diff options
Diffstat (limited to 'servers/rendering')
20 files changed, 312 insertions, 122 deletions
diff --git a/servers/rendering/renderer_rd/effects/ss_effects.cpp b/servers/rendering/renderer_rd/effects/ss_effects.cpp index 315bea2e67..582c5abbdd 100644 --- a/servers/rendering/renderer_rd/effects/ss_effects.cpp +++ b/servers/rendering/renderer_rd/effects/ss_effects.cpp @@ -443,6 +443,11 @@ void SSEffects::downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_ RD::get_singleton()->draw_command_begin_label("Downsample Depth"); if (p_invalidate_uniform_set || use_full_mips != ss_effects.used_full_mips_last_frame || use_half_size != ss_effects.used_half_size_last_frame || use_mips != ss_effects.used_mips_last_frame) { + if (ss_effects.downsample_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(ss_effects.downsample_uniform_set)) { + RD::get_singleton()->free(ss_effects.downsample_uniform_set); + ss_effects.downsample_uniform_set = RID(); + } + Vector<RD::Uniform> uniforms; { RD::Uniform u; @@ -516,6 +521,7 @@ void SSEffects::downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_ ss_effects.used_full_mips_last_frame = use_full_mips; ss_effects.used_half_size_last_frame = use_half_size; + ss_effects.used_mips_last_frame = use_mips; } /* SSIL */ @@ -1484,7 +1490,7 @@ void SSEffects::ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const Rend } } -void SSEffects::screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, const RID *p_metallic_slices, const Color &p_metallic_mask, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets) { +void SSEffects::screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets) { UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); ERR_FAIL_NULL(uniform_set_cache); MaterialStorage *material_storage = MaterialStorage::get_singleton(); @@ -1579,10 +1585,6 @@ void SSEffects::screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const R push_constant.proj_info[1] = -2.0f / (p_screen_size.height * p_projections[v].matrix[1][1]); push_constant.proj_info[2] = (1.0f - p_projections[v].matrix[0][2]) / p_projections[v].matrix[0][0]; push_constant.proj_info[3] = (1.0f + p_projections[v].matrix[1][2]) / p_projections[v].matrix[1][1]; - push_constant.metallic_mask[0] = CLAMP(p_metallic_mask.r * 255.0, 0, 255); - push_constant.metallic_mask[1] = CLAMP(p_metallic_mask.g * 255.0, 0, 255); - push_constant.metallic_mask[2] = CLAMP(p_metallic_mask.b * 255.0, 0, 255); - push_constant.metallic_mask[3] = CLAMP(p_metallic_mask.a * 255.0, 0, 255); ScreenSpaceReflectionMode mode = (p_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) ? SCREEN_SPACE_REFLECTION_ROUGH : SCREEN_SPACE_REFLECTION_NORMAL; RID shader = ssr.shader.version_get_shader(ssr.shader_version, mode); diff --git a/servers/rendering/renderer_rd/effects/ss_effects.h b/servers/rendering/renderer_rd/effects/ss_effects.h index a60f3a48ab..d50319c46f 100644 --- a/servers/rendering/renderer_rd/effects/ss_effects.h +++ b/servers/rendering/renderer_rd/effects/ss_effects.h @@ -168,7 +168,7 @@ public: }; void ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const RenderingDevice::DataFormat p_color_format, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, const Size2i &p_screen_size, const uint32_t p_view_count); - void screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, RS::EnvironmentSSRRoughnessQuality p_roughness_quality, const RID *p_metallic_slices, const Color &p_metallic_mask, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets); + void screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, RS::EnvironmentSSRRoughnessQuality p_roughness_quality, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets); void ssr_free(SSRRenderBuffers &p_ssr_buffers); /* subsurface scattering */ @@ -465,10 +465,7 @@ private: uint32_t orthogonal; // 4 - 52 float filter_mipmap_levels; // 4 - 56 uint32_t use_half_res; // 4 - 60 - uint8_t metallic_mask[4]; // 4 - 64 - - uint32_t view_index; // 4 - 68 - uint32_t pad[3]; // 12 - 80 + uint32_t view_index; // 4 - 64 // float projection[16]; // this is in our ScreenSpaceReflectionSceneData now }; diff --git a/servers/rendering/renderer_rd/environment/sky.cpp b/servers/rendering/renderer_rd/environment/sky.cpp index 65d1d9e705..307af99e91 100644 --- a/servers/rendering/renderer_rd/environment/sky.cpp +++ b/servers/rendering/renderer_rd/environment/sky.cpp @@ -907,6 +907,7 @@ void SkyRD::init() { actions.usage_defines["HALF_RES_COLOR"] = "\n#define USES_HALF_RES_COLOR\n"; actions.usage_defines["QUARTER_RES_COLOR"] = "\n#define USES_QUARTER_RES_COLOR\n"; actions.render_mode_defines["disable_fog"] = "#define DISABLE_FOG\n"; + actions.render_mode_defines["use_debanding"] = "#define USE_DEBANDING\n"; actions.sampler_array_name = "material_samplers"; actions.base_texture_binding_index = 1; @@ -1260,6 +1261,7 @@ void SkyRD::setup(RID p_env, Ref<RenderSceneBuffersRD> p_render_buffers, const P light_data_dirty = true; for (uint32_t i = sky_scene_state.ubo.directional_light_count; i < sky_scene_state.max_directional_lights; i++) { sky_scene_state.directional_lights[i].enabled = false; + sky_scene_state.last_frame_directional_lights[i].enabled = false; } } 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 a0f6e69fd5..33d7ddd6e8 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -124,7 +124,7 @@ void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RenderS render_buffers = p_render_buffers; ERR_FAIL_NULL(render_buffers); - bool msaa_3d = render_buffers->get_msaa_3d(); + RS::ViewportMSAA msaa_3d = render_buffers->get_msaa_3d(); if (msaa_3d != RS::VIEWPORT_MSAA_DISABLED) { RD::DataFormat format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; @@ -477,6 +477,10 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p prev_material_uniform_set = material_uniform_set; } + if (surf->owner->base_flags & INSTANCE_DATA_FLAG_MULTIMESH) { + mesh_storage->_multimesh_get_motion_vectors_offsets(surf->owner->data->base, push_constant.multimesh_motion_vectors_current_offset, push_constant.multimesh_motion_vectors_previous_offset); + } + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; @@ -924,7 +928,7 @@ _FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primit static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; } -void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi, bool