diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
47 files changed, 3822 insertions, 1202 deletions
diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h index ebb81abdad..c0c03eb26a 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.h +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -235,7 +235,7 @@ public: 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(); } diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index 5cf8895c8e..c609f25e31 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -37,6 +37,10 @@ #include "servers/rendering/renderer_rd/renderer_compositor_rd.h" #include "thirdparty/misc/cubemap_coeffs.h" +bool EffectsRD::get_prefer_raster_effects() { + return prefer_raster_effects; +} + static _FORCE_INLINE_ void store_camera(const CameraMatrix &p_mtx, float *p_array) { for (int i = 0; i < 4; i++) { for (int j = 0; j < 4; j++) { @@ -66,6 +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]; @@ -81,7 +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; @@ -383,6 +409,8 @@ 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) { + 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; @@ -416,6 +444,8 @@ 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) { + 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; @@ -449,6 +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(); @@ -719,11 +800,11 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone if (p_settings.view_count > 1) { // Use MULTIVIEW versions - mode += 4; + 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); @@ -735,7 +816,48 @@ 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; @@ -774,7 +896,41 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_end(); } +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(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) { + 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; @@ -924,6 +1080,78 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_end(); } +void EffectsRD::blur_dof_raster(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_base_fb, RID p_secondary_texture, RID p_secondary_fb, 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, 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(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode blur_mode; + int qsteps[4] = { 4, 4, 10, 20 }; + uint32_t base_flags = p_cam_orthogonal ? BLUR_FLAG_USE_ORTHOGONAL_PROJECTION : 0; + + Vector2 pixel_size = Vector2(1.0 / p_base_texture_size.width, 1.0 / p_base_texture_size.height); + + blur_raster.push_constant.dof_radius = (p_dof_blur_amount * p_dof_blur_amount) / qsteps[p_quality]; + blur_raster.push_constant.pixel_size[0] = pixel_size.x; + blur_raster.push_constant.pixel_size[1] = pixel_size.y; + blur_raster.push_constant.camera_z_far = p_cam_zfar; + blur_raster.push_constant.camera_z_near = p_cam_znear; + + if (p_dof_far || p_dof_near) { + if (p_quality == RS::DOF_BLUR_QUALITY_HIGH) { + blur_mode = BLUR_MODE_DOF_HIGH; + } else if (p_quality == RS::DOF_BLUR_QUALITY_MEDIUM) { + blur_mode = BLUR_MODE_DOF_MEDIUM; + } else { // for LOW or VERYLOW we use LOW + blur_mode = BLUR_MODE_DOF_LOW; + } + + if (p_dof_far) { + base_flags |= BLUR_FLAG_DOF_FAR; + blur_raster.push_constant.dof_far_begin = p_dof_far_begin; + blur_raster.push_constant.dof_far_end = p_dof_far_begin + p_dof_far_size; + } + + if (p_dof_near) { + base_flags |= BLUR_FLAG_DOF_NEAR; + blur_raster.push_constant.dof_near_begin = p_dof_near_begin; + blur_raster.push_constant.dof_near_end = p_dof_near_begin - p_dof_near_size; + } + + //HORIZONTAL + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_secondary_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].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_secondary_fb))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_depth_texture), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags | BLUR_FLAG_HORIZONTAL; + blur_raster.push_constant.dof_dir[0] = 1.0; + blur_raster.push_constant.dof_dir[1] = 0.0; + + 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(); + + //VERTICAL + draw_list = RD::get_singleton()->draw_list_begin(p_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].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_base_fb))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_secondary_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_depth_texture), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags; + blur_raster.push_constant.dof_dir[0] = 0.0; + blur_raster.push_constant.dof_dir[1] = 1.0; + + 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(); + } +} + 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) { @@ -1188,8 +1416,9 @@ 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++) { @@ -1283,7 +1512,9 @@ 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) { +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; @@ -1293,10 +1524,10 @@ 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)); @@ -1308,11 +1539,37 @@ 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); @@ -1326,7 +1583,27 @@ 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; @@ -1338,12 +1615,12 @@ 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); @@ -1355,6 +1632,29 @@ void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, RD::get_singleton()->compute_list_end(); } +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."); + + // TODO implement! + CubemapFilterRasterPushConstant push_constant; + push_constant.mip_level = p_mip_level; + push_constant.face_id = p_face_id; + + 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, &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_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; @@ -1464,7 +1764,35 @@ void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { RD::get_singleton()->compute_list_end(); } -EffectsRD::EffectsRD() { +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_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_DOF_BLUR\n#define DOF_QUALITY_LOW\n"); // BLUR_MODE_DOF_LOW + blur_modes.push_back("\n#define MODE_DOF_BLUR\n#define DOF_QUALITY_MEDIUM\n"); // BLUR_MODE_DOF_MEDIUM + blur_modes.push_back("\n#define MODE_DOF_BLUR\n#define DOF_QUALITY_HIGH\n"); // BLUR_MODE_DOF_HIGH + + 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(); + } + } + { // Initialize copy Vector<String> copy_modes; copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); @@ -1483,10 +1811,21 @@ EffectsRD::EffectsRD() { copy.shader.initialize(copy_modes); 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)); + } } } { @@ -1510,11 +1849,22 @@ 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); + + } else { + roughness.compute_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.compute_shader.version_create(); - roughness.pipeline = RD::get_singleton()->compute_pipeline_create(roughness.shader.version_get_shader(roughness.shader_version, 0)); + roughness.compute_pipeline = RD::get_singleton()->compute_pipeline_create(roughness.compute_shader.version_get_shader(roughness.shader_version, 0)); + roughness.raster_pipeline.clear(); + } } { @@ -1524,12 +1874,16 @@ 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); @@ -1538,6 +1892,8 @@ EffectsRD::EffectsRD() { 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(); @@ -1551,7 +1907,20 @@ EffectsRD::EffectsRD() { } } - { + 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"); @@ -1565,6 +1934,10 @@ 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(); + } } { @@ -1583,7 +1956,9 @@ EffectsRD::EffectsRD() { cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); } - { + if (prefer_raster_effects) { + // not supported + } else { // Initialize bokeh Vector<String> bokeh_modes; bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); @@ -1761,11 +2136,21 @@ 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(); + } } { @@ -1777,12 +2162,6 @@ 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)); @@ -1792,15 +2171,50 @@ 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); + filter.shader_version = filter.raster_shader.version_create(); + + // 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); + + 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); } { @@ -1974,15 +2388,23 @@ EffectsRD::~EffectsRD() { RD::get_singleton()->free(ssao.gather_constants_buffer); RD::get_singleton()->free(ssao.importance_map_load_counter); - bokeh.shader.version_free(bokeh.shader_version); + if (prefer_raster_effects) { + blur_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.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); + } copy.shader.version_free(copy.shader_version); 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); diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index 33d32f0c57..79d9a8b738 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -33,14 +33,19 @@ #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/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" @@ -60,6 +65,66 @@ #include "servers/rendering_server.h" class EffectsRD { +private: + bool prefer_raster_effects; + + enum BlurRasterMode { + BLUR_MODE_GAUSSIAN_BLUR, + BLUR_MODE_GAUSSIAN_GLOW, + BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, + + BLUR_MODE_DOF_LOW, + BLUR_MODE_DOF_MEDIUM, + BLUR_MODE_DOF_HIGH, + + BLUR_MODE_MAX + }; + + enum { + BLUR_FLAG_HORIZONTAL = (1 << 0), + BLUR_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 1), + BLUR_FLAG_GLOW_FIRST_PASS = (1 << 2), + BLUR_FLAG_DOF_FAR = (1 << 3), + BLUR_FLAG_DOF_NEAR = (1 << 4), + }; + + 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; + + //dof + float dof_far_begin; + float dof_far_end; + float dof_near_begin; + float dof_near_end; + + float dof_radius; + float dof_pad[3]; + + float dof_dir[2]; + float camera_z_far; + float camera_z_near; + }; + + 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, @@ -161,9 +226,11 @@ 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 { @@ -171,11 +238,15 @@ 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 }; @@ -239,6 +310,29 @@ 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; @@ -426,15 +520,17 @@ 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 { @@ -445,10 +541,19 @@ 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; @@ -617,6 +722,7 @@ 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; @@ -650,6 +756,7 @@ 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); @@ -657,6 +764,8 @@ 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); @@ -666,12 +775,17 @@ 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 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 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); + 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 blur_dof_raster(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_base_fb, RID p_secondary_texture, RID p_secondary_fb, 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, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); struct TonemapSettings { bool use_glow = false; @@ -734,13 +848,16 @@ 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, 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 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); @@ -751,7 +868,7 @@ public: 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 