE_RADIANCE_CUBEMAP_ARRAY + specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + specular_light *= horizon * horizon; + specular_light *= scene_data.ambient_light_color_energy.a; + } + +#if defined(CUSTOM_RADIANCE_USED) + specular_light = mix(specular_light, custom_radiance.rgb, custom_radiance.a); +#endif // CUSTOM_RADIANCE_USED + +#ifndef USE_LIGHTMAP + //lightmap overrides everything + if (scene_data.use_ambient_light) { + ambient_light = scene_data.ambient_light_color_energy.rgb; + + if (scene_data.use_ambient_cubemap) { + vec3 ambient_dir = scene_data.radiance_inverse_xform * normal; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + vec3 cubemap_ambient = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ambient_dir, MAX_ROUGHNESS_LOD)).rgb; +#else + vec3 cubemap_ambient = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ambient_dir, MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + + ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); + } + } +#endif // !USE_LIGHTMAP + +#if defined(CUSTOM_IRRADIANCE_USED) + ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); +#endif // CUSTOM_IRRADIANCE_USED + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + //radiance + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef USE_LIGHTMAP + + //lightmap + if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture + uint index = draw_call.gi_offset; + + vec3 wnormal = mat3(scene_data.camera_matrix) * normal; + const float c1 = 0.429043; + const float c2 = 0.511664; + const float c3 = 0.743125; + const float c4 = 0.886227; + const float c5 = 0.247708; + ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); + + } else if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap + bool uses_sh = bool(draw_call.flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); + uint ofs = draw_call.gi_offset & 0xFFFF; + vec3 uvw; + uvw.xy = uv2 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; + uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + + if (uses_sh) { + uvw.z *= 4.0; //SH textures use 4 times more data + vec3 lm_light_l0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 0.0), 0.0).rgb; + vec3 lm_light_l1n1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 1.0), 0.0).rgb; + vec3 lm_light_l1_0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 2.0), 0.0).rgb; + vec3 lm_light_l1p1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 3.0), 0.0).rgb; + + uint idx = draw_call.gi_offset >> 20; + vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); + + ambient_light += lm_light_l0 * 0.282095f; + ambient_light += lm_light_l1n1 * 0.32573 * n.y; + ambient_light += lm_light_l1_0 * 0.32573 * n.z; + ambient_light += lm_light_l1p1 * 0.32573 * n.x; + if (metallic > 0.01) { // since the more direct bounced light is lost, we can kind of fake it with this trick + vec3 r = reflect(normalize(-vertex), normal); + specular_light += lm_light_l1n1 * 0.32573 * r.y; + specular_light += lm_light_l1_0 * 0.32573 * r.z; + specular_light += lm_light_l1p1 * 0.32573 * r.x; + } + + } else { + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; + } + } + + // No GI nor non low end mode... + +#endif // USE_LIGHTMAP + + // skipping ssao, do we remove ssao totally? + + if (!sc_disable_reflection_probes) { //Reflection probes + vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); + vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); + + uint reflection_indices = draw_call.reflection_probes.x; + for (uint i = 0; i < 8; i++) { + uint reflection_index = reflection_indices & 0xFF; + if (i == 4) { + reflection_indices = draw_call.reflection_probes.y; + } else { + reflection_indices = reflection_indices >> 8; + } + + if (reflection_index == 0xFF) { + break; + } + + reflection_process(reflection_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + } + + if (reflection_accum.a > 0.0) { + specular_light = reflection_accum.rgb / reflection_accum.a; + } + } //Reflection probes + + // finalize ambient light here + ambient_light *= albedo.rgb; + ambient_light *= ao; + + // convert ao to direct light ao + ao = mix(1.0, ao, ao_light_affect); + + //this saves some VGPRs + vec3 f0 = F0(metallic, specular, albedo); + + { +#if defined(DIFFUSE_TOON) + //simplify for toon, as + specular_light *= specular * metallic * albedo * 2.0; +#else + + // scales the specular reflections, needs to be computed before lighting happens, + // but after environment, GI, and reflection probes are added + // Environment brdf approximation (Lazarov 2013) + // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile + const vec4 c0 = vec4(-1.0, -0.0275, -0.572, 0.022); + const vec4 c1 = vec4(1.0, 0.0425, 1.04, -0.04); + vec4 r = roughness * c0 + c1; + float ndotv = clamp(dot(normal, view), 0.0, 1.0); + float a004 = min(r.x * r.x, exp2(-9.28 * ndotv)) * r.x + r.y; + vec2 env = vec2(-1.04, 1.04) * a004 + r.zw; + + specular_light *= env.x * f0 + env.y; +#endif + } + +#endif // !