Merge commit godotengine/godot@8004c7524fb9f43425c4d6f614410a76678e0f7c

37 files changed, 655 insertions, 242 deletions

diff --git a/servers/rendering/dummy/rasterizer_dummy.h b/servers/rendering/dummy/rasterizer_dummy.h
index d7996a2ec1..11a019b563 100644
--- a/servers/rendering/dummy/rasterizer_dummy.h
+++ b/servers/rendering/dummy/rasterizer_dummy.h
@@ -68,14 +68,14 @@ protected:
 	RasterizerSceneDummy scene;
 
 public:
-	RendererUtilities *get_utilities() override { return &utilities; };
-	RendererLightStorage *get_light_storage() override { return &light_storage; };
-	RendererMaterialStorage *get_material_storage() override { return &material_storage; };
-	RendererMeshStorage *get_mesh_storage() override { return &mesh_storage; };
-	RendererParticlesStorage *get_particles_storage() override { return &particles_storage; };
-	RendererTextureStorage *get_texture_storage() override { return &texture_storage; };
-	RendererGI *get_gi() override { return &gi; };
-	RendererFog *get_fog() override { return &fog; };
+	RendererUtilities *get_utilities() override { return &utilities; }
+	RendererLightStorage *get_light_storage() override { return &light_storage; }
+	RendererMaterialStorage *get_material_storage() override { return &material_storage; }
+	RendererMeshStorage *get_mesh_storage() override { return &mesh_storage; }
+	RendererParticlesStorage *get_particles_storage() override { return &particles_storage; }
+	RendererTextureStorage *get_texture_storage() override { return &texture_storage; }
+	RendererGI *get_gi() override { return &gi; }
+	RendererFog *get_fog() override { return &fog; }
 	RendererCanvasRender *get_canvas() override { return &canvas; }
 	RendererSceneRender *get_scene() override { return &scene; }
 
diff --git a/servers/rendering/dummy/storage/material_storage.h b/servers/rendering/dummy/storage/material_storage.h
index 423544201c..f55ef24473 100644
--- a/servers/rendering/dummy/storage/material_storage.h
+++ b/servers/rendering/dummy/storage/material_storage.h
@@ -94,7 +94,7 @@ public:
 	virtual RID shader_get_default_texture_parameter(RID p_shader, const StringName &p_name, int p_index) const override { return RID(); }
 	virtual Variant shader_get_parameter_default(RID p_material, const StringName &p_param) const override { return Variant(); }
 
-	virtual RS::ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const override { return RS::ShaderNativeSourceCode(); };
+	virtual RS::ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const override { return RS::ShaderNativeSourceCode(); }
 
 	/* MATERIAL API */
 	virtual RID material_allocate() override { return RID(); }
diff --git a/servers/rendering/dummy/storage/mesh_storage.h b/servers/rendering/dummy/storage/mesh_storage.h
index 41381e2b2f..37959ffb7f 100644
--- a/servers/rendering/dummy/storage/mesh_storage.h
+++ b/servers/rendering/dummy/storage/mesh_storage.h
@@ -66,7 +66,7 @@ public:
 
 	/* MESH API */
 
-	bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); };
+	bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); }
 
 	virtual RID mesh_allocate() override;
 	virtual void mesh_initialize(RID p_rid) override;
diff --git a/servers/rendering/dummy/storage/texture_storage.h b/servers/rendering/dummy/storage/texture_storage.h
index e5167af9ed..4b61ac1722 100644
--- a/servers/rendering/dummy/storage/texture_storage.h
+++ b/servers/rendering/dummy/storage/texture_storage.h
@@ -55,7 +55,7 @@ public:
 
 	/* Canvas Texture API */
 
-	virtual RID canvas_texture_allocate() override { return RID(); };
+	virtual RID canvas_texture_allocate() override { return RID(); }
 	virtual void canvas_texture_initialize(RID p_rid) override {}
 	virtual void canvas_texture_free(RID p_rid) override {}
 
@@ -67,13 +67,13 @@ public:
 