22bfd03115..0d304556c2 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -183,7 +183,7 @@ 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, uint32_t p_view_count) { +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"); @@ -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 { @@ -339,7 +339,7 @@ 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; @@ -369,7 +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; @@ -384,14 +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; @@ -452,7 +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 @@ -467,8 +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); } @@ -573,11 +589,6 @@ void RenderForwardClustered::_setup_environment(const RenderDataRD *p_render_dat 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; @@ -848,7 +859,7 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, i 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 { @@ -868,6 +879,8 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, i 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; @@ -1375,7 +1388,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, nullptr, RID()); 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, 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); + 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(); @@ -1432,7 +1445,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } 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(), reverse_cull, 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(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->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 @@ -1529,7 +1542,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _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(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->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); } @@ -1542,6 +1555,14 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } 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() { @@ -1631,7 +1652,7 @@ 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); } @@ -1672,7 +1693,7 @@ void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, con { //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(); @@ -1707,7 +1728,7 @@ void RenderForwardClustered::_render_material(const Transform3D &p_cam_transform 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)); @@ -1750,7 +1771,7 @@ void RenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p 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)); @@ -1868,7 +1889,7 @@ 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); } @@ -1924,13 +1945,67 @@ 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); @@ -1938,35 +2013,35 @@ 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); @@ -1974,7 +2049,7 @@ 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); @@ -1982,7 +2057,7 @@ 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); @@ -1991,7 +2066,7 @@ 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); } @@ -1999,7 +2074,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { 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); } @@ -2511,12 +2586,14 @@ 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) { @@ -2911,6 +2988,56 @@ void RenderForwardClustered::geometry_instance_pair_voxel_gi_instances(GeometryI } } +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; @@ -2948,6 +3075,8 @@ RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : } 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 750c0167e7..676f633d33 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -51,6 +51,15 @@ 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_VOXEL_GI_INSTANCESS = 8, MAX_LIGHTMAPS = 8, @@ -103,7 +112,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { void ensure_specular(); 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, uint32_t p_view_count); + 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(); }; @@ -156,8 +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; @@ -172,6 +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; } }; @@ -220,11 +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; @@ -353,7 +359,10 @@ class RenderForwardClustered : public RendererSceneRenderRD { 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; @@ -402,12 +411,14 @@ class RenderForwardClustered : public RendererSceneRenderRD { 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; @@ -455,6 +466,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { 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; Transform3D transform; @@ -564,6 +577,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { RenderList render_list[RENDER_LIST_MAX]; + virtual void _update_shader_quality_settings() override; + protected: virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; @@ -604,6 +619,8 @@ public: 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); 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 d39823a1a3..89d14e871a 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 @@ -131,8 +131,9 @@ 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); + if (err != OK) { + return; + } if (version.is_null()) { version = shader_singleton->shader.version_create(); @@ -287,7 +288,7 @@ 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 @@ -305,7 +306,7 @@ 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 @@ -324,7 +325,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { } 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); } } } @@ -408,7 +409,8 @@ 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; } @@ -424,6 +426,7 @@ 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; } @@ -483,7 +486,6 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin 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 USE_FORWARD_GI\n"); // SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI 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 @@ -682,7 +684,19 @@ 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); @@ -699,7 +713,16 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin 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, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.1; }"); + 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); @@ -729,3 +752,16 @@ 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 810b1f3876..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 @@ -52,7 +52,6 @@ public: 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, @@ -60,6 +59,13 @@ public: 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, @@ -154,10 +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(); @@ -203,10 +212,12 @@ public: 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); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); }; } // namespace RendererSceneRenderImplementation diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp index 7fbd6e23b0..f810ae2454 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -35,6 +35,34 @@ 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() { @@ -50,12 +78,16 @@ void RenderForwardMobile::RenderBufferDataForwardMobile::clear() { color = RID(); depth = RID(); - color_fb = 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, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { +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; @@ -65,22 +97,57 @@ void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_b color = p_color_buffer; depth = p_depth_buffer; - // re-introduce setting up msaa? For now we ignore this... + // 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); - fb.push_back(depth); + 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); - color_fb = RD::get_singleton()->framebuffer_create(fb, 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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = color_format; tf.width = p_width; tf.height = p_height; tf.array_layers = view_count; // create a layer for every view @@ -106,14 +173,87 @@ void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_b { Vector<RID> fb; - fb.push_back(color_msaa); - fb.push_back(depth_msaa); + 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); + } - color_fb = RD::get_singleton()->framebuffer_create(fb, 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(); } @@ -131,6 +271,17 @@ bool RenderForwardMobile::free(RID p_rid) { /* 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()); @@ -326,14 +477,29 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color 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 opaque_framebuffer; - RID alpha_framebuffer; + 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 - // I don't think we support either of these in our mobile renderer so probably should phase them out - bool using_ssr = false; - bool using_sss = false; + // 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(); @@ -345,15 +511,36 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color screen_size.x = render_buffer->width; screen_size.y = render_buffer->height; - opaque_framebuffer = render_buffer->color_fb; - alpha_framebuffer = opaque_framebuffer; + 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; - opaque_framebuffer = reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); - alpha_framebuffer = opaque_framebuffer; + 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 @@ -361,6 +548,8 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color } 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? } @@ -372,17 +561,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color _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_render_data, PASS_MODE_COLOR); - render_list[RENDER_LIST_OPAQUE].sort_by_key(); - render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); - - // we no longer use this... - _fill_instance_data(RENDER_LIST_OPAQUE); - _fill_instance_data(RENDER_LIST_ALPHA); - - RD::get_singleton()->draw_command_end_label(); - - // note, no depth prepass here! + RD::get_singleton()->draw_command_end_label(); // Render Setup // setup environment RID radiance_texture; @@ -457,97 +636,201 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color // do not try to draw sky if invalid draw_sky = false; } - RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_command_end_label(); // Setup Sky } } else { clear_color = p_default_bg_color; } - // opaque pass + // 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 + } - // !BAS! Look into this, seems most of the code in here related to clustered only, may want to move this code into ForwardClustered/RenderForwardMobile before calling it from here - // does trigger shadow map rendering so kinda important _pre_opaque_render(p_render_data, false, false, RID(), RID()); - RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); + { + 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"); + } - scene_state.ubo.directional_light_count = p_render_data->directional_light_count; + // opaque pass - _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()); + RD::get_singleton()->draw_command_begin_label("Render Opaque Subpass"); - RENDER_TIMESTAMP("Render Opaque Pass"); + scene_state.ubo.directional_light_count = p_render_data->directional_light_count; - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + _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()); - 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; + 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"); + } - { - 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); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); - // regular forward for now - Vector<Color> c; - c.push_back(clear_color.to_linear()); + 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; - 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, 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_with_threads(&render_list_params, opaque_framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); - } + { + // 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::get_singleton()->draw_command_end_label(); + 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, 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); + } + } - if (draw_sky || draw_sky_fog_only) { - RENDER_TIMESTAMP("Render Sky"); + RD::get_singleton()->draw_command_end_label(); //Render Opaque Subpass - RD::get_singleton()->draw_command_begin_label("Draw Sky"); + if (draw_sky || draw_sky_fog_only) { + RD::get_singleton()->draw_command_begin_label("Draw Sky Subpass"); - 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(env, can_continue_color, can_continue_depth, opaque_framebuffer, 1, &projection, p_render_data->cam_transform, time); + 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 { - sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time); + // 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(); } - RD::get_singleton()->draw_command_end_label(); - } - if (render_buffer && !can_continue_color && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); - /* - if (using_separate_specular) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular); + 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 (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); - } + // transparent pass + + RD::get_singleton()->draw_command_begin_label("Render Transparent Subpass"); - // transparent pass - RENDER_TIMESTAMP("Render Transparent Pass"); + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); - RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); + 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, 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); + } - rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass - _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); + // note if we are using MSAA we should get an automatic resolve through our subpass configuration. - { - 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, 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_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); + // 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, 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 + } } - RD::get_singleton()->draw_command_end_label(); + if (render_buffer && !using_subpass_post_process) { + RD::get_singleton()->draw_command_begin_label("Post process pass"); - RD::get_singleton()->draw_command_begin_label("Resolve"); + // If we need extra effects we do this in its own pass + RENDER_TIMESTAMP("Tonemap"); - if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); + _render_buffers_post_process_and_tonemap(p_render_data); + + RD::get_singleton()->draw_command_end_label(); // Post process pass } - RD::get_singleton()->draw_command_end_label(); + if (render_buffer) { + _disable_clear_request(p_render_data); + } } /* these are being called from RendererSceneRenderRD::_pre_opaque_render */ @@ -596,7 +879,7 @@ void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedAr _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_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); + _fill_element_info(RENDER_LIST_SECONDARY, render_list_from, render_list_size); { //regular forward for now @@ -674,7 +957,7 @@ void RenderForwardMobile::_render_material(const Transform3D &p_cam_transform, c 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_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); @@ -715,7 +998,7 @@ void RenderForwardMobile::_render_uv2(const PagedArray<GeometryInstance *> &p_in 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_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); @@ -788,7 +1071,7 @@ void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); - _fill_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); @@ -853,13 +1136,67 @@ void RenderForwardMobile::_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); @@ -867,35 +1204,35 @@ void RenderForwardMobile::_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); @@ -903,7 +1240,7 @@ void RenderForwardMobile::_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); @@ -911,7 +1248,7 @@ void RenderForwardMobile::_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); @@ -920,7 +1257,7 @@ void RenderForwardMobile::_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); } @@ -1157,11 +1494,6 @@ void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, 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; @@ -1325,9 +1657,7 @@ void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, 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_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { - // !BAS! Rename this to make clear this is not the same as with the forward renderer and remove p_update_buffer? - +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(); @@ -1392,6 +1722,44 @@ void RenderForwardMobile::_render_list_with_threads(RenderListParameters *p_para } } +void RenderForwardMobile::_fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, 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; + + 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; @@ -1441,8 +1809,6 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr push_constant.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; }; - _fill_instance_indices(inst->omni_lights, inst->omni_light_count, push_constant.omni_lights, inst->spot_lights, inst->spot_light_count, push_constant.spot_lights, inst->reflection_probes, inst->reflection_probe_count, push_constant.reflection_probes, inst->decals, inst->decals_count, push_constant.decals, push_constant.layer_mask); - RID material_uniform_set; SceneShaderForwardMobile::ShaderData *shader; void *mesh_surface; @@ -1453,6 +1819,8 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr mesh_surface = surf->surface_shadow; } else { + _fill_push_constant_instance_indices(&push_constant, inst); + #ifdef DEBUG_ENABLED if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { material_uniform_set = scene_shader.default_material_uniform_set; @@ -1543,7 +1911,7 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr 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, p_params->subpass); if (pipeline_rd != prev_pipeline_rd) { // checking with prev shader does not make so much sense, as @@ -1558,8 +1926,8 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr } 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); } @@ -1761,13 +2129,13 @@ void RenderForwardMobile::geometry_instance_pair_light_instances(GeometryInstanc switch (type) { case RS::LIGHT_OMNI: { if (ginstance->omni_light_count < (uint32_t)MAX_RDL_CULL) { - ginstance->omni_lights[ginstance->omni_light_count] = p_light_instances[i]; + 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] = p_light_instances[i]; + ginstance->spot_lights[ginstance->spot_light_count] = light_instance_get_forward_id(p_light_instances[i]); ginstance->spot_light_count++; } } break; @@ -1783,7 +2151,7 @@ void RenderForwardMobile::geometry_instance_pair_reflection_probe_instances(Geom 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] = p_reflection_probe_instances[i]; + ginstance->reflection_probes[i] = reflection_probe_instance_get_forward_id(p_reflection_probe_instances[i]); } } @@ -1793,7 +2161,7 @@ void RenderForwardMobile::geometry_instance_pair_decal_instances(GeometryInstanc 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] = p_decal_instances[i]; + ginstance->decals[i] = decal_instance_get_forward_id(p_decal_instances[i]); } } @@ -1801,6 +2169,9 @@ void RenderForwardMobile::geometry_instance_pair_voxel_gi_instances(GeometryInst // We do not have this here! } +void RenderForwardMobile::geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) { +} + 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()) { @@ -2150,10 +2521,54 @@ uint32_t RenderForwardMobile::get_max_elements() const { 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_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 +} + 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"; @@ -2185,6 +2600,8 @@ RenderForwardMobile::RenderForwardMobile(RendererStorageRD *p_storage) : // !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() { diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h index f40f713c03..b981592840 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h @@ -42,6 +42,18 @@ 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 */ @@ -53,6 +65,15 @@ protected: }; 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 { 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 @@ -71,6 +92,18 @@ protected: /* 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; @@ -83,12 +116,12 @@ protected: RID depth_msaa; // RID normal_roughness_buffer_msaa; - RID color_fb; + 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, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); + 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(); }; @@ -131,6 +164,7 @@ protected: 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, 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; @@ -151,6 +185,9 @@ protected: } }; + 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; @@ -171,7 +208,7 @@ protected: 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_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + 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; @@ -213,11 +250,6 @@ protected: 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; @@ -513,14 +545,14 @@ protected: GeometryInstanceLightmapSH *lightmap_sh = nullptr; // culled light info - uint32_t reflection_probe_count; - RID reflection_probes[MAX_RDL_CULL]; - uint32_t omni_light_count; - RID omni_lights[MAX_RDL_CULL]; - uint32_t spot_light_count; - RID spot_lights[MAX_RDL_CULL]; - uint32_t decals_count; - RID decals[MAX_RDL_CULL]; + 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; @@ -552,7 +584,13 @@ protected: dirty_list_element(this) {} }; + _FORCE_INLINE_ void _fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, 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); @@ -594,6 +632,8 @@ public: 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; 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 index 7709c8aadc..bcdcb05653 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -318,7 +318,7 @@ void SceneShaderForwardMobile::ShaderData::set_code(const String &p_code) { } 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); } } } @@ -402,7 +402,8 @@ RS::ShaderNativeSourceCode SceneShaderForwardMobile::ShaderData::get_native_sour