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#if !defined(MODE_RENDER_DEPTH) + //this saves some VGPRs + uint orms = packUnorm4x8(vec4(ao, roughness, metallic, specular)); +#endif + +// LIGHTING +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (!sc_disable_directional_lights) { //directional light + + // Do shadow and lighting in two passes to reduce register pressure + uint shadow0 = 0; + uint shadow1 = 0; + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + float shadow = 1.0; + + // Directional light shadow code is basically the same as forward clustered at this point in time minus `LIGHT_TRANSMITTANCE_USED` support. + // Not sure if there is a reason to change this seeing directional lights are part of our global data + // Should think about whether we may want to move this code into an include file or function?? + +#ifdef USE_SOFT_SHADOWS + //version with soft shadows, more expensive + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + vec3 shadow_color = vec3(0.0); + vec3 light_dir = directional_lights.data[i].direction; + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))) * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.x; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color1.rgb; + + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color2.rgb; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color3.rgb; + + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color4.rgb; + } + + if (directional_lights.data[i].blend_splits) { + vec3 shadow_color_blend = vec3(0.0); + float pssm_blend; + float shadow2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + + shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + shadow_color_blend = directional_lights.data[i].shadow_color4.rgb; + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_blend = sqrt(pssm_blend); + + shadow = mix(shadow, shadow2, pssm_blend); + shadow_color = mix(shadow_color, shadow_color_blend, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#else + // Soft shadow disabled version + + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + } + + pssm_coord /= pssm_coord.w; + + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#endif + + if (i < 4) { + shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); + } else { + shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); + } + } + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + // We're not doing light transmittence + + float shadow = 1.0; + + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } + + blur_shadow(shadow); + + light_compute(normal, directional_lights.data[i].direction, normalize(view), 0.0, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* not supported here +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, rim_tint, albedo, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif +#ifdef USE_SOFT_SHADOW + directional_lights.data[i].size, +#endif +#ifdef USE_SHADOW_TO_OPACITY + alpha, +#endif + diffuse_light, + specular_light); + } + } //directional light + + if (!sc_disable_omni_lights) { //omni lights + uint light_indices = draw_call.omni_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.omni_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_omni_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, + albedo, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, binormal, anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + alpha, +#endif + diffuse_light, specular_light); + } + } //omni lights + + if (!sc_disable_spot_lights) { //spot lights + + uint light_indices = draw_call.spot_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.spot_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_spot_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, + albedo, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, binormal, anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + alpha, +#endif + diffuse_light, specular_light); + } + } //spot lights + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + +#if defined(ALPHA_SCISSOR_USED) + if (alpha < alpha_scissor) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +#ifdef USE_OPAQUE_PREPASS + + if (alpha < opaque_prepass_threshold) { + discard; + } + +#endif // USE_OPAQUE_PREPASS + +#endif // USE_SHADOW_TO_OPACITY + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; + + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; + + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; + + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; +#endif // MODE_RENDER_MATERIAL + +#else // MODE_RENDER_DEPTH + + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end + + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; + + // apply metallic + metallic = unpackUnorm4x8(orms).z; + diffuse_light *= 1.0 - metallic; + ambient_light *= 1.0 - metallic; + + //restore fog + fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + diffuse_buffer = vec4(albedo.rgb, 0.0); + specular_buffer = vec4(0.0); + +#else // MODE_UNSHADED + +#ifdef SSS_MODE_SKIN + sss_strength = -sss_strength; +#endif // SSS_MODE_SKIN + diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength); + specular_buffer = vec4(specular_light, metallic); +#endif // MODE_UNSHADED + + diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); + specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); + +#else //MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + frag_color = vec4(albedo, alpha); +#else // MODE_UNSHADED + frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); + //frag_color = vec4(1.0); +#endif // MODE_UNSHADED + + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + +#endif //MODE_MULTIPLE_RENDER_TARGETS + +#endif //MODE_RENDER_DEPTH +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl new file mode 100644 index 0000000000..dd8879acb4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -0,0 +1,225 @@ +#define M_PI 3.14159265359 +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#include "decal_data_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) +#ifndef NORMAL_USED +#define NORMAL_USED +#endif +#endif + +/* don't exceed 128 bytes!! */ +/* put instance data into our push content, not a array */ +layout(push_constant, binding = 0, std430) uniform DrawCall { + highp mat4 transform; // 64 - 64 + uint flags; // 04 - 68 + uint instance_uniforms_ofs; //base offset in global buffer for instance variables // 04 - 72 + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) // 04 - 76 + uint layer_mask; // 04 - 80 + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + + uvec2 reflection_probes; // 08 - 104 + uvec2 omni_lights; // 08 - 112 + uvec2 spot_lights; // 08 - 120 + uvec2 decals; // 08 - 128 +} +draw_call; + +/* Set 0: Base Pass (never changes) */ + +#include "light_data_inc.glsl" + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 2) uniform sampler shadow_sampler; + +layout(set = 0, binding = 3) uniform sampler decal_sampler; +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) +#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) +#define INSTANCE_FLAGS_USE_SDFGI (1 << 7) +#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) +#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 9) +#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 10) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 11) +#define INSTANCE_FLAGS_MULTIMESH (1 << 12) +#define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) +#define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) +#define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 +//3 bits of stride +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF + +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { + ReflectionData data[]; +} +reflections; + +layout(set = 0, binding = 8, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +#define LIGHTMAP_FLAG_USE_DIRECTION 1 +#define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 + +struct Lightmap { + mediump mat3 normal_xform; +}; + +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { + Lightmap data[]; +} +lightmaps; + +struct LightmapCapture { + mediump vec4 sh[9]; +}; + +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { + LightmapCapture data[]; +} +lightmap_captures; + +layout(set = 0, binding = 11) uniform mediump texture2D decal_atlas; +layout(set = 0, binding = 12) uniform mediump texture2D decal_atlas_srgb; + +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { + DecalData data[]; +} +decals; + +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { + highp vec4 data[]; +} +global_variables; + +/* Set 1: Render Pass (changes per render pass) */ + +layout(set = 1, binding = 0, std140) uniform SceneData { + highp mat4 projection_matrix; + highp mat4 inv_projection_matrix; + highp mat4 camera_matrix; + highp mat4 inv_camera_matrix; + + // only used for multiview + highp mat4 projection_matrix_view[MAX_VIEWS]; + highp mat4 inv_projection_matrix_view[MAX_VIEWS]; + + highp vec2 viewport_size; + highp vec2 screen_pixel_size; + + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. + highp vec4 directional_penumbra_shadow_kernel[32]; + highp vec4 directional_soft_shadow_kernel[32]; + highp vec4 penumbra_shadow_kernel[32]; + highp vec4 soft_shadow_kernel[32]; + + mediump vec4 ambient_light_color_energy; + + mediump float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; + + mediump mat3 radiance_inverse_xform; + + highp vec2 shadow_atlas_pixel_size; + highp vec2 directional_shadow_pixel_size; + + uint directional_light_count; + mediump float dual_paraboloid_side; + highp float z_far; + highp float z_near; + + bool ssao_enabled; + mediump float ssao_light_affect; + mediump float ssao_ao_affect; + bool roughness_limiter_enabled; + + mediump float roughness_limiter_amount; + mediump float roughness_limiter_limit; + uvec2 roughness_limiter_pad; + + mediump vec4 ao_color; + + bool fog_enabled; + highp float fog_density; + highp float fog_height; + highp float fog_height_density; + + mediump vec3 fog_light_color; + mediump float fog_sun_scatter; + + mediump float fog_aerial_perspective; + bool material_uv2_mode; + + highp float time; + mediump float reflection_multiplier; // one normally, zero when rendering reflections + + bool pancake_shadows; + uint pad1; + uint pad2; + uint pad3; +} +scene_data; + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + +layout(set = 1, binding = 