Shadow map rendering optimization

-All shadow rendering is done with raster now (no compute)
-All shadow rendering is done by rendering directly to the shadow atlas
-Improved how buffer clearing is done to optimize the above.
-Ability to set shadows as 16 bits.

servers/rendering/renderer_rd/renderer_scene_render_forward.cpp

@@ -1722,7 +1722,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
 
 	RENDER_TIMESTAMP("Render Opaque Pass");
 
-	RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps);
+	RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true);
 
 	bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss;
 	bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss;
@@ -1838,7 +1838,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
 	}
 }
 
-void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
+void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) {
 	RENDER_TIMESTAMP("Setup Rendering Shadow");
 
 	_update_render_base_uniform_set();
@@ -1847,7 +1847,7 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedAr
 
 	scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1;
 
-	_setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_zfar, false, p_use_pancake);
+	_setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), !p_flip_y, Color(), 0, p_zfar, false, p_use_pancake);
 
 	if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) {
 		p_screen_lod_threshold = 0.0;
@@ -1865,8 +1865,12 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedAr
 
 	{
 		//regular forward for now
-		RenderListParameters render_list_params(render_list.elements, render_list.element_count, p_use_dp_flip, pass_mode, true, rp_uniform_set, false, Vector2(), p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
-		_render_list_with_threads(&render_list_params, p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ);
+		bool flip_cull = p_use_dp_flip;
+		if (p_flip_y) {
+			flip_cull = !flip_cull;
+		}
+		RenderListParameters render_list_params(render_list.elements, render_list.element_count, flip_cull, pass_mode, true, rp_uniform_set, false, Vector2(), p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
+		_render_list_with_threads(&render_list_params, p_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : RD::INITIAL_ACTION_CONTINUE, p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, Vector<Color>(), 1.0, 0, p_rect);
 	}
 }
 
@@ -2128,14 +2132,14 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 
 		{
 			RD::Uniform u;
-			u.binding = 5;
+			u.binding = 4;
 			u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 			u.ids.push_back(get_omni_light_buffer());
 			uniforms.push_back(u);
 		}
 		{
 			RD::Uniform u;
-			u.binding = 6;
+			u.binding = 5;
 			u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 			u.ids.push_back(get_spot_light_buffer());
 			uniforms.push_back(u);
@@ -2143,35 +2147,35 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 
 		{
 			RD::Uniform u;
-			u.binding = 7;
+			u.binding = 6;
 			u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 			u.ids.push_back(get_reflection_probe_buffer());
 			uniforms.push_back(u);
 		}
 		{
 			RD::Uniform u;
-			u.binding = 8;
+			u.binding = 7;
 			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 = 10;
+			u.binding = 8;
 			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 = 11;
+			u.binding = 9;
 			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 = 12;
+			u.binding = 10;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID decal_atlas = storage->decal_atlas_get_texture();
 			u.ids.push_back(decal_atlas);
@@ -2179,7 +2183,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 		}
 		{
 			RD::Uniform u;
-			u.binding = 13;
+			u.binding = 11;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID decal_atlas = storage->decal_atlas_get_texture_srgb();
 			u.ids.push_back(decal_atlas);
@@ -2187,28 +2191,16 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 		}
 		{
 			RD::Uniform u;
-			u.binding = 14;
+			u.binding = 12;
 			u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 			u.ids.push_back(get_decal_buffer());
 			uniforms.push_back(u);
 		}
 
-		{
-			RD::Uniform u;
-			u.binding = 15;
-			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
-			if (directional_shadow_get_texture().is_valid()) {
-				u.ids.push_back(directional_shadow_get_texture());
-			} else {
-				u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
-			}
-			uniforms.push_back(u);
-		}
-
 		{
 			RD::Uniform u;
 			u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
-			u.binding = 16;
+			u.binding = 13;
 			u.ids.push_back(storage->global_variables_get_storage_buffer());
 			uniforms.push_back(u);
 		}
@@ -2216,7 +2208,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 		if (!low_end) {
 			RD::Uniform u;
 			u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
-			u.binding = 17;
+			u.binding = 14;
 			u.ids.push_back(sdfgi_get_ubo());
 			uniforms.push_back(u);
 		}
@@ -2225,7 +2217,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
 	}
 }
 
-RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps) {
+RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas) {
 	if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) {
 		RD::get_singleton()->free(render_pass_uniform_set);
 	}
@@ -2284,6 +2276,17 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 		RD::Uniform u;
 		u.binding = 3;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
+		if (p_use_directional_shadow_atlas && directional_shadow_get_texture().is_valid()) {
+			u.ids.push_back(directional_shadow_get_texture());
+		} else {
+			u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
+		}
+		uniforms.push_back(u);
+	}
+	{
+		RD::Uniform u;
+		u.binding = 4;
+		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		u.ids.resize(scene_state.max_lightmaps);
 		RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE);
 		for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) {
@@ -2301,7 +2304,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 	}
 	{
 		RD::Uniform u;
-		u.binding = 4;
+		u.binding = 5;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		u.ids.resize(MAX_GI_PROBES);
 		RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
@@ -2322,7 +2325,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 
 	{
 		RD::Uniform u;
-		u.binding = 5;
+		u.binding = 6;
 		u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 		RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : default_vec4_xform_buffer;
 		u.ids.push_back(cb);
@@ -2331,7 +2334,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 
 	{
 		RD::Uniform u;
-		u.binding = 6;
+		u.binding = 7;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE);
 		u.ids.push_back(texture);
@@ -2339,7 +2342,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 	}
 	{
 		RD::Uniform u;
-		u.binding = 7;
+		u.binding = 8;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_buffers) : RID();
 		RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
@@ -2350,7 +2353,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 	if (!low_end) {
 		{
 			RD::Uniform u;
-			u.binding = 8;
+			u.binding = 9;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID texture = rb && rb->normal_roughness_buffer.is_valid() ? rb->normal_roughness_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_NORMAL);
 			u.ids.push_back(texture);
@@ -2359,7 +2362,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 
 		{
 			RD::Uniform u;
-			u.binding = 9;
+			u.binding = 10;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID aot = rb ? render_buffers_get_ao_texture(p_render_buffers) : RID();
 			RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
@@ -2369,7 +2372,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 
 		{
 			RD::Uniform u;
-			u.binding = 10;
+			u.binding = 11;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID ambient_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_ambient_texture(p_render_buffers) : RID();
 			RID texture = ambient_buffer.is_valid() ? ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
@@ -2379,7 +2382,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 
 		{
 			RD::Uniform u;
-			u.binding = 11;
+			u.binding = 12;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID reflection_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_reflection_texture(p_render_buffers) : RID();
 			RID texture = reflection_buffer.is_valid() ? reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
@@ -2388,7 +2391,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 		}
 		{
 			RD::Uniform u;
-			u.binding = 12;
+			u.binding = 13;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID t;
 			if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) {
@@ -2401,7 +2404,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 		}
 		{
 			RD::Uniform u;
-			u.binding = 13;
+			u.binding = 14;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) {
 				u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers));
@@ -2412,14 +2415,14 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
 		}
 		{
 			RD::Uniform u;
-			u.binding = 14;
+			u.binding = 15;
 			u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
 			u.ids.push_back(rb ? render_buffers_get_gi_probe_buffer(p_render_buffers) : render_buffers_get_default_gi_probe_buffer());
 			uniforms.push_back(u);
 		}
 		{
 			RD::Uniform u;
-			u.binding = 15;
+			u.binding = 16;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			RID vfog = RID();
 			if (rb && render_buffers_has_volumetric_fog(p_render_buffers)) {
@@ -2477,10 +2480,20 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed
 	}
 
 	{
-		// No Lightmaps
+		// No directional shadow atlas.
 		RD::Uniform u;
 		u.binding = 3;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
+		RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE);
+		u.ids.push_back(texture);
+		uniforms.push_back(u);
+	}
+
+	{
+		// No Lightmaps
+		RD::Uniform u;
+		u.binding = 4;
+		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		u.ids.resize(scene_state.max_lightmaps);
 		RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE);
 		for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) {
@@ -2493,7 +2506,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed
 	{
 		// No GIProbes
 		RD::Uniform u;
-		u.binding = 4;
+		u.binding = 5;
 		u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 		u.ids.resize(MAX_GI_PROBES);
 		RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
@@ -2506,7 +2519,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed
 
 	{
 		RD::Uniform u;
-		u.binding = 5;
+		u.binding = 6;
 		u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
 		RID cb = default_vec4_xform_buffer;
 		u.ids.push_back(cb);
@@ -2518,28 +2531,28 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed
 	{
 		RD::Uniform u;
 		u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
-		u.binding = 6;
+		u.binding = 7;
 		u.ids.push_back(p_albedo_texture);
 		uniforms.push_back(u);
 	}
 	{
 		RD::Uniform u;
 		u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
-		u.binding = 7;
+		u.binding = 8;
 		u.ids.push_back(p_emission_texture);
 		uniforms.push_back(u);
 	}
 	{
 		RD::Uniform u;
 		u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
-		u.binding = 8;
+		u.binding = 9;
 		u.ids.push_back(p_emission_aniso_texture);
 		uniforms.push_back(u);
 	}
 	{
 		RD::Uniform u;
 		u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
-		u.binding = 9;
+		u.binding = 10;
 		u.ids.push_back(p_geom_facing_texture);
 		uniforms.push_back(u);
 	}