return shader_singleton->shader.version_get_native_source_code(version); } -SceneShaderForwardMobile::ShaderData::ShaderData() { +SceneShaderForwardMobile::ShaderData::ShaderData() : + shader_list_element(this) { valid = false; uses_screen_texture = false; } @@ -418,6 +419,7 @@ SceneShaderForwardMobile::ShaderData::~ShaderData() { RendererStorageRD::ShaderData *SceneShaderForwardMobile::_create_shader_func() { ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); return shader_data; } @@ -671,7 +673,19 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p //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); @@ -687,7 +701,16 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p 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, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.1; }"); + 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); @@ -718,6 +741,19 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p } } +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); 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 index 5c9e35fd0d..e1c10f0206 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h @@ -151,6 +151,8 @@ 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(); }; @@ -174,6 +176,8 @@ public: 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)); @@ -202,7 +206,10 @@ public: 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 diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp index 22888ddbe5..aefe926cb0 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp @@ -31,14 +31,30 @@ #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, uint32_t p_render_pass) { +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, 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, p_render_pass); + 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; @@ -46,11 +62,15 @@ RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD 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 @@ -64,7 +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; @@ -75,6 +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 387a8a038f..e52f47fa47 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.h +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.h @@ -46,27 +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, uint32_t p_render_pass); + 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, uint32_t p_render_pass = 0) { + _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),"); @@ -75,13 +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 && versions[i].render_pass == p_render_pass) { + 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, p_render_pass); + 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 1e3dbe69a3..14b2768c29 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -480,6 +480,10 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend 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); @@ -620,7 +624,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend 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; @@ -1089,7 +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 { @@ -2570,8 +2575,19 @@ 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_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index a8f086b0f9..02d548bf13 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -173,7 +173,6 @@ void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color } Size2 window_size = DisplayServer::get_singleton()->window_get_size(); - print_line("window size: " + window_size); Rect2 imgrect(0, 0, p_image->get_width(), p_image->get_height()); Rect2 screenrect; @@ -216,8 +215,6 @@ void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color blit.push_constant.upscale = 1.0; blit.push_constant.aspect_ratio = 1.0; - print_line("rect: " + screenrect); - RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); RD::get_singleton()->draw_list_draw(draw_list, true); @@ -225,7 +222,7 @@ void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color RD::get_singleton()->swap_buffers(); - RD::get_singleton()->free(texture); + storage->free(texture); } RendererCompositorRD *RendererCompositorRD::singleton = nullptr; @@ -283,6 +280,9 @@ RendererCompositorRD::RendererCompositorRD() { // default to our high end renderer scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered(storage)); } + + // now we're ready to create our effects, + storage->init_effects(!scene->_render_buffers_can_be_storage()); } RendererCompositorRD::~RendererCompositorRD() { diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index 15b3b77ed9..8639362da9 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -80,7 +80,7 @@ protected: Map<RID, RID> render_target_descriptors; double time; - float delta; + double delta; static uint64_t frame; @@ -100,7 +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() { diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 46057bddab..fd797b1cab 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -567,6 +567,8 @@ 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); } @@ -627,6 +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); @@ -654,12 +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()); } @@ -673,12 +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); } } @@ -709,15 +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); @@ -1229,6 +1243,9 @@ 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; } @@ -1302,6 +1319,7 @@ 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); } @@ -1373,12 +1391,20 @@ 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()); @@ -1398,11 +1424,40 @@ 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); } @@ -1425,26 +1480,48 @@ 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(); @@ -1456,19 +1533,43 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { } 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(); @@ -1740,17 +1841,27 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende 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); } - 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_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + if (can_use_storage) { + 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_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + } else { + storage->get_effects()->blur_dof_raster(rb->texture, rb->depth_texture, Size2i(rb->width, rb->height), rb->texture_fb, rb->blur[0].mipmaps[0].texture, rb->blur[0].mipmaps[0].fb, 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, camfx->dof_blur_amount, 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); } @@ -1759,16 +1870,26 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende 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()) { @@ -1794,14 +1915,26 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const 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; @@ -1860,11 +1993,81 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const 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::_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(); @@ -2114,8 +2317,16 @@ float RendererSceneRenderRD::render_buffers_get_volumetric_fog_detail_spread(RID return rb->volumetric_fog->spread; } +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 atleast 1 view"); + 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; @@ -2140,16 +2351,15 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p if (rb->view_count > 1) { tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; } - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); tf.width = rb->width; tf.height = rb->height; tf.array_layers = rb->view_count; // create a layer for every view - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + 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()); } @@ -2179,7 +2389,16 @@ 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, p_view_count); + 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); @@ -2241,6 +2460,8 @@ 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) { @@ -2281,6 +2502,23 @@ 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 { @@ -2320,10 +2558,13 @@ 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; @@ -2373,7 +2614,7 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti } } -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) { +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; @@ -2385,6 +2626,8 @@ 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) { @@ -2423,6 +2666,9 @@ 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; } @@ -2615,6 +2861,8 @@ 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]; @@ -2622,6 +2870,10 @@ 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; + 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; @@ -2754,7 +3006,6 @@ 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); if (current_cluster_builder != nullptr) { @@ -2823,11 +3074,15 @@ 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; - di->render_index = i; + 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; @@ -2944,116 +3199,6 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const } } -void RendererSceneRenderRD::_fill_instance_indices(const RID *p_omni_light_instances, uint32_t p_omni_light_instance_count, uint32_t *p_omni_light_indices, const RID *p_spot_light_instances, uint32_t p_spot_light_instance_count, uint32_t *p_spot_light_indices, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count, uint32_t *p_reflection_probe_indices, const RID *p_decal_instances, uint32_t p_decal_instance_count, uint32_t *p_decal_instance_indices, uint32_t p_layer_mask, uint32_t p_max_dst_words) { - // first zero out our indices - for (uint32_t i = 0; i < p_max_dst_words; i++) { - p_omni_light_indices[i] = 0; - p_spot_light_indices[i] = 0; - p_reflection_probe_indices[i] = 0; - p_decal_instance_indices[i] = 0; - } - - { - // process omni lights - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_omni_light_instance_count && dword < p_max_dst_words; i++) { - LightInstance *li = light_instance_owner.getornull(p_omni_light_instances[i]); - - if ((li->cull_mask & p_layer_mask) && (li->light_index < 255)) { - p_omni_light_indices[dword] += li->light_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_omni_light_indices[dword] += 0xFF << shift; - } - } - - { - // process spot lights - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_spot_light_instance_count && dword < p_max_dst_words; i++) { - LightInstance *li = light_instance_owner.getornull(p_spot_light_instances[i]); - - if ((li->cull_mask & p_layer_mask) && (li->light_index < 255)) { - p_spot_light_indices[dword] += li->light_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_spot_light_indices[dword] += 0xFF << shift; - } - } - - { - // process reflection probes - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_reflection_probe_instance_count && dword < p_max_dst_words; i++) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflection_probe_instances[i]); - - if ((rpi->cull_mask & p_layer_mask) && (rpi->render_index < 255)) { - p_reflection_probe_indices[dword] += rpi->render_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_reflection_probe_indices[dword] += 0xFF << shift; - } - } - - { - // process