2) uniform mediump textureCubeArray radiance_cubemap; + +#else + +layout(set = 1, binding = 2) uniform mediump textureCube radiance_cubemap; + +#endif + +layout(set = 1, binding = 3) uniform mediump textureCubeArray reflection_atlas; + +layout(set = 1, binding = 4) uniform highp texture2D shadow_atlas; + +layout(set = 1, binding = 5) uniform highp texture2D directional_shadow_atlas; + +// this needs to change to providing just the lightmap we're using.. +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; + +layout(set = 1, binding = 9) uniform highp texture2D depth_buffer; +layout(set = 1, binding = 10) uniform mediump texture2D color_buffer; + +/* Set 2 Skeleton & Instancing (can change per item) */ + +layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { + highp vec4 data[]; +} +transforms; + +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl index 06dc4b13de..78e0a85341 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl index a5afe74cb2..62d1cffb0a 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl index 218605a962..2328effe7b 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -36,12 +36,12 @@ void main() { float divisor = 0.0; vec4 color; float depth; - vec3 normal; + vec4 normal; if (params.filtered) { color = vec4(0.0); depth = 0.0; - normal = vec3(0.0); + normal = vec4(0.0); for (int i = 0; i < 4; i++) { ivec2 ofs = ssC << 1; @@ -53,7 +53,9 @@ void main() { } color += texelFetch(source_ssr, ofs, 0); float d = texelFetch(source_depth, ofs, 0).r; - normal += texelFetch(source_normal, ofs, 0).xyz * 2.0 - 1.0; + vec4 nr = texelFetch(source_normal, ofs, 0); + normal.xyz += nr.xyz * 2.0 - 1.0; + normal.w += nr.w; d = d * 2.0 - 1.0; if (params.orthogonal) { @@ -66,11 +68,12 @@ void main() { color /= 4.0; depth /= 4.0; - normal = normalize(normal / 4.0) * 0.5 + 0.5; + normal.xyz = normalize(normal.xyz / 4.0) * 0.5 + 0.5; + normal.w /= 4.0; } else { color = texelFetch(source_ssr, ssC << 1, 0); depth = texelFetch(source_depth, ssC << 1, 0).r; - normal = texelFetch(source_normal, ssC << 1, 0).xyz; + normal = texelFetch(source_normal, ssC << 1, 0); depth = depth * 2.0 - 1.0; if (params.orthogonal) { @@ -83,5 +86,5 @@ void main() { imageStore(dest_ssr, ssC, color); imageStore(dest_depth, ssC, vec4(depth)); - imageStore(dest_normal, ssC, vec4(normal, 0.0)); + imageStore(dest_normal, ssC, normal); } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index e4c3f3a84b..8b58796962 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl index 08da283dad..0eacbc5363 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define MAX_CASCADES 8 @@ -153,7 +153,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) out vec4 frag_color; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index dc7238abed..d6e5c6a92e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -20,10 +20,10 @@ layout(set = 0, binding = 3, std430) restrict readonly buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours - uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours - uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; // rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors. + uint light; // rgbe8985 encoded total saved light, extra 2 bits for neighbors. + uint light_aniso; // 55555 light anisotropy, extra 2 bits for neighbors. //total neighbours: 26 }; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl deleted file mode 100644 index 69d8824d8a..0000000000 --- a/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl +++ /dev/null @@ -1,182 +0,0 @@ -/* clang-format off */ -[compute] - -#version 450 - -VERSION_DEFINES - -layout(local_size_x = OCT_RES, local_size_y = OCT_RES, local_size_z = 1) in; - -/* clang-format on */ - -#define MAX_CASCADES 8 - -layout(rgba16f, set = 0, binding = 1) uniform restrict image2DArray irradiance_texture; -layout(rg16f, set = 0, binding = 2) uniform restrict image2DArray depth_texture; - -layout(rgba32ui, set = 0, binding = 3) uniform restrict uimage2DArray irradiance_history_texture; -layout(rg32ui, set = 0, binding = 4) uniform restrict uimage2DArray depth_history_texture; - -struct CascadeData { - vec3 offset; //offset of (0,0,0) in world coordinates - float to_cell; // 1/bounds * grid_size -}; - -layout(set = 0, binding = 5, std140) uniform Cascades { - CascadeData data[MAX_CASCADES]; -} -cascades; - -#define DEPTH_HISTORY_BITS 24 -#define IRRADIANCE_HISTORY_BITS 16 - -layout(push_constant, binding = 0, std430) uniform Params { - vec3 grid_size; - uint max_cascades; - - uint probe_axis_size; - uint cascade; - uint history_size; - uint pad0; - - ivec3 scroll; //scroll in probes - uint pad1; -} -params; - -void main() { - ivec2 local = ivec2(gl_LocalInvocationID.xy); - ivec2 probe = ivec2(gl_WorkGroupID.xy); - - ivec3 probe_cell; - probe_cell.x = probe.x % int(params.probe_axis_size); - probe_cell.y = probe.y; - probe_cell.z = probe.x / int(params.probe_axis_size); - -#ifdef MODE_SCROLL_BEGIN - - ivec3 read_cell = probe_cell - params.scroll; - - uint src_layer = (params.history_size + 1) * params.cascade; - uint dst_layer = (params.history_size + 1) * params.max_cascades; - - for (uint i = 0; i <= params.history_size; i++) { - ivec3 write_pos = ivec3(probe * OCT_RES + local, int(i)); - - if (any(lessThan(read_pos, ivec3(0))) || any(greaterThanEqual(read_pos, ivec3(params.probe_axis_size)))) { - // nowhere to read from for scrolling, try finding the value from upper probes - -#ifdef MODE_IRRADIANCE - imageStore(irradiance_history_texture, write_pos, uvec4(0)); -#endif -#ifdef MODE_DEPTH - imageStore(depth_history_texture, write_pos, uvec4(0)); -#endif - } else { - ivec3 read_pos; - read_pos.xy = read_cell.xy; - read_pos.x += read_cell.z * params.probe_axis_size; - read_pos.xy = read_pos.xy * OCT_RES + local; - read_pos.z = int(i); - -#ifdef MODE_IRRADIANCE - uvec4 value = imageLoad(irradiance_history_texture, read_pos); - imageStore(irradiance_history_texture, write_pos, value); -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - imageStore(depth_history_texture, write_pos, value); -#endif - } - } - -#endif // MODE_SCROLL_BEGIN - -#ifdef MODE_SCROLL_END - - uint src_layer = (params.history_size + 1) * params.max_cascades; - uint dst_layer = (params.history_size + 1) * params.cascade; - - for (uint i = 0; i <= params.history_size; i++) { - ivec3 pos = ivec3(probe * OCT_RES + local, int(i)); - -#ifdef MODE_IRRADIANCE - uvec4 value = imageLoad(irradiance_history_texture, read_pos); - imageStore(irradiance_history_texture, write_pos, value); -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - imageStore(depth_history_texture, write_pos, value); -#endif - } - -#endif //MODE_SCROLL_END - -#ifdef MODE_STORE - - uint src_layer = (params.history_size + 1) * params.cascade + params.history_size; - ivec3 read_pos = ivec3(probe * OCT_RES + local, int(src_layer)); - - ivec3 write_pos = ivec3(probe * (OCT_RES + 2) + ivec2(1), int(params.cascade)); - - ivec3 copy_to[4] = ivec3[](write_pos, ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2)); - -#ifdef MODE_IRRADIANCE - uvec4 average = imageLoad(irradiance_history_texture, read_pos); - vec4 light_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS); - -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - vec2 depth_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS); - - float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; - float max_depth = length(params.grid_size / cascades.data[params.max_cascades - 1].to_cell); - max_depth /= probe_cell_size; - - depth_value = (vec2(average / params.history_size) / float(1 << DEPTH_HISTORY_BITS)) * vec2(max_depth, max_depth * max_depth); - -#endif - - /* Fill the border if required */ - - if (local == ivec2(0, 0)) { - copy_to[1] = texture_pos + ivec3(OCT_RES - 1, -1, 0); - copy_to[2] = texture_pos + ivec3(-1, OCT_RES - 1, 0); - copy_to[3] = texture_pos + ivec3(OCT_RES, OCT_RES, 0); - } else if (local == ivec2(OCT_RES - 1, 0)) { - copy_to[1] = texture_pos + ivec3(0, -1, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES, OCT_RES - 1, 0); - copy_to[3] = texture_pos + ivec3(-1, OCT_RES, 0); - } else if (local == ivec2(0, OCT_RES - 1)) { - copy_to[1] = texture_pos + ivec3(-1, 0, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES - 1, OCT_RES, 0); - copy_to[3] = texture_pos + ivec3(OCT_RES, -1, 0); - } else if (local == ivec2(OCT_RES - 1, OCT_RES - 1)) { - copy_to[1] = texture_pos + ivec3(0, OCT_RES, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES, 0, 0); - copy_to[3] = texture_pos + ivec3(-1, -1, 0); - } else if (local.y == 0) { - copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y - 1, 0); - } else if (local.x == 0) { - copy_to[1] = texture_pos + ivec3(local.x - 1, OCT_RES - local.y - 1, 0); - } else if (local.y == OCT_RES - 1) { - copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y + 1, 0); - } else if (local.x == OCT_RES - 1) { - copy_to[1] = texture_pos + ivec3(local.x + 1, OCT_RES - local.y - 1, 0); - } - - for (int i = 0; i < 4; i++) { - if (copy_to[i] == ivec3(-2, -2, -2)) { - continue; - } -#ifdef MODE_IRRADIANCE - imageStore(irradiance_texture, copy_to[i], light_accum); -#endif -#ifdef MODE_DEPTH - imageStore(depth_texture, copy_to[i], vec4(depth_value, 0.0, 0.0)); -#endif - } - -#endif // MODE_STORE -} diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index 007e4c113a..eedd28959c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -266,9 +266,9 @@ void main() { } else if (params.sky_mode == SKY_MODE_SKY) { #ifdef USE_CUBEMAP_ARRAY - light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #else - light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #endif light.rgb *= params.sky_energy; light.a = 0.0; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl index 916c60ac89..4d9fa85a74 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef MODE_JUMPFLOOD_OPTIMIZED #define GROUP_SIZE 8 @@ -101,7 +101,7 @@ layout(set = 0, binding = 10, std430) restrict buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -134,7 +134,7 @@ layout(set = 0, binding = 5, std430) restrict buffer readonly DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -183,7 +183,7 @@ void main() { ivec3 write_pos = read_pos + params.scroll; if (any(lessThan(write_pos, ivec3(0))) || any(greaterThanEqual(write_pos, ivec3(params.grid_size)))) { - return; //fits outside the 3D texture, dont do anything + return; // Fits outside the 3D texture, don't do anything. } uint albedo = ((src_process_voxels.data[index].albedo & 0x7FFF) << 1) | 1; //add solid bit diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index 680d1045cd..b831005256 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -74,6 +74,53 @@ void main() { #ifdef MODE_2D vec2 vertex = uintBitsToFloat(uvec2(src_vertices.data[src_offset + 0], src_vertices.data[src_offset + 1])); + + if (params.has_blend_shape) { + float blend_total = 0.0; + vec2 blend_vertex = vec2(0.0); + + for (uint i = 0; i < params.blend_shape_count; i++) { + float w = blend_shape_weights.data[i]; + if (abs(w) > 0.0001) { + uint base_offset = (params.vertex_count * i + index) * params.vertex_stride; + + blend_vertex += uintBitsToFloat(uvec2(src_blend_shapes.data[base_offset + 0], src_blend_shapes.data[base_offset + 1])) * w; + + base_offset += 2; + + blend_total += w; + } + } + + if (params.normalized_blend_shapes) { + vertex = (1.0 - blend_total) * vertex; + } + + vertex += blend_vertex; + } + + if (params.has_skeleton) { + uint skin_offset = params.skin_stride * index; + + uvec2 bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset + uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + + skin_offset += params.skin_weight_offset; + + uvec2 weights = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + + vec2 weights_01 = unpackUnorm2x16(weights.x); + vec2 weights_23 = unpackUnorm2x16(weights.y); + + mat4 m = mat4(bone_transforms.data[bones_01.x], bone_transforms.data[bones_01.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; + m += mat4(bone_transforms.data[bones_01.y], bone_transforms.data[bones_01.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; + m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; + m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; + + //reverse order because its transposed + vertex = (vec4(vertex, 0.0, 1.0) * m).xy; + } #else vec3 vertex; vec3 normal; diff --git a/servers/rendering/renderer_rd/shaders/sky.glsl b/servers/rendering/renderer_rd/shaders/sky.glsl index 6c985e1f5c..41c6325bc5 100644 --- a/servers/rendering/renderer_rd/shaders/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/sky.glsl @@ -2,13 +2,19 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif layout(location = 0) out vec2 uv_interp; layout(push_constant, binding = 1, std430) uniform Params { mat3 orientation; - vec4 proj; + vec4 projections[MAX_VIEWS]; vec4 position_multiplier; float time; } @@ -24,17 +30,31 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW #define M_PI 3.14159265359 +#define MAX_VIEWS 2 layout(location = 0) in vec2 uv_interp; layout(push_constant, binding = 1, std430) uniform Params { mat3 orientation; - vec4 proj; + vec4 projections[MAX_VIEWS]; vec4 position_multiplier; - float time; //TODO consider adding vec2 screen res, and float radiance size + float time; } params; @@ -85,16 +105,11 @@ struct DirectionalLightData { layout(set = 0, binding = 3, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } - directional_lights; -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ - -MATERIAL_UNIFORMS - - /* clang-format on */ +#MATERIAL_UNIFORMS } material; #endif @@ -127,11 +142,7 @@ layout(set = 3, binding = 0) uniform texture3D volumetric_fog_texture; #define AT_QUARTER_RES_PASS false #endif -/* clang-format off */ - -FRAGMENT_SHADER_GLOBALS - -/* clang-format on */ +#GLOBALS layout(location = 0) out vec4 frag_color; @@ -162,8 +173,8 @@ vec4 fog_process(vec3 view, vec3 sky_color) { void main() { vec3 cube_normal; cube_normal.z = -1.0; - cube_normal.x = (cube_normal.z * (-uv_interp.x - params.proj.x)) / params.proj.y; - cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.proj.z)) / params.proj.w; + cube_normal.x = (cube_normal.z * (-uv_interp.x - params.projections[ViewIndex].x)) / params.projections[ViewIndex].y; + cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.projections[ViewIndex].z)) / params.projections[ViewIndex].w; cube_normal = mat3(params.orientation) * cube_normal; cube_normal.z = -cube_normal.z; cube_normal = normalize(cube_normal); @@ -202,22 +213,10 @@ void main() { #endif #endif -// unused, just here to make our compiler happy, make sure we don't execute any light code the user adds in.. -#ifndef REALLYINCLUDETHIS - { - /* clang-format off */ - -LIGHT_SHADER_CODE - - /* clang-format on */ - } -#endif { - /* clang-format off */ -FRAGMENT_SHADER_CODE +#CODE : SKY - /* clang-format on */ } frag_color.rgb = color * params.position_multiplier.w; diff --git a/servers/rendering/renderer_rd/shaders/sort.glsl b/servers/rendering/renderer_rd/shaders/sort.glsl index e5ebb9c64b..307e60dc21 100644 --- a/servers/rendering/renderer_rd/shaders/sort.glsl +++ b/servers/rendering/renderer_rd/shaders/sort.