 	/* Texture API */
 
-	bool owns_texture(RID p_rid) { return texture_owner.owns(p_rid); };
+	bool owns_texture(RID p_rid) { return texture_owner.owns(p_rid); }
 
 	virtual RID texture_allocate() override {
 		DummyTexture *texture = memnew(DummyTexture);
 		ERR_FAIL_NULL_V(texture, RID());
 		return texture_owner.make_rid(texture);
-	};
+	}
 
 	virtual void texture_free(RID p_rid) override {
 		// delete the texture
@@ -81,13 +81,13 @@ public:
 		ERR_FAIL_NULL(texture);
 		texture_owner.free(p_rid);
 		memdelete(texture);
-	};
+	}
 
 	virtual void texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) override {
 		DummyTexture *t = texture_owner.get_or_null(p_texture);
 		ERR_FAIL_NULL(t);
 		t->image = p_image->duplicate();
-	};
+	}
 	virtual void texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) override {}
 	virtual void texture_3d_initialize(RID p_texture, Image::Format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) override {}
 	virtual void texture_external_initialize(RID p_texture, int p_width, int p_height, uint64_t p_external_buffer) override {}
@@ -109,15 +109,15 @@ public:
 		DummyTexture *t = texture_owner.get_or_null(p_texture);
 		ERR_FAIL_NULL_V(t, Ref<Image>());
 		return t->image;
-	};
-	virtual Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const override { return Ref<Image>(); };
-	virtual Vector<Ref<Image>> texture_3d_get(RID p_texture) const override { return Vector<Ref<Image>>(); };
+	}
+	virtual Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const override { return Ref<Image>(); }
+	virtual Vector<Ref<Image>> texture_3d_get(RID p_texture) const override { return Vector<Ref<Image>>(); }
 
-	virtual void texture_replace(RID p_texture, RID p_by_texture) override { texture_free(p_by_texture); };
+	virtual void texture_replace(RID p_texture, RID p_by_texture) override { texture_free(p_by_texture); }
 	virtual void texture_set_size_override(RID p_texture, int p_width, int p_height) override {}
 
 	virtual void texture_set_path(RID p_texture, const String &p_path) override {}
-	virtual String texture_get_path(RID p_texture) const override { return String(); };
+	virtual String texture_get_path(RID p_texture) const override { return String(); }
 
 	virtual Image::Format texture_get_format(RID p_texture) const override { return Image::FORMAT_MAX; }
 
@@ -129,11 +129,11 @@ public:
 
 	virtual void texture_set_force_redraw_if_visible(RID p_texture, bool p_enable) override {}
 
-	virtual Size2 texture_size_with_proxy(RID p_proxy) override { return Size2(); };
+	virtual Size2 texture_size_with_proxy(RID p_proxy) override { return Size2(); }
 
 	virtual void texture_rd_initialize(RID p_texture, const RID &p_rd_texture, const RS::TextureLayeredType p_layer_type = RS::TEXTURE_LAYERED_2D_ARRAY) override {}
-	virtual RID texture_get_rd_texture(RID p_texture, bool p_srgb = false) const override { return RID(); };
-	virtual uint64_t texture_get_native_handle(RID p_texture, bool p_srgb = false) const override { return 0; };
+	virtual RID texture_get_rd_texture(RID p_texture, bool p_srgb = false) const override { return RID(); }
+	virtual uint64_t texture_get_native_handle(RID p_texture, bool p_srgb = false) const override { return 0; }
 
 	/* DECAL API */
 	virtual RID decal_allocate() override { return RID(); }
diff --git a/servers/rendering/dummy/storage/utilities.h b/servers/rendering/dummy/storage/utilities.h
index 009d3ca5ad..d4faad59f5 100644
--- a/servers/rendering/dummy/storage/utilities.h
+++ b/servers/rendering/dummy/storage/utilities.h
@@ -126,7 +126,7 @@ public:
 	virtual RenderingDevice::DeviceType get_video_adapter_type() const override { return RenderingDevice::DeviceType::DEVICE_TYPE_OTHER; }
 	virtual String get_video_adapter_api_version() const override { return String(); }
 
-	virtual Size2i get_maximum_viewport_size() const override { return Size2i(); };
+	virtual Size2i get_maximum_viewport_size() const override { return Size2i(); }
 };
 