decals - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_decal_instance_count && dword < p_max_dst_words; i++) { - DecalInstance *decal = decal_instance_owner.getornull(p_decal_instances[i]); - - if ((decal->cull_mask & p_layer_mask) && (decal->render_index < 255)) { - p_decal_instance_indices[dword] += decal->render_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_decal_instance_indices[dword] += 0xFF << shift; - } - } -} - void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { ERR_FAIL_COND(!rb->volumetric_fog); @@ -3617,7 +3762,7 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*p_render_data->lights, p_render_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count); + _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()); p_render_data->directional_light_count = directional_light_count; @@ -3773,9 +3918,28 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData _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; @@ -3791,17 +3955,7 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData default: { } } - if (current_cluster_builder != nullptr) { - current_cluster_builder->debug(elem_type); - } - } - - 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); + current_cluster_builder->debug(elem_type); } } } @@ -4029,11 +4183,13 @@ 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); @@ -4068,6 +4224,9 @@ bool RendererSceneRenderRD::free(RID p_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)) { @@ -4357,6 +4516,9 @@ 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); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index be3d3551c7..336dd6e1a7 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -80,6 +80,7 @@ struct RenderDataRD { uint32_t cluster_max_elements = 0; uint32_t directional_light_count = 0; + bool directional_light_soft_shadows = false; RendererScene::RenderInfo *render_info = nullptr; }; @@ -94,12 +95,12 @@ protected: double time_step = 0; struct RenderBufferData { - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) = 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 Transform3D &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_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); @@ -116,6 +117,7 @@ protected: 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); @@ -132,6 +134,10 @@ protected: 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 _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; PagedArray<GeometryInstance *> cull_argument; //need this to exist @@ -145,7 +151,26 @@ 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; @@ -188,9 +213,10 @@ 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; + Transform3D transform; }; @@ -201,8 +227,8 @@ private: struct DecalInstance { RID decal; Transform3D transform; - uint32_t render_index; uint32_t cull_mask; + ForwardID forward_id = -1; }; mutable RID_Owner<DecalInstance> decal_instance_owner; @@ -286,6 +312,8 @@ 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 */ @@ -346,7 +374,6 @@ 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; @@ -358,6 +385,8 @@ private: Set<RID> shadow_atlases; //shadow atlases where this light is registered + ForwardID forward_id = -1; + LightInstance() {} }; @@ -426,6 +455,7 @@ 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; @@ -441,6 +471,11 @@ private: RID texture; int width; int height; + + // only used on mobile renderer + RID fb; + RID half_texture; + RID half_fb; }; Vector<Mipmap> mipmaps; @@ -451,6 +486,10 @@ private: struct Luminance { Vector<RID> reduce; RID current; + + // used only on mobile renderer + Vector<RID> fb; + RID current_fb; } luminance; struct SSAO { @@ -490,7 +529,6 @@ private: void _allocate_luminance_textures(RenderBuffers *rb); void _render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer); - void _render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data); /* Cluster */ @@ -889,6 +927,12 @@ public: 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; + bool camera_effects_uses_dof(RID p_camera_effects) { + CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects); + + 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; @@ -1005,14 +1049,9 @@ 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) { @@ -1036,6 +1075,7 @@ public: 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); @@ -1049,17 +1089,11 @@ 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) { @@ -1097,6 +1131,11 @@ public: return decal->decal; } + _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; @@ -1117,8 +1156,6 @@ public: return li->transform; } - void _fill_instance_indices(const RID *p_omni_light_instances, uint32_t p_omni_light_instance_count, uint32_t *p_omni_light_indices, const RID *p_spot_light_instances, uint32_t p_spot_light_instance_count, uint32_t *p_spot_light_indices, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count, uint32_t *p_reflection_probe_indices, const RID *p_decal_instances, uint32_t p_decal_instance_count, uint32_t *p_decal_instance_indices, uint32_t p_layer_mask, uint32_t p_max_dst_words = 2); - /* gi light probes */ virtual RID voxel_gi_instance_create(RID p_base) override; @@ -1129,6 +1166,8 @@ public: /* render buffers */ + 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; @@ -1183,6 +1222,10 @@ public: 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; + _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; } @@ -1198,6 +1241,9 @@ 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; diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp index e701219617..9331cab090 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp @@ -283,14 +283,17 @@ void RendererSceneSkyRD::_render_sky(RD::DrawListID p_list, float p_time, RID p_ RD::DrawListID draw_list = p_list; - 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_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format, false, RD::get_singleton()->draw_list_get_current_pass())); - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.uniform_set, 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); + // Update uniform sets. + { + 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); } - 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); RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); @@ -313,12 +316,16 @@ 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; @@ -377,9 +384,9 @@ 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; @@ -388,6 +395,7 @@ 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; @@ -397,6 +405,18 @@ 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); @@ -405,50 +425,128 @@ 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(); } //////////////////////////////////////////////////////////////////////////////// @@ -757,7 +855,13 @@ 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 sky() { 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); @@ -838,7 +942,15 @@ 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 sky() { 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); @@ -888,6 +1000,10 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { } } +void RendererSceneSkyRD::set_texture_format(RD::DataFormat p_texture_format) { + texture_format = p_texture_format; +} + RendererSceneSkyRD::~RendererSceneSkyRD() { // TODO cleanup anything created in init... @@ -1156,6 +1272,7 @@ 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; @@ -1163,17 +1280,18 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform3D 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); - _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.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; @@ -1181,28 +1299,29 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform3D 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); - _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.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++) { - Transform3D 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); - _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.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); @@ -1348,6 +1467,179 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont 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; @@ -1379,7 +1671,7 @@ 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; @@ -1388,13 +1680,13 @@ 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; @@ -1403,7 +1695,7 @@ 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; } @@ -1411,7 +1703,7 @@ 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; @@ -1426,7 +1718,7 @@ 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 4f852e55a7..7b670bddd5 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h @@ -64,6 +64,7 @@ public: private: RendererStorageRD *storage; + RD::DataFormat texture_format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; RID index_buffer; RID index_array; @@ -190,6 +191,10 @@ public: struct Mipmap { RID view; Size2i size; + + // for mobile only + RID views[6]; + RID framebuffers[6]; }; Vector<Mipmap> mipmaps; }; @@ -204,7 +209,7 @@ 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); @@ -284,11 +289,14 @@ public: RendererSceneSkyRD(); void init(RendererStorageRD *p_storage); + 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 6738f499bd..4feaf08808 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -2531,6 +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 { @@ -2714,9 +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); } @@ -3027,8 +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; @@ -3454,6 +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 } @@ -4092,7 +4092,7 @@ void RendererStorageRD::particles_set_amount(RID p_particles, int p_amount) { 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; @@ -4104,7 +4104,7 @@ 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; @@ -4127,7 +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); @@ -4443,7 +4443,7 @@ void RendererStorageRD::particles_set_canvas_sdf_collision(RID p_particles, bool particles->sdf_collision_texture = p_texture; } -void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta) { +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; @@ -4492,7 +4492,7 @@ 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); //move back history (if there is any) for (uint32_t i = p_particles->frame_history.size() - 1; i > 0; i--) { @@ -4960,7 +4960,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & 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; @@ -5132,14 +5132,14 @@ void RendererStorageRD::update_particles() { bool zero_time_scale = Engine::get_singleton()->get_time_scale() <= 0.0; if (particles->clear && particles->pre_process_time > 0.0) { - float frame_time; + 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); @@ -5148,8 +5148,8 @@ void RendererStorageRD::update_particles() { } if (fixed_fps > 0) { - float frame_time; - float decr; + double frame_time; + double decr; if (zero_time_scale) { frame_time = 0.0; decr = 1.0 / fixed_fps; @@ -5157,13 +5157,13 @@ void RendererStorageRD::update_particles() { 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); @@ -5886,6 +5886,10 @@ 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: @@ -5899,6 +5903,12 @@ 