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES // Original version here: // https://github.com/GPUOpen-LibrariesAndSDKs/GPUParticles11/blob/master/gpuparticles11/src/Shaders diff --git a/servers/rendering/renderer_rd/shaders/specular_merge.glsl b/servers/rendering/renderer_rd/shaders/specular_merge.glsl index 0b8f406213..3579c35cce 100644 --- a/servers/rendering/renderer_rd/shaders/specular_merge.glsl +++ b/servers/rendering/renderer_rd/shaders/specular_merge.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) out vec2 uv_interp; @@ -17,7 +17,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec2 uv_interp; diff --git a/servers/rendering/renderer_rd/shaders/ssao.glsl b/servers/rendering/renderer_rd/shaders/ssao.glsl index 231f8f91ec..6e945edfcd 100644 --- a/servers/rendering/renderer_rd/shaders/ssao.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao.glsl @@ -21,7 +21,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define SSAO_ADAPTIVE_TAP_BASE_COUNT 5 diff --git a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl index 510a777048..d9cd2b4e85 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl @@ -21,7 +21,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl b/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl index cb2d31f70d..ee0db6a6f0 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl @@ -21,7 +21,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl index 6aa7624261..687fe1e6e2 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl @@ -21,7 +21,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl index 4fdf334aa5..0907423d5d 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl @@ -20,7 +20,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl b/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl index 88a953562f..9367b641c2 100644 --- a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl +++ b/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/tonemap.glsl b/servers/rendering/renderer_rd/shaders/tonemap.glsl index 7de91fd541..3c685c25b9 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/tonemap.glsl @@ -2,7 +2,13 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#endif +#endif layout(location = 0) out vec2 uv_interp; @@ -16,11 +22,28 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW layout(location = 0) in vec2 uv_interp; +#ifdef SUBPASS +layout(input_attachment_index = 0, set = 0, binding = 0) uniform subpassInput input_color; +#else +#if MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif +#endif layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; layout(set = 2, binding = 0) uniform sampler2D source_glow; #ifdef USE_1D_LUT @@ -161,10 +184,6 @@ vec3 tonemap_aces(vec3 color, float white) { } vec3 tonemap_reinhard(vec3 color, float white) { - // Ensure color values are positive. - // They can be negative in the case of negative lights, which leads to undesired behavior. - color = max(vec3(0.0), color); - return (white * color + color) / (color * white + white); } @@ -188,7 +207,7 @@ vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always o return tonemap_reinhard(color, white); } else if (params.tonemapper == TONEMAPPER_FILMIC) { return tonemap_filmic(color, white); - } else { //aces + } else { // TONEMAPPER_ACES return tonemap_aces(color, white); } } @@ -272,15 +291,23 @@ vec3 apply_color_correction(vec3 color) { } #endif +#ifndef SUBPASS vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { const float FXAA_REDUCE_MIN = (1.0 / 128.0); const float FXAA_REDUCE_MUL = (1.0 / 8.0); const float FXAA_SPAN_MAX = 8.0; +#ifdef MULTIVIEW + vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; + vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; + vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; + vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; +#else vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; +#endif vec3 rgbM = color; vec3 luma = vec3(0.299, 0.587, 0.114); float lumaNW = dot(rgbNW, luma); @@ -305,8 +332,13 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { dir * rcpDirMin)) * params.pixel_size; +#ifdef MULTIVIEW + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz); + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz); +#else vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz); vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz); +#endif float lumaB = dot(rgbB, luma); if ((lumaB < lumaMin) || (lumaB > lumaMax)) { @@ -315,6 +347,7 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { return rgbB; } } +#endif // !SUBPASS // From http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf // and https://www.shadertoy.com/view/MslGR8 (5th one starting from the bottom) @@ -329,20 +362,29 @@ vec3 screen_space_dither(vec2 frag_coord) { } void main() { +#ifdef SUBPASS + // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer + vec3 color = subpassLoad(input_color).rgb; +#elif MULTIVIEW + vec3 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f).rgb; +#else vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb; +#endif // Exposure float exposure = params.exposure; +#ifndef SUBPASS if (params.use_auto_exposure) { exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.auto_exposure_grey); } +#endif color *= exposure; // Early Tonemap & SRGB Conversion - +#ifndef SUBPASS if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { vec3 glow = gather_glow(source_glow, uv_interp); color.rgb = mix(color.rgb, glow, params.glow_intensity); @@ -351,15 +393,21 @@ void main() { if (params.use_fxaa) { color = do_fxaa(color, exposure, uv_interp); } +#endif + if (params.use_debanding) { // For best results, debanding should be done before tonemapping. // Otherwise, we're adding noise to an already-quantized image. color += screen_space_dither(gl_FragCoord.xy); } - color = apply_tonemapping(color, params.white); + + // Ensure color values passed to tonemappers are positive. + // They can be negative in the case of negative lights, which leads to undesired behavior. + color = apply_tonemapping(max(vec3(0.0), color), params.white); color = linear_to_srgb(color); // regular linear -> SRGB conversion +#ifndef SUBPASS // Glow if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) { @@ -371,6 +419,7 @@ void main() { color = apply_glow(color, glow); } +#endif // Additional effects diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index ce8a459b24..f2010222e5 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES /* Do not use subgroups here, seems there is not much advantage and causes glitches #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) @@ -26,6 +26,7 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; #endif #include "cluster_data_inc.glsl" +#include "light_data_inc.glsl" #define M_PI 3.14159265359 @@ -71,9 +72,9 @@ layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_ma layout(set = 0, binding = 10) uniform sampler shadow_sampler; -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 -struct GIProbeData { +struct VoxelGIData { mat4 xform; vec3 bounds; float dynamic_range; @@ -89,12 +90,12 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 0, binding = 11, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; +layout(set = 0, binding = 11, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } -gi_probes; +voxel_gi_instances; -layout(set = 0, binding = 12) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +layout(set = 0, binding = 12) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; layout(set = 0, binding = 13) uniform sampler linear_sampler_with_mipmaps; @@ -103,7 +104,7 @@ layout(set = 0, binding = 13) uniform sampler linear_sampler_with_mipmaps; // SDFGI Integration on set 1 #define SDFGI_MAX_CASCADES 8 -struct SDFGIProbeCascadeData { +struct SDFVoxelGICascadeData { vec3 position; float to_probe; ivec3 probe_world_offset; @@ -134,7 +135,7 @@ layout(set = 1, binding = 0, std140) uniform SDFGI { vec3 cascade_probe_size; uint pad5; - SDFGIProbeCascadeData cascades[SDFGI_MAX_CASCADES]; + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; } sdfgi; @@ -161,7 +162,7 @@ layout(set = 0, binding = 14, std140) uniform Params { float detail_spread; float gi_inject; - uint max_gi_probes; + uint max_voxel_gi_instances; uint cluster_type_size; vec2 screen_size; @@ -532,21 +533,21 @@ void main() { vec3 world_pos = mat3(params.cam_rotation) * view_pos; - for (uint i = 0; i < params.max_gi_probes; i++) { - vec3 position = (gi_probes.data[i].xform * vec4(world_pos, 1.0)).xyz; + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + vec3 position = (voxel_gi_instances.data[i].xform * vec4(world_pos, 1.0)).xyz; //this causes corrupted pixels, i have no idea why.. - if (all(bvec2(all(greaterThanEqual(position, vec3(0.0))), all(lessThan(position, gi_probes.data[i].bounds))))) { - position /= gi_probes.data[i].bounds; + if (all(bvec2(all(greaterThanEqual(position, vec3(0.0))), all(lessThan(position, voxel_gi_instances.data[i].bounds))))) { + position /= voxel_gi_instances.data[i].bounds; vec4 light = vec4(0.0); - for (uint j = 0; j < gi_probes.data[i].mipmaps; j++) { - vec4 slight = textureLod(sampler3D(gi_probe_textures[i], linear_sampler_with_mipmaps), position, float(j)); + for (uint j = 0; j < voxel_gi_instances.data[i].mipmaps; j++) { + vec4 slight = textureLod(sampler3D(voxel_gi_textures[i], linear_sampler_with_mipmaps), position, float(j)); float a = (1.0 - light.a); light += a * slight; } - light.rgb *= gi_probes.data[i].dynamic_range * params.gi_inject; + light.rgb *= voxel_gi_instances.data[i].dynamic_range * params.gi_inject; total_light += light.rgb; } diff --git a/servers/rendering/renderer_rd/shaders/giprobe.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl index b931461b31..49a493cdc7 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef MODE_DYNAMIC layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; diff --git a/servers/rendering/renderer_rd/shaders/giprobe_debug.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl index 515cc35507..7d4d72967a 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES struct CellData { uint position; // xyz 10 bits @@ -172,7 +172,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec4 color_interp; layout(location = 0) out vec4 frag_color; diff --git a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl index 5b3dec0ee7..e20b3f680d 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; |