p_using_opaque_gi, bool p_append) { +void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, uint32_t p_color_pass_flags = 0, bool p_using_sdfgi, bool p_using_opaque_gi, bool p_append) { RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); if (p_render_list == RENDER_LIST_OPAQUE) { @@ -1149,6 +1153,12 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con scene_state.used_depth_texture = true; } + if (p_color_pass_flags & COLOR_PASS_FLAG_MOTION_VECTORS && flags & INSTANCE_DATA_FLAG_MULTIMESH) { + if (RendererRD::MeshStorage::get_singleton()->_multimesh_enable_motion_vectors(inst->data->base)) { + inst->transforms_uniform_set = mesh_storage->multimesh_get_3d_uniform_set(inst->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + } + } + } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { rl->add_element(surf); @@ -1322,7 +1332,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _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, using_sdfgi, using_sdfgi || using_voxelgi); + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR, color_pass_flags, using_sdfgi, using_sdfgi || using_voxelgi); render_list[RENDER_LIST_OPAQUE].sort_by_key(); render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); _fill_instance_data(RENDER_LIST_OPAQUE, p_render_data->render_info ? p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE] : (int *)nullptr); @@ -1609,7 +1619,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co for (uint32_t v = 0; v < p_render_data->view_count; v++) { specular_views[v] = rb_data->get_specular(v); } - _process_ssr(rb, color_only_framebuffer, normal_roughness_views, rb_data->get_specular(), specular_views, Color(0, 0, 0, 1), p_render_data->environment, p_render_data->view_projection, p_render_data->view_eye_offset, rb->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED); + _process_ssr(rb, color_only_framebuffer, normal_roughness_views, rb_data->get_specular(), specular_views, p_render_data->environment, p_render_data->view_projection, p_render_data->view_eye_offset, rb->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED); RD::get_singleton()->draw_command_end_label(); } else { //just mix specular back @@ -1723,7 +1733,7 @@ void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const Page PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); - _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, false, false, true); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, 0, false, false, true); uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); _fill_instance_data(RENDER_LIST_SECONDARY, p_render_info ? p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW] : (int *)nullptr, render_list_from, render_list_size, false); 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 35379cd69b..9e2586542a 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -337,7 +337,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { struct PushConstant { uint32_t base_index; // uint32_t uv_offset; //packed - uint32_t pad[2]; + uint32_t multimesh_motion_vectors_current_offset; + uint32_t multimesh_motion_vectors_previous_offset; }; struct InstanceData { @@ -429,7 +430,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { void _update_instance_data_buffer(RenderListType p_render_list); void _fill_instance_data(RenderListType p_render_list, int *p_render_info = nullptr, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); - void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, bool p_append = false); + void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, uint32_t p_color_pass_flags, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, bool p_append = false); HashMap<Size2i, RID> sdfgi_framebuffer_size_cache; diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 78c83d1fb4..e2b06bc92f 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -1328,7 +1328,7 @@ void RendererSceneRenderRD::_process_sss(Ref<RenderSceneBuffersRD> p_render_buff } } -void RendererSceneRenderRD::_process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_slices, RID p_specular_buffer, const RID *p_metallic_slices, const Color &p_metallic_mask, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive) { +void RendererSceneRenderRD::_process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_slices, RID p_specular_buffer, const RID *p_metallic_slices, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive) { ERR_FAIL_NULL(ss_effects); ERR_FAIL_COND(p_render_buffers.is_null()); @@ -1355,7 +1355,7 @@ void RendererSceneRenderRD::_process_ssr(Ref<RenderSceneBuffersRD> p_render_buff texture_slices[v] = p_render_buffers->get_internal_texture(v); depth_slices[v] = p_render_buffers->get_depth_texture(v); } - ss_effects->screen_space_reflection(p_render_buffers->ssr, texture_slices, p_normal_slices, ssr_roughness_quality, p_metallic_slices, p_metallic_mask, depth_slices, half_size, environment_get_ssr_max_steps(p_environment), environment_get_ssr_fade_in(p_environment), environment_get_ssr_fade_out(p_environment), environment_get_ssr_depth_tolerance(p_environment), view_count, p_projections, p_eye_offsets); + ss_effects->screen_space_reflection(p_render_buffers->ssr, texture_slices, p_normal_slices, ssr_roughness_quality, p_metallic_slices, depth_slices, half_size, environment_get_ssr_max_steps(p_environment), environment_get_ssr_fade_in(p_environment), environment_get_ssr_fade_out(p_environment), environment_get_ssr_depth_tolerance(p_environment), view_count, p_projections, p_eye_offsets); copy_effects->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : p_render_buffers->get_internal_texture(), p_render_buffers->ssr.output, view_count); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 76d2bc68fe..d47394989f 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -138,7 +138,7 @@ protected: virtual RID _render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) = 0; void _process_ssao(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection); - void _process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_buffer_slices, RID p_specular_buffer, const RID *p_metallic_slices, const Color &p_metallic_mask, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive); + void _process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_buffer_slices, RID p_specular_buffer, const RID *p_metallic_slices, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive); void _process_sss(Ref<RenderSceneBuffersRD> p_render_buffers, const Projection &p_camera); void _process_ssil(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection, const Transform3D &p_transform); diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl index 246ef81cb2..9f86643e52 100644 --- a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl @@ -30,12 +30,7 @@ layout(push_constant, std430) uniform Params { bool orthogonal; float filter_mipmap_levels; bool use_half_res; - uint metallic_mask; - uint view_index; - uint pad1; - uint pad2; - uint pad3; } params; @@ -167,7 +162,7 @@ void main() { if (depth > z_to) { // if depth was surpassed - if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far) { + if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far * 0.95) { // check the depth tolerance and far clip // check that normal is valid found = true; @@ -231,18 +226,20 @@ void main() { } } - // Isn't this going to be overwritten after our endif? - final_color = imageLoad(source_diffuse, ivec2((final_pos - 0.5) * pixel_size)); - imageStore(blur_radius_image, ssC, vec4(blur_radius / 255.0)); //stored in r8 #endif // MODE_ROUGH final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend); - //change blend by metallic - vec4 metallic_mask = unpackUnorm4x8(params.metallic_mask); - final_color.a *= dot(metallic_mask, texelFetch(source_metallic, ssC << 1, 0)); + // Schlick term. + float metallic = texelFetch(source_metallic, ssC << 1, 0).w; + float f0 = mix(0.04, 1.0, metallic); // Assume a "specular" amount of 0.5 + normal.y = -normal.y; + float m = clamp(1.0 - dot(normalize(normal), -view_dir), 0.0, 1.0); + float m2 = m * m; + m = m2 * m2 * m; // pow(m,5) + final_color.a *= f0 + (1.0 - f0) * m; // Fresnel Schlick term. imageStore(ssr_image, ssC, final_color); diff --git a/servers/rendering/renderer_rd/shaders/environment/sky.glsl b/servers/rendering/renderer_rd/shaders/environment/sky.glsl index 0eb0f5f8fd..d523461600 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sky.glsl @@ -153,6 +153,15 @@ layout(set = 3, binding = 0) uniform texture3D volumetric_fog_texture; layout(location = 0) out vec4 frag_color; +#ifdef USE_DEBANDING +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +vec3 interleaved_gradient_noise(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + float res = fract(magic.z * fract(dot(pos, magic.xy))) * 2.0 - 1.0; + return vec3(res, -res, res) / 255.0; +} +#endif + vec4 volumetric_fog_process(vec2 screen_uv) { vec3 fog_pos = vec3(screen_uv, 1.0); @@ -252,4 +261,8 @@ void main() { // For mobile renderer we're multiplying by 0.5 as we're using a UNORM buffer. // For both mobile and clustered, we also bake in the exposure value for the environment and camera. frag_color.rgb = frag_color.rgb * params.luminance_multiplier; + +#ifdef USE_DEBANDING + frag_color.rgb += interleaved_gradient_noise(gl_FragCoord.xy); +#endif } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 26b96b358f..8e8ec6883a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -129,7 +129,7 @@ invariant gl_Position; #GLOBALS -void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { +void vertex_shader(in uint instance_index, in bool is_multimesh, in uint multimesh_offset, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; @@ -208,7 +208,7 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc } } - uint offset = stride * gl_InstanceIndex; + uint offset = stride * (gl_InstanceIndex + multimesh_offset); if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); @@ -397,13 +397,13 @@ void main() { mat4 model_matrix = instances.data[instance_index].transform; #if defined(MOTION_VECTORS) global_time = scene_data_block.prev_data.time; - vertex_shader(instance_index, is_multimesh, scene_data_block.prev_data, instances.data[instance_index].prev_transform, prev_screen_position); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_previous_offset, scene_data_block.prev_data, instances.data[instance_index].prev_transform, prev_screen_position); global_time = scene_data_block.data.time; - vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, scene_data_block.data, model_matrix, screen_position); #else global_time = scene_data_block.data.time; vec4 screen_position; - vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, scene_data_block.data, model_matrix, screen_position); #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 45484b8c47..d97da0c571 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -29,8 +29,8 @@ layout(push_constant, std430) uniform DrawCall { uint instance_index; uint uv_offset; - uint pad0; - uint pad1; + uint multimesh_motion_vectors_current_offset; + uint multimesh_motion_vectors_previous_offset; } draw_call; 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 4e6e29b315..2fba1351f7 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -97,11 +97,12 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance #if defined(DIFFUSE_LAMBERT_WRAP) - // energy conserving lambert wrap shader - diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); + // Energy conserving lambert wrap shader. + // https://web.archive.org/web/20210228210901/http://blog.stevemcauley.com/2011/12/03/energy-conserving-wrapped-diffuse/ + diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))) * (1.0 / M_PI); #elif defined(DIFFUSE_TOON) - diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); + diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL) * (1.0 / M_PI); #elif defined(DIFFUSE_BURLEY) diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp index 49e3543ba5..1827b73507 100644 --- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp @@ -29,6 +29,7 @@ /*************************************************************************/ #include "mesh_storage.h" +#include "../../rendering_server_globals.h" using namespace RendererRD; @@ -1211,7 +1212,13 @@ void MeshStorage::multimesh_allocate_data(RID p_multimesh, int p_instances, RS:: if (multimesh->data_cache_dirty_regions) { memdelete_arr(multimesh->data_cache_dirty_regions); multimesh->data_cache_dirty_regions = nullptr; - multimesh->data_cache_used_dirty_regions = 