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: { } } @@ -5935,8 +5945,11 @@ 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); } } @@ -7523,7 +7536,7 @@ 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; @@ -7539,7 +7552,7 @@ 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; } } @@ -7562,7 +7575,7 @@ 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) { @@ -7592,7 +7605,7 @@ 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; } } @@ -7913,14 +7926,14 @@ 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 { @@ -8398,10 +8411,10 @@ 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")); @@ -8569,8 +8582,8 @@ 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); @@ -8582,8 +8595,8 @@ 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); @@ -8792,8 +8805,13 @@ 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() { @@ -9112,25 +9130,41 @@ RendererStorageRD::RendererStorageRD() { 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; + 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; + 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")); @@ -9373,7 +9407,13 @@ 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 process() { 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); @@ -9517,4 +9557,9 @@ 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 1a33569c33..71c3d8b09a 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -715,12 +715,12 @@ 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; + double lifetime = 1.0; + double pre_process_time = 0.0; float explosiveness = 0.0; float randomness = 0.0; bool restart_request = false; @@ -773,12 +773,12 @@ private: uint32_t cycle_number = 0; - float speed_scale = 1.0; + double speed_scale = 1.0; int fixed_fps = 30; bool interpolate = true; bool fractional_delta = false; - float frame_remainder = 0; + double frame_remainder = 0; float collision_base_size = 0.01; bool clear = true; @@ -805,14 +805,14 @@ private: } }; - 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 { - float lifetime; + double lifetime; uint32_t clear; uint32_t total_particles; uint32_t trail_size; @@ -1290,7 +1290,7 @@ private: void _update_global_variables(); /* EFFECTS */ - EffectsRD effects; + EffectsRD *effects = NULL; public: virtual bool can_create_resources_async() const; @@ -1888,6 +1888,13 @@ 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); @@ -2137,13 +2144,13 @@ public: 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_pre_process_time(RID p_particles, double 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_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); @@ -2367,6 +2374,7 @@ 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 b347197289..9c1068ea2e 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -571,7 +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++; @@ -605,7 +605,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge if (uniform.scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { //insert, but don't generate any code. p_actions.uniforms->insert(uniform_name, uniform); - continue; //instances are indexed directly, dont need index uniforms + continue; // Instances are indexed directly, don't need index uniforms. } 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 "; @@ -760,11 +760,11 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge 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"; @@ -1351,7 +1351,13 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide 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); @@ -1388,8 +1394,8 @@ void ShaderCompilerRD::initialize(DefaultIdentifierActions p_actions) { 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"); diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index 27305cc938..5bb12fc168 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -116,8 +116,10 @@ void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, con } StringBuilder tohash; - tohash.append("[VersionKey]"); - tohash.append(RenderingDevice::get_singleton()->shader_get_cache_key()); + 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]"); @@ -144,12 +146,14 @@ 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_stages) { - memdelete_arr(p_version->variant_stages); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); } p_version->variants = nullptr; } @@ -203,7 +207,7 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { return; //variant is disabled, return } - Vector<RD::ShaderStageData> &stages = p_version->variant_stages[p_variant]; + Vector<RD::ShaderStageSPIRVData> stages; String error; String current_source; @@ -217,8 +221,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { _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 { @@ -235,8 +239,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { _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 { @@ -254,8 +258,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { 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 { @@ -275,10 +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); + + 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; } } @@ -364,14 +373,12 @@ String ShaderRD::_version_get_sha1(Version *p_version) const { } static const char *shader_file_header = "GDSC"; -static const uint32_t cache_file_version = 1; +static const uint32_t cache_file_version = 2; 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"; - uint64_t time_from = OS::get_singleton()->get_ticks_usec(); - FileAccessRef f = FileAccess::open(path, FileAccess::READ); if (!f) { return false; @@ -390,76 +397,43 @@ bool ShaderRD::_load_from_cache(Version *p_version) { ERR_FAIL_COND_V(variant_count != (uint32_t)variant_defines.size(), false); //should not happen but check - bool success = true; for (uint32_t i = 0; i < variant_count; i++) { - uint32_t stage_count = f->get_32(); - p_version->variant_stages[i].resize(stage_count); - for (uint32_t j = 0; j < stage_count; j++) { - p_version->variant_stages[i].write[j].shader_stage = RD::ShaderStage(f->get_32()); - - int compression = f->get_32(); - uint32_t length = f->get_32(); - - if (compression == 0) { - Vector<uint8_t> data; - data.resize(length); - - f->get_buffer(data.ptrw(), length); - - p_version->variant_stages[i].write[j].spir_v = data; - } else { - Vector<uint8_t> data; - - if (compression == 2) { - //zstd - int smol_length = f->get_32(); - Vector<uint8_t> zstd_data; - - zstd_data.resize(smol_length); - f->get_buffer(zstd_data.ptrw(), smol_length); - - data.resize(length); - Compression::decompress(data.ptrw(), data.size(), zstd_data.ptr(), zstd_data.size(), Compression::MODE_ZSTD); - - } else { - data.resize(length); - f->get_buffer(data.ptrw(), length); - } - - Vector<uint8_t> spirv; - uint32_t spirv_size = smolv::GetDecodedBufferSize(data.ptr(), data.size()); - spirv.resize(spirv_size); - if (!smolv::Decode(data.ptr(), data.size(), spirv.ptrw(), spirv_size)) { - ERR_PRINT("Malformed smolv input uncompressing shader " + name + ", variant #" + itos(i) + " stage :" + itos(j)); - success = false; - break; - } - p_version->variant_stages[i].write[j].spir_v = spirv; - } + 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); - if (!success) { - for (uint32_t i = 0; i < variant_count; i++) { - p_version->variant_stages[i].resize(0); - } - return false; - } + uint32_t br = f->get_buffer(variant_bytes.ptrw(), variant_size); - float time_ms = double(OS::get_singleton()->get_ticks_usec() - time_from) / 1000.0; + ERR_FAIL_COND_V(br != variant_size, false); - print_verbose("Shader cache load success '" + path + "' " + rtos(time_ms) + "ms."); + p_version->variant_data[i] = variant_bytes; + } for (uint32_t i = 0; i < variant_count; i++) { - RID shader = RD::get_singleton()->shader_create(p_version->variant_stages[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; } } - memdelete_arr(p_version->variant_stages); //clear stages - p_version->variant_stages = nullptr; + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; p_version->valid = true; return true; } @@ -476,49 +450,8 @@ void ShaderRD::_save_to_cache(Version *p_version) { f->store_32(variant_count); //variant count for (uint32_t i = 0; i < variant_count; i++) { - f->store_32(p_version->variant_stages[i].size()); //stage count - for (int j = 0; j < p_version->variant_stages[i].size(); j++) { - f->store_32(p_version->variant_stages[i][j].shader_stage); //stage count - Vector<uint8_t> spirv = p_version->variant_stages[i][j].spir_v; - - bool save_uncompressed = true; - if (shader_cache_save_compressed) { - smolv::ByteArray smolv; - bool strip_debug = !shader_cache_save_debug; - if (!smolv::Encode(spirv.ptr(), spirv.size(), smolv, strip_debug ? smolv::kEncodeFlagStripDebugInfo : 0)) { - ERR_PRINT("Error compressing shader " + name + ", variant #" + itos(i) + " stage :" + itos(i)); - } else { - bool compress_zstd = shader_cache_save_compressed_zstd; - - if (compress_zstd) { - Vector<uint8_t> zstd; - zstd.resize(Compression::get_max_compressed_buffer_size(smolv.size(), Compression::MODE_ZSTD)); - int dst_size = Compression::compress(zstd.ptrw(), &smolv[0], smolv.size(), Compression::MODE_ZSTD); - if (dst_size >= 0 && (uint32_t)dst_size < smolv.size()) { - f->store_32(2); //compressed zstd - f->store_32(smolv.size()); //size of smolv buffer - f->store_32(dst_size); //size of smolv buffer - f->store_buffer(zstd.ptr(), dst_size); //smolv buffer - } else { - compress_zstd = false; - } - } - - if (!compress_zstd) { - f->store_32(1); //compressed - f->store_32(smolv.size()); //size of smolv buffer - f->store_buffer(&smolv[0], smolv.size()); //smolv buffer - } - save_uncompressed = false; - } - } - - if (save_uncompressed) { - f->store_32(0); //uncompressed - f->store_32(spirv.size()); //stage count - f->store_buffer(spirv.ptr(), spirv.size()); //stage count - } - } + 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()); } f->close(); @@ -531,8 +464,8 @@ void ShaderRD::_compile_version(Version *p_version) { p_version->dirty = false; p_version->variants = memnew_arr(RID, variant_defines.size()); - typedef Vector<RD::ShaderStageData> ShaderStageArray; - p_version->variant_stages = memnew_arr(ShaderStageArray, 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)) { @@ -571,19 +504,19 @@ void ShaderRD::_compile_version(Version *p_version) { } } memdelete_arr(p_version->variants); - if (p_version->variant_stages) { - memdelete_arr(p_version->variant_stages); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); } p_version->variants = nullptr; - p_version->variant_stages = 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_stages); //clear stages - p_version->variant_stages = nullptr; + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; p_version->valid = true; } diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index 9a68e02007..44705b2bd6 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -59,7 +59,7 @@ class ShaderRD { Map<StringName, CharString> code_sections; Vector<CharString> custom_defines; - Vector<RD::ShaderStageData> *variant_stages = nullptr; + Vector<uint8_t> *variant_data = nullptr; RID *variants = nullptr; //same size as version defines bool valid; @@ -142,7 +142,9 @@ public: ERR_FAIL_COND_V(!variants_enabled[p_variant], RID()); Version *version = version_owner.getornull(p_version); - ERR_FAIL_COND_V(!version, RID()); + if (!version) { + return RID(); + } if (version->dirty) { _compile_version(version); 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..b1d1c2365e --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster.glsl @@ -0,0 +1,228 @@ +/* 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; + +//DOF +#ifdef MODE_DOF_BLUR + +layout(set = 1, binding = 0) uniform sampler2D dof_source_depth; + +#ifdef DOF_QUALITY_LOW +const int dof_kernel_size = 5; +const int dof_kernel_from = 2; +const float dof_kernel[5] = float[](0.153388, 0.221461, 0.250301, 0.221461, 0.153388); +#endif + +#ifdef DOF_QUALITY_MEDIUM +const int dof_kernel_size = 11; +const int dof_kernel_from = 5; +const float dof_kernel[11] = float[](0.055037, 0.072806, 0.090506, 0.105726, 0.116061, 0.119726, 0.116061, 0.105726, 0.090506, 0.072806, 0.055037); + +#endif + +#ifdef DOF_QUALITY_HIGH +const int dof_kernel_size = 21; +const int dof_kernel_from = 10; +const float dof_kernel[21] = float[](0.028174, 0.032676, 0.037311, 0.041944, 0.046421, 0.050582, 0.054261, 0.057307, 0.059587, 0.060998, 0.061476, 0.060998, 0.059587, 0.057307, 0.054261, 0.050582, 0.046421, 0.041944, 0.037311, 0.032676, 0.028174); +#endif + +#endif + +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_DOF_BLUR + + vec4 color_accum = vec4(0.0); + + float depth = texture(dof_source_depth, uv_interp, 0.0).r; + depth = depth * 2.0 - 1.0; + + if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) { + depth = ((depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0; + } else { + depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - depth * (blur.camera_z_far - blur.camera_z_near)); + } + + // mix near and far blur amount + float amount = 1.0; + if (bool(blur.flags & FLAG_DOF_FAR)) { + amount *= 1.0 - smoothstep(blur.dof_far_begin, blur.dof_far_end, depth); + } + if (bool(blur.flags & FLAG_DOF_NEAR)) { + amount *= smoothstep(blur.dof_near_end, blur.dof_near_begin, depth); + } + amount = 1.0 - amount; + + if (amount > 0.0) { + float k_accum = 0.0; + + for (int i = 0; i < dof_kernel_size; i++) { + int int_ofs = i - dof_kernel_from; + vec2 tap_uv = uv_interp + blur.dof_dir * float(int_ofs) * amount * blur.dof_radius; + + float tap_k = dof_kernel[i]; + + float tap_depth = texture(dof_source_depth, tap_uv, 0.0).r; + tap_depth = tap_depth * 2.0 - 1.0; + + if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) { + tap_depth = ((tap_depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0; + } else { + tap_depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - tap_depth * (blur.camera_z_far - blur.camera_z_near)); + } + + // mix near and far blur amount + float tap_amount = 1.0; + if (bool(blur.flags & FLAG_DOF_FAR)) { + tap_amount *= mix(1.0 - smoothstep(blur.dof_far_begin, blur.dof_far_end, tap_depth), 0.0, int_ofs == 0); + } + if (bool(blur.flags & FLAG_DOF_NEAR)) { + tap_amount *= mix(smoothstep(blur.dof_near_end, blur.dof_near_begin, tap_depth), 0.0, int_ofs == 0); + } + tap_amount = 1.0 - tap_amount; + + tap_amount *= tap_amount * tap_amount; //prevent undesired glow effect + + vec4 tap_color = texture(source_color, tap_uv, 0.0) * tap_k; + + k_accum += tap_k * tap_amount; + color_accum += tap_color * tap_amount; + } + + if (k_accum > 0.0) { + color_accum /= k_accum; + } + + frag_color = color_accum; ///k_accum; + } else { + // we are in focus, don't waste time + frag_color = texture(source_color, uv_interp, 0.0); + } + +#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..6ea968e595 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl @@ -0,0 +1,36 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) +#define FLAG_DOF_FAR (1 << 3) +#define FLAG_DOF_NEAR (1 << 4) + +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; + + // DOF. + float dof_far_begin; + float dof_far_end; + float dof_near_begin; + float dof_near_end; + + float dof_radius; + float dof_pad[3]; + + vec2 dof_dir; + float camera_z_far; + float camera_z_near; +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl index 9fa84657d1..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl @@ -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_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 ce7c03c1d4..28f4dc59ec 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl @@ -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/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl index 2fce258cff..9155216d7e 100644 --- a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -3,7 +3,7 @@ #define LIGHT_BAKE_STATIC 2 struct LightData { //this structure needs to be as packed as possible - vec3 position; + highp vec3 position; float inv_radius; vec3 direction; @@ -17,8 +17,8 @@ struct LightData { //this structure needs to be as packed as possible float specular_amount; bool shadow_enabled; - vec4 atlas_rect; // rect in the shadow atlas - mat4 shadow_matrix; + highp vec4 atlas_rect; // rect in the shadow atlas + highp mat4 shadow_matrix; float shadow_bias; float shadow_normal_bias; float transmittance_bias; @@ -27,7 +27,7 @@ struct LightData { //this structure needs to be as packed as possible uint mask; float shadow_volumetric_fog_fade; uint bake_mode; - vec4 projector_rect; //projector rect in srgb decal atlas + highp vec4 projector_rect; //projector rect in srgb decal atlas }; #define REFLECTION_AMBIENT_DISABLED 0 @@ -69,13 +69,13 @@ struct DirectionalLightData { vec4 shadow_bias; vec4 shadow_normal_bias; vec4 shadow_transmittance_bias; - vec4 shadow_z_range; - vec4 shadow_range_begin; + highp vec4 shadow_z_range; + highp vec4 shadow_range_begin; vec4 shadow_split_offsets; - mat4 shadow_matrix1; - mat4 shadow_matrix2; - mat4 shadow_matrix3; - mat4 shadow_matrix4; + highp mat4 shadow_matrix1; + highp mat4 shadow_matrix2; + highp mat4 shadow_matrix3; + highp mat4 shadow_matrix4; vec4 shadow_color1; vec4 shadow_color2; vec4 shadow_color3; 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_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index 4dceeea995..e88e68b511 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -138,7 +138,7 @@ void main() { 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 dont fit precisely in the history frames, must do a tiny bit of + // 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; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 74d5af5cb6..4f140dd10d 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -356,6 +356,24 @@ void main() { #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" /* Varyings */ @@ -448,14 +466,15 @@ layout(location = 0) out vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -#include "scene_forward_lights_inc.glsl" +/* 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 USE_FORWARD_GI +#include "scene_forward_lights_inc.glsl" #include "scene_forward_gi_inc.glsl" -#endif //USE_FORWARD_GI - #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH @@ -793,25 +812,35 @@ void main() { 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 @@ -821,7 +850,12 @@ void main() { } 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); @@ -830,7 +864,11 @@ void main() { 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; + } } } } @@ -874,6 +912,8 @@ void main() { 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; } @@ -963,9 +1003,9 @@ void main() { 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; @@ -1037,7 +1077,7 @@ void main() { } } - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + 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)); @@ -1068,9 +1108,8 @@ void main() { specular_light = spec_accum.rgb; ambient_light = amb_accum.rgb; } -#else - 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; @@ -1101,7 +1140,7 @@ void main() { 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 if (scene_data.ssao_enabled) { float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; @@ -1182,7 +1221,7 @@ void main() { 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 @@ -1228,14 +1267,13 @@ void main() { 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]; \ @@ -1243,168 +1281,105 @@ void main() { 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; - - } else { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 3) + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - 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]; \ @@ -1412,70 +1387,70 @@ void main() { 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 (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); @@ -1554,7 +1529,9 @@ void main() { 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, #endif @@ -1573,9 +1550,6 @@ void main() { #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 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 6599a42bab..b53bf6a6d4 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -52,6 +52,10 @@ 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) @@ -67,22 +71,22 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -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, std430) restrict readonly buffer 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; @@ -94,7 +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; @@ -103,20 +107,20 @@ 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; @@ -128,7 +132,7 @@ struct SDFVoxelGICascadeData { 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; @@ -173,17 +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; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl index b6e89acb46..5920444f7a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -73,7 +73,7 @@ vec3 F0(float metallic, float specular, vec3 albedo) { 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, +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 @@ -92,9 +92,6 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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 @@ -111,11 +108,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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); @@ -125,19 +118,11 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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; @@ -223,26 +208,20 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, //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; + float blinn = pow(cNdotH, shininess); + blinn *= (shininess + 2.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * blinn * f0 * unpackUnorm4x8(orms).w; #elif defined(SPECULAR_PHONG) vec3 R = normalize(-reflect(L, N)); -#ifdef USE_SOFT_SHADOWS float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); -#else - float cRdotV = clamp(dot(R, V), 0.0, 1.0); -#endif 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); + phong *= (shininess + 1.