 } // namespace RendererDummy
diff --git a/servers/rendering/renderer_compositor.h b/servers/rendering/renderer_compositor.h
index 75567a8f1a..ddfc29cd2c 100644
--- a/servers/rendering/renderer_compositor.h
+++ b/servers/rendering/renderer_compositor.h
@@ -108,7 +108,7 @@ public:
 	virtual double get_total_time() const = 0;
 	virtual bool can_create_resources_async() const = 0;
 
-	static bool is_low_end() { return low_end; };
+	static bool is_low_end() { return low_end; }
 	virtual bool is_xr_enabled() const;
 
 	static RendererCompositor *get_singleton() { return singleton; }
diff --git a/servers/rendering/renderer_rd/environment/fog.cpp b/servers/rendering/renderer_rd/environment/fog.cpp
index 4cdb38fd29..48300571d4 100644
--- a/servers/rendering/renderer_rd/environment/fog.cpp
+++ b/servers/rendering/renderer_rd/environment/fog.cpp
@@ -159,7 +159,7 @@ RendererRD::MaterialStorage::ShaderData *Fog::_create_fog_shader_func() {
 
 RendererRD::MaterialStorage::ShaderData *Fog::_create_fog_shader_funcs() {
 	return Fog::get_singleton()->_create_fog_shader_func();
-};
+}
 
 RendererRD::MaterialStorage::MaterialData *Fog::_create_fog_material_func(FogShaderData *p_shader) {
 	FogMaterialData *material_data = memnew(FogMaterialData);
@@ -170,7 +170,7 @@ RendererRD::MaterialStorage::MaterialData *Fog::_create_fog_material_func(FogSha
 
 RendererRD::MaterialStorage::MaterialData *Fog::_create_fog_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) {
 	return Fog::get_singleton()->_create_fog_material_func(static_cast<FogShaderData *>(p_shader));
-};
+}
 
 ////////////////////////////////////////////////////////////////////////////////
 // FOG VOLUMES INSTANCE
diff --git a/servers/rendering/renderer_rd/environment/fog.h b/servers/rendering/renderer_rd/environment/fog.h
index a05a48b064..390829962f 100644
--- a/servers/rendering/renderer_rd/environment/fog.h
+++ b/servers/rendering/renderer_rd/environment/fog.h
@@ -235,7 +235,7 @@ public:
 
 	/* FOG VOLUMES */
 
-	bool owns_fog_volume(RID p_rid) { return fog_volume_owner.owns(p_rid); };
+	bool owns_fog_volume(RID p_rid) { return fog_volume_owner.owns(p_rid); }
 
 	virtual RID fog_volume_allocate() override;
 	virtual void fog_volume_initialize(RID p_rid) override;
@@ -252,7 +252,7 @@ public:
 
 	/* FOG VOLUMES INSTANCE */
 
-	bool owns_fog_volume_instance(RID p_rid) { return fog_volume_instance_owner.owns(p_rid); };
+	bool owns_fog_volume_instance(RID p_rid) { return fog_volume_instance_owner.owns(p_rid); }
 
 	RID fog_volume_instance_create(RID p_fog_volume);
 	void fog_instance_free(RID p_rid);
diff --git a/servers/rendering/renderer_rd/environment/gi.h b/servers/rendering/renderer_rd/environment/gi.h
index 7c83f7b954..e3c5f4fea2 100644
--- a/servers/rendering/renderer_rd/environment/gi.h
+++ b/servers/rendering/renderer_rd/environment/gi.h
@@ -469,7 +469,7 @@ public:
 
 	/* VOXEL GI API */
 
-	bool owns_voxel_gi(RID p_rid) { return voxel_gi_owner.owns(p_rid); };
+	bool owns_voxel_gi(RID p_rid) { return voxel_gi_owner.owns(p_rid); }
 
 	virtual RID voxel_gi_allocate() override;
 	virtual void voxel_gi_free(RID p_voxel_gi) override;
@@ -526,14 +526,14 @@ public:
 		VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe);
 		ERR_FAIL_NULL_V(voxel_gi, RID());
 		return voxel_gi->texture;
-	};
+	}
 