0; + multimesh->data_cache_dirty_region_count = 0; + } + + if (multimesh->previous_data_cache_dirty_regions) { + memdelete_arr(multimesh->previous_data_cache_dirty_regions); + multimesh->previous_data_cache_dirty_regions = nullptr; + multimesh->previous_data_cache_dirty_region_count = 0; } multimesh->instances = p_instances; @@ -1228,14 +1235,67 @@ void MeshStorage::multimesh_allocate_data(RID p_multimesh, int p_instances, RS:: multimesh->aabb = AABB(); multimesh->aabb_dirty = false; multimesh->visible_instances = MIN(multimesh->visible_instances, multimesh->instances); + multimesh->motion_vectors_current_offset = 0; + multimesh->motion_vectors_previous_offset = 0; + multimesh->motion_vectors_last_change = -1; if (multimesh->instances) { - multimesh->buffer = RD::get_singleton()->storage_buffer_create(multimesh->instances * multimesh->stride_cache * 4); + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache * sizeof(float); + if (multimesh->motion_vectors_enabled) { + buffer_size *= 2; + } + multimesh->buffer = RD::get_singleton()->storage_buffer_create(buffer_size); } multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MULTIMESH); } +bool MeshStorage::_multimesh_enable_motion_vectors(RID p_multimesh) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, false); + + if (multimesh->motion_vectors_enabled) { + return false; + } + + multimesh->motion_vectors_enabled = true; + + multimesh->motion_vectors_current_offset = 0; + multimesh->motion_vectors_previous_offset = 0; + multimesh->motion_vectors_last_change = -1; + + if (!multimesh->data_cache.is_empty()) { + multimesh->data_cache.append_array(multimesh->data_cache); + } + + if (multimesh->buffer_set) { + RD::get_singleton()->barrier(); + Vector<uint8_t> buffer_data = RD::get_singleton()->buffer_get_data(multimesh->buffer); + if (!multimesh->data_cache.is_empty()) { + memcpy(buffer_data.ptrw(), multimesh->data_cache.ptr(), buffer_data.size()); + } + + RD::get_singleton()->free(multimesh->buffer); + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache * sizeof(float) * 2; + multimesh->buffer = RD::get_singleton()->storage_buffer_create(buffer_size); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, buffer_data.size(), buffer_data.ptr(), RD::BARRIER_MASK_NO_BARRIER); + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_data.size(), buffer_data.size(), buffer_data.ptr()); + multimesh->uniform_set_3d = RID(); // Cleared by dependency + return true; + } + return false; // Update the transforms uniform set cache +} + +void MeshStorage::_multimesh_get_motion_vectors_offsets(RID p_multimesh, uint32_t &r_current_offset, uint32_t &r_prev_offset) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + r_current_offset = multimesh->motion_vectors_current_offset; + if (RSG::rasterizer->get_frame_number() - multimesh->motion_vectors_last_change >= 2) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + } + r_prev_offset = multimesh->motion_vectors_previous_offset; +} + int MeshStorage::multimesh_get_instance_count(RID p_multimesh) const { MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); ERR_FAIL_COND_V(!multimesh, 0); @@ -1261,7 +1321,7 @@ void MeshStorage::multimesh_set_mesh(RID p_multimesh, RID p_mesh) { //need to re-create AABB unfortunately, calling this has a penalty if (multimesh->buffer_set) { Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); - const uint8_t *r = buffer.ptr(); + const uint8_t *r = buffer.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); const float *data = reinterpret_cast<const float *>(r); _multimesh_re_create_aabb(multimesh, data, multimesh->instances); } @@ -1276,10 +1336,14 @@ void MeshStorage::_multimesh_make_local(MultiMesh *multimesh) const { if (multimesh->data_cache.size() > 0) { return; //already local } - ERR_FAIL_COND(multimesh->data_cache.size() > 0); + // this means that the user wants to load/save individual elements, // for this, the data must reside on CPU, so just copy it there. - multimesh->data_cache.resize(multimesh->instances * multimesh->stride_cache); + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache; + if (multimesh->motion_vectors_enabled) { + buffer_size *= 2; + } + multimesh->data_cache.resize(buffer_size); { float *w = multimesh->data_cache.ptrw(); @@ -1290,15 +1354,48 @@ void MeshStorage::_multimesh_make_local(MultiMesh *multimesh) const { memcpy(w, r, buffer.size()); } } else { - memset(w, 0, (size_t)multimesh->instances * multimesh->stride_cache * sizeof(float)); + memset(w, 0, buffer_size * sizeof(float)); } } uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; multimesh->data_cache_dirty_regions = memnew_arr(bool, data_cache_dirty_region_count); - for (uint32_t i = 0; i < data_cache_dirty_region_count; i++) { - multimesh->data_cache_dirty_regions[i] = false; + memset(multimesh->data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); + multimesh->data_cache_dirty_region_count = 0; + + multimesh->previous_data_cache_dirty_regions = memnew_arr(bool, data_cache_dirty_region_count); + memset(multimesh->previous_data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); + multimesh->previous_data_cache_dirty_region_count = 0; +} + +void MeshStorage::_multimesh_update_motion_vectors_data_cache(MultiMesh *multimesh) { + ERR_FAIL_COND(multimesh->data_cache.is_empty()); + + if (!multimesh->motion_vectors_enabled) { + return; + } + + uint32_t frame = RSG::rasterizer->get_frame_number(); + if (multimesh->motion_vectors_last_change != frame) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_current_offset = multimesh->instances - multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_last_change = frame; + + if (multimesh->previous_data_cache_dirty_region_count > 0) { + uint8_t *data = (uint8_t *)multimesh->data_cache.ptrw(); + uint32_t current_ofs = multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + uint32_t previous_ofs = multimesh->motion_vectors_previous_offset * multimesh->stride_cache * sizeof(float); + uint32_t visible_instances = multimesh->visible_instances >= 0 ? multimesh->visible_instances : multimesh->instances; + uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); + uint32_t size = multimesh->stride_cache * (uint32_t)multimesh->instances * (uint32_t)sizeof(float); + for (uint32_t i = 0; i < visible_region_count; i++) { + if (multimesh->previous_data_cache_dirty_regions[i]) { + uint32_t offset = i * region_size; + memcpy(data + current_ofs + offset, data + previous_ofs + offset, MIN(region_size, size - offset)); + } + } + } } - multimesh->data_cache_used_dirty_regions = 0; } void MeshStorage::_multimesh_mark_dirty(MultiMesh *multimesh, int p_index, bool p_aabb) { @@ -1309,7 +1406,7 @@ void MeshStorage::_multimesh_mark_dirty(MultiMesh *multimesh, int p_index, bool #endif if (!multimesh->data_cache_dirty_regions[region_index]) { multimesh->data_cache_dirty_regions[region_index] = true; - multimesh->data_cache_used_dirty_regions++; + multimesh->data_cache_dirty_region_count++; } if (p_aabb) { @@ -1330,7 +1427,7 @@ void MeshStorage::_multimesh_mark_all_dirty(MultiMesh *multimesh, bool p_data, b for (uint32_t i = 0; i < data_cache_dirty_region_count; i++) { if (!multimesh->data_cache_dirty_regions[i]) { multimesh->data_cache_dirty_regions[i] = true; - multimesh->data_cache_used_dirty_regions++; + multimesh->data_cache_dirty_region_count++; } } } @@ -1395,11 +1492,12 @@ void MeshStorage::multimesh_instance_set_transform(RID p_multimesh, int p_index, ERR_FAIL_COND(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_3D); _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); { float *w = multimesh->data_cache.ptrw(); - float *dataptr = w + p_index * multimesh->stride_cache; + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; dataptr[0] = p_transform.basis.rows[0][0]; dataptr[1] = p_transform.basis.rows[0][1]; @@ -1425,11 +1523,12 @@ void MeshStorage::multimesh_instance_set_transform_2d(RID p_multimesh, int p_ind ERR_FAIL_COND(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_2D); _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); { float *w = multimesh->data_cache.ptrw(); - float *dataptr = w + p_index * multimesh->stride_cache; + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; dataptr[0] = p_transform.columns[0][0]; dataptr[1] = p_transform.columns[1][0]; @@ -1451,11 +1550,12 @@ void MeshStorage::multimesh_instance_set_color(RID p_multimesh, int p_index, con ERR_FAIL_COND(!multimesh->uses_colors); _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); { float *w = multimesh->data_cache.ptrw(); - float *dataptr = w + p_index * multimesh->stride_cache + multimesh->color_offset_cache; + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->color_offset_cache; dataptr[0] = p_color.r; dataptr[1] = p_color.g; @@ -1473,11 +1573,12 @@ void MeshStorage::multimesh_instance_set_custom_data(RID p_multimesh, int p_inde ERR_FAIL_COND(!multimesh->uses_custom_data); _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); { float *w = multimesh->data_cache.ptrw(); - float *dataptr = w + p_index * multimesh->stride_cache + multimesh->custom_data_offset_cache; + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->custom_data_offset_cache; dataptr[0] = p_color.r; dataptr[1] = p_color.g; @@ -1514,7 +1615,7 @@ Transform3D MeshStorage::multimesh_instance_get_transform(RID p_multimesh, int p { const float *r = multimesh->data_cache.ptr(); - const float *dataptr = r + p_index * multimesh->stride_cache; + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; t.basis.rows[0][0] = dataptr[0]; t.basis.rows[0][1] = dataptr[1]; @@ -1545,7 +1646,7 @@ Transform2D MeshStorage::multimesh_instance_get_transform_2d(RID p_multimesh, in { const float *r = multimesh->data_cache.ptr(); - const float *dataptr = r + p_index * multimesh->stride_cache; + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; t.columns[0][0] = dataptr[0]; t.columns[1][0] = dataptr[1]; @@ -1570,7 +1671,7 @@ Color MeshStorage::multimesh_instance_get_color(RID p_multimesh, int p_index) co { const float *r = multimesh->data_cache.ptr(); - const float *dataptr = r + p_index * multimesh->stride_cache + multimesh->color_offset_cache; + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->color_offset_cache; c.r = dataptr[0]; c.g = dataptr[1]; @@ -1593,7 +1694,7 @@ Color MeshStorage::multimesh_instance_get_custom_data(RID p_multimesh, int p_ind { const float *r = multimesh->data_cache.ptr(); - const float *dataptr = r + p_index * multimesh->stride_cache + multimesh->custom_data_offset_cache; + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->custom_data_offset_cache; c.r = dataptr[0]; c.g = dataptr[1]; @@ -1609,25 +1710,26 @@ void MeshStorage::multimesh_set_buffer(RID p_multimesh, const Vector<float> &p_b ERR_FAIL_COND(!multimesh); ERR_FAIL_COND(p_buffer.size() != (multimesh->instances * (int)multimesh->stride_cache)); + if (multimesh->motion_vectors_enabled) { + uint32_t frame = RSG::rasterizer->get_frame_number(); + + if (multimesh->motion_vectors_last_change != frame) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_current_offset = multimesh->instances - multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_last_change = frame; + } + } + { const float *r = p_buffer.ptr(); - RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r); + RD::get_singleton()->buffer_update(multimesh->buffer, multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float), p_buffer.size() * sizeof(float), r); multimesh->buffer_set = true; } if (multimesh->data_cache.size()) { - //if we have a data cache, just update it - multimesh->data_cache = p_buffer; - { - //clear dirty since nothing will be dirty anymore - uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; - for (uint32_t i = 0; i < data_cache_dirty_region_count; i++) { - multimesh->data_cache_dirty_regions[i] = false; - } - multimesh->data_cache_used_dirty_regions = 0; - } - - _multimesh_mark_all_dirty(multimesh, false, true); //update AABB + float *cache_data = multimesh->data_cache.ptrw(); + memcpy(cache_data + (multimesh->motion_vectors_current_offset * multimesh->stride_cache), p_buffer.ptr(), p_buffer.size() * sizeof(float)); + _multimesh_mark_all_dirty(multimesh, true, true); //update AABB } else if (multimesh->mesh.is_valid()) { //if we have a mesh set, we need to re-generate the AABB from the new data const float *data = p_buffer.ptr(); @@ -1642,20 +1744,19 @@ Vector<float> MeshStorage::multimesh_get_buffer(RID p_multimesh) const { ERR_FAIL_COND_V(!multimesh, Vector<float>()); if (multimesh->buffer.is_null()) { return Vector<float>(); - } else if (multimesh->data_cache.size()) { - return multimesh->data_cache; } else { - //get from memory - - Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); Vector<float> ret; ret.resize(multimesh->instances * multimesh->stride_cache); - { - float *w = ret.ptrw(); - const uint8_t *r = buffer.ptr(); - memcpy(w, r, buffer.size()); - } + float *w = ret.ptrw(); + if (multimesh->data_cache.size()) { + const uint8_t *r = (uint8_t *)multimesh->data_cache.