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * phong * f0 * unpackUnorm4x8(orms).w; #elif defined(SPECULAR_TOON) @@ -320,7 +299,7 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float depth = coord.z; //if only one sample is taken, take it from the center - if (scene_data.directional_soft_shadow_samples == 1) { + if (sc_directional_soft_shadow_samples == 1) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -334,11 +313,11 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float avg = 0.0; - for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { + 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(scene_data.directional_soft_shadow_samples)); + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); } float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { @@ -346,7 +325,7 @@ float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { float depth = coord.z; //if only one sample is taken, take it from the center - if (scene_data.soft_shadow_samples == 1) { + if (sc_soft_shadow_samples == 1) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -360,11 +339,11 @@ float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { float avg = 0.0; - for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { + 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(scene_data.soft_shadow_samples)); + return avg * (1.0 / float(sc_soft_shadow_samples)); } float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { @@ -380,7 +359,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + 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) { @@ -396,12 +375,12 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex tex_scale *= penumbra; float s = 0.0; - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + 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(scene_data.directional_penumbra_shadow_samples); + return s / float(sc_directional_penumbra_shadow_samples); } else { //no blockers found, so no shadow @@ -442,8 +421,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { float shadow; -#ifdef USE_SOFT_SHADOWS - if (omni_lights.data[idx].soft_shadow_size > 0.0) { + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -468,7 +446,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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++) { + for (uint i = 0; i < sc_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; @@ -505,7 +483,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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++) { + for (uint i = 0; i < sc_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; @@ -526,14 +504,13 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_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; @@ -553,9 +530,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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; } @@ -592,14 +567,12 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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) { + 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)); } -#endif #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; //no transmittance by default @@ -633,9 +606,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } #endif -#if 0 - - if (omni_lights.data[idx].projector_rect != vec4(0.0)) { + 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); @@ -653,46 +624,50 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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 (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; - } + 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; + 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; + 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); + 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; - } + 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; + 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; - } + 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); + 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; + } } -#endif light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, + 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 @@ -711,9 +686,6 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif -#ifdef USE_SOFT_SHADOWS - size_A, -#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -747,8 +719,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { 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) { + if (sc_use_light_soft_shadows && spot_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -768,11 +739,11 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { 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++) { + 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 < z_norm) { + if (d < splane.z) { blocker_average += d; blocker_count += 1.0; } @@ -785,13 +756,13 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { uv_size *= penumbra; shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + 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, z_norm, 1.0)); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_penumbra_shadow_samples); } else { //no blockers found, so no shadow @@ -799,14 +770,11 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { } } 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; } @@ -816,6 +784,18 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { 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, @@ -850,20 +830,12 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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) { + 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)); } -#endif - - /* - if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; @@ -886,9 +858,32 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } #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, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, + 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 @@ -907,9 +902,6 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif -#ifdef USE_SOFT_SHADOW - size_A, -#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl index 1bc17e140f..2babe92c1c 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -59,27 +59,27 @@ layout(location = 11) in vec4 weight_attrib; /* Varyings */ -layout(location = 0) out vec3 vertex_interp; +layout(location = 0) highp out vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) out vec3 normal_interp; +layout(location = 1) mediump out vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) out vec4 color_interp; +layout(location = 2) mediump out vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) out vec2 uv_interp; +layout(location = 3) mediump out vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) out vec2 uv2_interp; +layout(location = 4) mediump out vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) out vec3 tangent_interp; -layout(location = 6) out vec3 binormal_interp; +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; #endif #ifdef MATERIAL_UNIFORMS_USED @@ -370,37 +370,61 @@ void main() { #VERSION_DEFINES +//use medium precision for floats on mobile. + +precision mediump float; + +/* Specialization Constants */ + +/* 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 our forward mobile UBOs definitions etc. */ #include "scene_forward_mobile_inc.glsl" /* Varyings */ -layout(location = 0) in vec3 vertex_interp; +layout(location = 0) highp in vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) in vec3 normal_interp; +layout(location = 1) mediump in vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) in vec4 color_interp; +layout(location = 2) mediump in vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) in vec2 uv_interp; +layout(location = 3) mediump in vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) in vec2 uv2_interp; +layout(location = 4) mediump in vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) in vec3 tangent_interp; -layout(location = 6) in vec3 binormal_interp; +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; #endif #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) in float dp_clip; +layout(location = 8) highp in float dp_clip; #endif @@ -471,6 +495,11 @@ layout(location = 0) out vec4 frag_color; #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) @@ -823,6 +852,8 @@ void main() { 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; } @@ -961,7 +992,7 @@ void main() { 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 @@ -1284,7 +1315,7 @@ void main() { 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, + 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 diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl index d4ebcbeec4..f1e554d01c 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -16,12 +16,12 @@ /* don't exceed 128 bytes!! */ /* put instance data into our push content, not a array */ layout(push_constant, binding = 0, std430) uniform DrawCall { - mat4 transform; // 64 - 64 + 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 - vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed uvec2 reflection_probes; // 08 - 104 uvec2 omni_lights; // 08 - 112 @@ -51,6 +51,9 @@ 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) @@ -66,22 +69,22 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -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, std430) restrict readonly buffer 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; @@ -93,7 +96,7 @@ 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; @@ -102,20 +105,20 @@ 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; @@ -123,29 +126,24 @@ global_variables; /* Set 1: Render Pass (changes per render pass) */ layout(set = 1, binding = 0, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - mat4 camera_matrix; - mat4 inv_camera_matrix; + highp mat4 projection_matrix; + highp mat4 inv_projection_matrix; + highp mat4 camera_matrix; + highp mat4 inv_camera_matrix; // only used for multiview - mat4 projection_matrix_view[MAX_VIEWS]; - mat4 inv_projection_matrix_view[MAX_VIEWS]; + highp mat4 projection_matrix_view[MAX_VIEWS]; + highp mat4 inv_projection_matrix_view[MAX_VIEWS]; vec2 viewport_size; vec2 screen_pixel_size; - //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; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index 99db35bb34..d6e5c6a92e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -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_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index bc376e9522..eedd28959c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -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 aa4ded146f..4d9fa85a74 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl @@ -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/tonemap.glsl b/servers/rendering/renderer_rd/shaders/tonemap.glsl index 23f83b3b9c..3c685c25b9 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/tonemap.glsl @@ -35,11 +35,15 @@ void main() { layout(location = 0) in vec2 uv_interp; -#ifdef MULTIVIEW +#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 @@ -180,10 +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); } @@ -207,7 +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); } } @@ -291,6 +291,7 @@ 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); @@ -346,6 +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) @@ -360,7 +362,10 @@ vec3 screen_space_dither(vec2 frag_coord) { } void main() { -#ifdef MULTIVIEW +#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; @@ -370,14 +375,16 @@ void main() { 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); @@ -386,15 +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) { @@ -406,6 +419,7 @@ void main() { color = apply_glow(color, glow); } +#endif // Additional effects |