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + memcpy(w, r, ret.size() * sizeof(float)); + } else { + Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); + const uint8_t *r = buffer.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + memcpy(w, r, ret.size() * sizeof(float)); + } return ret; } } @@ -1698,36 +1799,38 @@ void MeshStorage::_update_dirty_multimeshes() { MultiMesh *multimesh = multimesh_dirty_list; if (multimesh->data_cache.size()) { //may have been cleared, so only process if it exists - const float *data = multimesh->data_cache.ptr(); uint32_t visible_instances = multimesh->visible_instances >= 0 ? multimesh->visible_instances : multimesh->instances; + uint32_t buffer_offset = multimesh->motion_vectors_current_offset * multimesh->stride_cache; + const float *data = multimesh->data_cache.ptr() + buffer_offset; - if (multimesh->data_cache_used_dirty_regions) { + uint32_t total_dirty_regions = multimesh->data_cache_dirty_region_count + multimesh->previous_data_cache_dirty_region_count; + if (total_dirty_regions != 0) { uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); - - if (multimesh->data_cache_used_dirty_regions > 32 || multimesh->data_cache_used_dirty_regions > visible_region_count / 2) { + if (total_dirty_regions > 32 || total_dirty_regions > visible_region_count / 2) { //if there too many dirty regions, or represent the majority of regions, just copy all, else transfer cost piles up too much - RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * (uint32_t)multimesh->stride_cache * (uint32_t)sizeof(float)), data); + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_offset * sizeof(float), MIN(visible_region_count * region_size, multimesh->instances * (uint32_t)multimesh->stride_cache * (uint32_t)sizeof(float)), data); } else { //not that many regions? update them all for (uint32_t i = 0; i < visible_region_count; i++) { - if (multimesh->data_cache_dirty_regions[i]) { + if (multimesh->data_cache_dirty_regions[i] || multimesh->previous_data_cache_dirty_regions[i]) { uint32_t offset = i * region_size; uint32_t size = multimesh->stride_cache * (uint32_t)multimesh->instances * (uint32_t)sizeof(float); uint32_t region_start_index = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * i; - RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[region_start_index]); + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_offset * sizeof(float) + offset, MIN(region_size, size - offset), &data[region_start_index], RD::BARRIER_MASK_NO_BARRIER); } } + RD::get_singleton()->barrier(RD::BARRIER_MASK_NO_BARRIER, RD::BARRIER_MASK_ALL); } - for (uint32_t i = 0; i < data_cache_dirty_region_count; i++) { - multimesh->data_cache_dirty_regions[i] = false; - } + memcpy(multimesh->previous_data_cache_dirty_regions, multimesh->data_cache_dirty_regions, data_cache_dirty_region_count * sizeof(bool)); + memset(multimesh->data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); - multimesh->data_cache_used_dirty_regions = 0; + multimesh->previous_data_cache_dirty_region_count = multimesh->data_cache_dirty_region_count; + multimesh->data_cache_dirty_region_count = 0; } if (multimesh->aabb_dirty) { diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h index 5c0d019c15..622f3911c7 100644 --- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h @@ -205,13 +205,19 @@ private: AABB aabb; bool aabb_dirty = false; bool buffer_set = false; + bool motion_vectors_enabled = false; + uint32_t motion_vectors_current_offset = 0; + uint32_t motion_vectors_previous_offset = 0; + uint64_t motion_vectors_last_change = -1; uint32_t stride_cache = 0; uint32_t color_offset_cache = 0; uint32_t custom_data_offset_cache = 0; Vector<float> data_cache; //used if individual setting is used bool *data_cache_dirty_regions = nullptr; - uint32_t data_cache_used_dirty_regions = 0; + uint32_t data_cache_dirty_region_count = 0; + bool *previous_data_cache_dirty_regions = nullptr; + uint32_t previous_data_cache_dirty_region_count = 0; RID buffer; //storage buffer RID uniform_set_3d; @@ -228,6 +234,7 @@ private: MultiMesh *multimesh_dirty_list = nullptr; _FORCE_INLINE_ void _multimesh_make_local(MultiMesh *multimesh) const; + _FORCE_INLINE_ void _multimesh_update_motion_vectors_data_cache(MultiMesh *multimesh); _FORCE_INLINE_ void _multimesh_mark_dirty(MultiMesh *multimesh, int p_index, bool p_aabb); _FORCE_INLINE_ void _multimesh_mark_all_dirty(MultiMesh *multimesh, bool p_data, bool p_aabb); _FORCE_INLINE_ void _multimesh_re_create_aabb(MultiMesh *multimesh, const float *p_data, int p_instances); @@ -579,6 +586,8 @@ public: virtual AABB multimesh_get_aabb(RID p_multimesh) const override; void _update_dirty_multimeshes(); + bool _multimesh_enable_motion_vectors(RID p_multimesh); + void _multimesh_get_motion_vectors_offsets(RID p_multimesh, uint32_t &r_current_offset, uint32_t &r_prev_offset); _FORCE_INLINE_ RS::MultimeshTransformFormat multimesh_get_transform_format(RID p_multimesh) const { MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp index 576ec81124..16fdbc07f5 100644 --- a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp @@ -128,6 +128,11 @@ void RenderSceneBuffersRD::cleanup() { ss_effects.linear_depth_slices.clear(); } + if (ss_effects.downsample_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(ss_effects.downsample_uniform_set)) { + RD::get_singleton()->free(ss_effects.downsample_uniform_set); + ss_effects.downsample_uniform_set = RID(); + } + sse->ssao_free(ss_effects.ssao); sse->ssil_free(ss_effects.ssil); sse->ssr_free(ssr); @@ -535,7 +540,9 @@ void RenderSceneBuffersRD::ensure_velocity() { RD::TEXTURE_SAMPLES_8, }; - create_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA, RD::DATA_FORMAT_R16G16_SFLOAT, msaa_usage_bits, ts[msaa_3d]); + RD::TextureSamples texture_samples = ts[msaa_3d]; + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA, RD::DATA_FORMAT_R16G16_SFLOAT, msaa_usage_bits, texture_samples); } create_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY, RD::DATA_FORMAT_R16G16_SFLOAT, usage_bits); diff --git a/servers/rendering/renderer_viewport.cpp b/servers/rendering/renderer_viewport.cpp index d466f90e79..eacd9bbbc2 100644 --- a/servers/rendering/renderer_viewport.cpp +++ b/servers/rendering/renderer_viewport.cpp @@ -73,33 +73,36 @@ static Transform2D _canvas_get_transform(RendererViewport::Viewport *p_viewport, } Vector<RendererViewport::Viewport *> RendererViewport::_sort_active_viewports() { - // We need to sort the viewports in a "topological order", - // children first and parents last, we use the Kahn's algorithm to achieve that. + // We need to sort the viewports in a "topological order", children first and + // parents last. We also need to keep sibling viewports in the original order + // from top to bottom. Vector<Viewport *> result; List<Viewport *> nodes; - for (Viewport *viewport : active_viewports) { + for (int i = active_viewports.size() - 1; i >= 0; --i) { + Viewport *viewport = active_viewports[i]; if (viewport->parent.is_valid()) { continue; } nodes.push_back(viewport); + result.insert(0, viewport); } while (!nodes.is_empty()) { - Viewport *node = nodes[0]; + const Viewport *node = nodes[0]; nodes.pop_front(); - result.insert(0, node); - - for (Viewport *child : active_viewports) { + for (int i = active_viewports.size() - 1; i >= 0; --i) { + Viewport *child = active_viewports[i]; if (child->parent != node->self) { continue; } if (!nodes.find(child)) { nodes.push_back(child); + result.insert(0, child); } } } @@ -714,7 +717,14 @@ void RendererViewport::draw_viewports() { blit_to_screen_list[vp->viewport_to_screen] = Vector<BlitToScreen>(); } - blit_to_screen_list[vp->viewport_to_screen].push_back(blit); + if (OS::get_singleton()->get_current_rendering_driver_name() == "opengl3") { + Vector<BlitToScreen> blit_to_screen_vec; + blit_to_screen_vec.push_back(blit); + RSG::rasterizer->blit_render_targets_to_screen(vp->viewport_to_screen, blit_to_screen_vec.ptr(), 1); + RSG::rasterizer->end_frame(true); + } else { + blit_to_screen_list[vp->viewport_to_screen].push_back(blit); + } } } diff --git a/servers/rendering/rendering_server_default.cpp b/servers/rendering/rendering_server_default.cpp index 1f686069bd..9103b0cf56 100644 --- a/servers/rendering/rendering_server_default.cpp +++ b/servers/rendering/rendering_server_default.cpp @@ -91,7 +91,10 @@ void RenderingServerDefault::_draw(bool p_swap_buffers, double frame_step) { RSG::viewport->draw_viewports(); RSG::canvas_render->update(); - RSG::rasterizer->end_frame(p_swap_buffers); + if (OS::get_singleton()->get_current_rendering_driver_name() != "opengl3") { + // Already called for gl_compatibility renderer. + RSG::rasterizer->end_frame(p_swap_buffers); + } XRServer *xr_server = XRServer::get_singleton(); if (xr_server != nullptr) { diff --git a/servers/rendering/shader_language.cpp b/servers/rendering/shader_language.cpp index 2bbc5e4dfb..6754d84cd4 100644 --- a/servers/rendering/shader_language.cpp +++ b/servers/rendering/shader_language.cpp @@ -257,6 +257,7 @@ enum ContextFlag : uint32_t { CF_UNIFORM_KEYWORD = 2048U, // "uniform" CF_CONST_KEYWORD = 4096U, // "const" CF_UNIFORM_QUALIFIER = 8192U, // "<x> uniform float t;" + CF_SHADER_TYPE = 16384U, // "shader_type" }; const uint32_t KCF_DATATYPE = CF_BLOCK | CF_GLOBAL_SPACE | CF_DATATYPE | CF_FUNC_DECL_PARAM_TYPE | CF_UNIFORM_TYPE; @@ -318,7 +319,7 @@ const ShaderLanguage::KeyWord ShaderLanguage::keyword_list[] = { { TK_VARYING, "varying", CF_GLOBAL_SPACE, { "particles", "sky", "fog" }, {} }, { TK_CONST, "const", CF_BLOCK | CF_GLOBAL_SPACE | CF_CONST_KEYWORD, {}, {} }, { TK_STRUCT, "struct", CF_GLOBAL_SPACE, {}, {} }, - { TK_SHADER_TYPE, "shader_type", CF_GLOBAL_SPACE, {}, {} }, + { TK_SHADER_TYPE, "shader_type", CF_SHADER_TYPE, {}, {} }, { TK_RENDER_MODE, "render_mode", CF_GLOBAL_SPACE, {}, {} }, // uniform qualifiers @@ -1068,7 +1069,7 @@ String ShaderLanguage::get_uniform_hint_name(ShaderNode::Uniform::Hint p_hint) { result = "hint_range"; } break; case ShaderNode::Uniform::HINT_SOURCE_COLOR: { - result = "hint_color"; + result = "source_color"; } break; case ShaderNode::Uniform::HINT_NORMAL: { result = "hint_normal"; @@ -1170,6 +1171,10 @@ void ShaderLanguage::clear() { last_type = IDENTIFIER_MAX; current_uniform_group_name = ""; current_uniform_subgroup_name = ""; + current_uniform_hint = ShaderNode::Uniform::HINT_NONE; + current_uniform_filter = FILTER_DEFAULT; + current_uniform_repeat = REPEAT_DEFAULT; + current_uniform_instance_index_defined = false; completion_type = COMPLETION_NONE; completion_block = nullptr; @@ -1183,7 +1188,7 @@ void ShaderLanguage::clear() { include_positions.push_back(FilePosition()); #ifdef DEBUG_ENABLED - keyword_completion_context = CF_GLOBAL_SPACE; + keyword_completion_context = CF_UNSPECIFIED; used_constants.clear(); used_varyings.clear(); used_uniforms.clear(); @@ -7821,6 +7826,9 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f Token next; if (!is_shader_inc) { +#ifdef DEBUG_ENABLED + keyword_completion_context = CF_SHADER_TYPE; +#endif // DEBUG_ENABLED tk = _get_token(); if (tk.type != TK_SHADER_TYPE) { @@ -8613,6 +8621,7 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f } custom_instance_index = tk.constant; + current_uniform_instance_index_defined = true; if (custom_instance_index >= MAX_INSTANCE_UNIFORM_INDICES) { _set_error(vformat(RTR("Allowed instance uniform indices must be within [0..%d] range."), MAX_INSTANCE_UNIFORM_INDICES - 1)); @@ -8678,6 +8687,7 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f return ERR_PARSE_ERROR; } else { uniform.hint = new_hint; + current_uniform_hint = new_hint; } } @@ -8691,6 +8701,7 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f return ERR_PARSE_ERROR; } else { uniform.filter = new_filter; + current_uniform_filter = new_filter; } } @@ -8704,6 +8715,7 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f return ERR_PARSE_ERROR; } else { uniform.repeat = new_repeat; + current_uniform_repeat = new_repeat; } } @@ -8771,6 +8783,11 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f keyword_completion_context = CF_GLOBAL_SPACE; #endif // DEBUG_ENABLED completion_type = COMPLETION_NONE; + + current_uniform_hint = ShaderNode::Uniform::HINT_NONE; + current_uniform_filter = FILTER_DEFAULT; + current_uniform_repeat = REPEAT_DEFAULT; + current_uniform_instance_index_defined = false; } else { // varying ShaderNode::Varying varying; varying.type = type; @@ -10307,28 +10324,33 @@ Error ShaderLanguage::complete(const String &p_code, const ShaderCompileInfo &p_ } break; case COMPLETION_HINT: { if (completion_base == DataType::TYPE_VEC3 || completion_base == DataType::TYPE_VEC4) { - ScriptLanguage::CodeCompletionOption option("source_color", ScriptLanguage::CODE_COMPLETION_KIND_PLAIN_TEXT); - r_options->push_back(option); + if (current_uniform_hint == ShaderNode::Uniform::HINT_NONE) { + ScriptLanguage::CodeCompletionOption option("source_color", ScriptLanguage::CODE_COMPLETION_KIND_PLAIN_TEXT); + r_options->push_back(option); + } } else if ((completion_base == DataType::TYPE_INT || completion_base == DataType::TYPE_FLOAT) && !completion_base_array) { - ScriptLanguage::CodeCompletionOption option("hint_range", ScriptLanguage::CODE_COMPLETION_KIND_PLAIN_TEXT); + if (current_uniform_hint == ShaderNode::Uniform::HINT_NONE) { + ScriptLanguage::CodeCompletionOption option("hint_range", ScriptLanguage::CODE_COMPLETION_KIND_PLAIN_TEXT); - if (completion_base == DataType::TYPE_INT) { - option.insert_text = "hint_range(0, 100, 1)"; - } else { - option.insert_text = "hint_range(0.0, 1.0, 0.1)"; - } + if (completion_base == DataType::TYPE_INT) { + option.insert_text = "hint_range(0, 100, 1)"; + } else { + option.insert_text = "hint_range(0.0, 1.0, 0.1)"; + } - r_options->push_back(option); + r_options->push_back(option); + } } else if ((int(completion_base) > int(TYPE_MAT4) && int(completion_base) < int(TYPE_STRUCT)) && !completion_base_array) { - static Vector<String> options; - - if (options.is_empty()) { + Vector<String> options; + if (current_uniform_filter == FILTER_DEFAULT) { options.push_back("filter_linear"); options.push_back("filter_linear_mipmap"); options.push_back("filter_linear_mipmap_anisotropic"); options.push_back("filter_nearest"); options.push_back("filter_nearest_mipmap"); options.push_back("filter_nearest_mipmap_anisotropic"); + } + if (current_uniform_hint == ShaderNode::Uniform::HINT_NONE) { options.push_back("hint_anisotropy"); options.push_back("hint_default_black"); options.push_back("hint_default_white"); @@ -10344,6 +10366,8 @@ Error ShaderLanguage::complete(const String &p_code, const ShaderCompileInfo &p_ options.push_back("hint_normal_roughness_texture"); options.push_back("hint_depth_texture"); options.push_back("source_color"); + } + if (current_uniform_repeat == REPEAT_DEFAULT) { options.push_back("repeat_enable"); options.push_back("repeat_disable"); } @@ -10353,7 +10377,7 @@ Error ShaderLanguage::complete(const String &p_code, const ShaderCompileInfo &p_ r_options->push_back(option); } } - if (!completion_base_array) { + if (!completion_base_array && !current_uniform_instance_index_defined) { ScriptLanguage::CodeCompletionOption option("instance_index", ScriptLanguage::CODE_COMPLETION_KIND_PLAIN_TEXT); option.insert_text = "instance_index(0)"; r_options->push_back(option); diff --git a/servers/rendering/shader_language.h b/servers/rendering/shader_language.h index 75b713d167..e9f8c3b289 100644 --- a/servers/rendering/shader_language.h +++ b/servers/rendering/shader_language.h @@ -1050,6 +1050,10 @@ private: }; CompletionType completion_type; + ShaderNode::Uniform::Hint current_uniform_hint = ShaderNode::Uniform::HINT_NONE; + TextureFilter current_uniform_filter = FILTER_DEFAULT; + TextureRepeat current_uniform_repeat = REPEAT_DEFAULT; + bool current_uniform_instance_index_defined = false; int completion_line = 0; BlockNode *completion_block = nullptr; DataType completion_base; diff --git a/servers/rendering/shader_types.cpp b/servers/rendering/shader_types.cpp index 43c483a00d..cd063f91ac 100644 --- a/servers/rendering/shader_types.cpp +++ b/servers/rendering/shader_types.cpp @@ -340,6 +340,11 @@ ShaderTypes::ShaderTypes() { shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["INDEX"] = constt(ShaderLanguage::TYPE_UINT); shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["EMISSION_TRANSFORM"] = constt(ShaderLanguage::TYPE_MAT4); shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["RANDOM_SEED"] = constt(ShaderLanguage::TYPE_UINT); + shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["FLAG_EMIT_POSITION"] = constt(ShaderLanguage::TYPE_UINT); + shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["FLAG_EMIT_ROT_SCALE"] = constt(ShaderLanguage::TYPE_UINT); + shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["FLAG_EMIT_VELOCITY"] = constt(ShaderLanguage::TYPE_UINT); + shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["FLAG_EMIT_COLOR"] = constt(ShaderLanguage::TYPE_UINT); + shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["FLAG_EMIT_CUSTOM"] = constt(ShaderLanguage::TYPE_UINT); shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["RESTART_POSITION"] = constt(ShaderLanguage::TYPE_BOOL); shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["RESTART_ROT_SCALE"] = constt(ShaderLanguage::TYPE_BOOL); shader_modes[RS::SHADER_PARTICLES].functions["start"].built_ins["RESTART_VELOCITY"] = constt(ShaderLanguage::TYPE_BOOL); @@ -385,6 +390,7 @@ ShaderTypes::ShaderTypes() { emit_vertex_func.arguments.push_back(ShaderLanguage::StageFunctionInfo::Argument("custom", ShaderLanguage::TYPE_VEC4)); emit_vertex_func.arguments.push_back(ShaderLanguage::StageFunctionInfo::Argument("flags", ShaderLanguage::TYPE_UINT)); emit_vertex_func.return_type = ShaderLanguage::TYPE_BOOL; //whether it could emit + shader_modes[RS::SHADER_PARTICLES].functions["start"].stage_functions["emit_subparticle"] = emit_vertex_func; shader_modes[RS::SHADER_PARTICLES].functions["process"].stage_functions["emit_subparticle"] = emit_vertex_func; } @@ -444,6 +450,7 @@ ShaderTypes::ShaderTypes() { shader_modes[RS::SHADER_SKY].modes.push_back({ PNAME("use_half_res_pass") }); shader_modes[RS::SHADER_SKY].modes.push_back({ PNAME("use_quarter_res_pass") }); shader_modes[RS::SHADER_SKY].modes.push_back({ PNAME("disable_fog") }); + shader_modes[RS::SHADER_SKY].modes.push_back({ PNAME("use_debanding") }); } /************ FOG **************************/ |