 	_FORCE_INLINE_ void voxel_gi_instance_set_render_index(RID p_probe, uint32_t p_index) {
 		VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe);
 		ERR_FAIL_NULL(voxel_gi);
 
 		voxel_gi->render_index = p_index;
-	};
+	}
 
 	bool voxel_gi_instance_owns(RID p_rid) const {
 		return voxel_gi_instance_owner.owns(p_rid);
diff --git a/servers/rendering/renderer_rd/environment/sky.cpp b/servers/rendering/renderer_rd/environment/sky.cpp
index 9464d6e249..e71e3f6efb 100644
--- a/servers/rendering/renderer_rd/environment/sky.cpp
+++ b/servers/rendering/renderer_rd/environment/sky.cpp
@@ -703,7 +703,7 @@ RendererRD::MaterialStorage::ShaderData *SkyRD::_create_sky_shader_func() {
 RendererRD::MaterialStorage::ShaderData *SkyRD::_create_sky_shader_funcs() {
 	// !BAS! Why isn't _create_sky_shader_func not just static too?
 	return static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton)->sky._create_sky_shader_func();
-};
+}
 
 RendererRD::MaterialStorage::MaterialData *SkyRD::_create_sky_material_func(SkyShaderData *p_shader) {
 	SkyMaterialData *material_data = memnew(SkyMaterialData);
@@ -715,7 +715,7 @@ RendererRD::MaterialStorage::MaterialData *SkyRD::_create_sky_material_func(SkyS
 RendererRD::MaterialStorage::MaterialData *SkyRD::_create_sky_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) {
 	// !BAS! same here, we could just make _create_sky_material_func static?
 	return static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton)->sky._create_sky_material_func(static_cast<SkyShaderData *>(p_shader));
-};
+}
 
 SkyRD::SkyRD() {
 	roughness_layers = GLOBAL_GET("rendering/reflections/sky_reflections/roughness_layers");
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 b58ed162ef..014bdeba8d 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
@@ -110,10 +110,10 @@ public:
 				uint32_t projector_use_mipmaps : 1;
 				uint32_t use_depth_fog : 1;
 				uint32_t use_lightmap_bicubic_filter : 1;
-				uint32_t soft_shadow_samples : 4;
-				uint32_t penumbra_shadow_samples : 4;
-				uint32_t directional_soft_shadow_samples : 4;
-				uint32_t directional_penumbra_shadow_samples : 4;
+				uint32_t soft_shadow_samples : 6;
+				uint32_t penumbra_shadow_samples : 6;
+				uint32_t directional_soft_shadow_samples : 6;
+				uint32_t directional_penumbra_shadow_samples : 6;
 			};
 
 			uint32_t packed_0;
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 2de13593c3..71f071e44e 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
@@ -324,7 +324,12 @@ void SceneShaderForwardMobile::ShaderData::_create_pipeline(PipelineKey p_pipeli
 	specialization_constants.push_back(sc);
 
 	sc.constant_id = 1;
-	sc.float_value = p_pipeline_key.shader_specialization.packed_1;
+	sc.int_value = p_pipeline_key.shader_specialization.packed_1;
+	sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT;
+	specialization_constants.push_back(sc);
+
+	sc.constant_id = 2;
+	sc.float_value = p_pipeline_key.shader_specialization.packed_2;
 	sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_FLOAT;
 	specialization_constants.push_back(sc);
 
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 e338ef09df..60216e9b48 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
@@ -76,22 +76,23 @@ public:
 				uint32_t use_depth_fog : 1;
 				uint32_t is_multimesh : 1;
 				uint32_t use_lightmap_bicubic_filter : 1;
-				uint32_t pad : 2;
-				uint32_t soft_shadow_samples : 4;
-				uint32_t penumbra_shadow_samples : 4;
-				uint32_t directional_soft_shadow_samples : 4;
-				uint32_t directional_penumbra_shadow_samples : 4;
+				uint32_t soft_shadow_samples : 6;
+				uint32_t penumbra_shadow_samples : 6;
+				uint32_t directional_soft_shadow_samples : 6;
 			};
 
 			uint32_t packed_0;
 		};
 
 		union {
-			float luminance_multiplier;
-			float packed_1;
+			uint32_t directional_penumbra_shadow_samples : 6;
+			uint32_t packed_1;
 		};
 
-		uint32_t packed_2;
+		union {
+			float luminance_multiplier;
+			float packed_2;
+		};
 	};
 
 	struct UbershaderConstants {
diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
index 16b4a5792f..42c73c544a 100644
--- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
+++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
@@ -913,7 +913,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p
 		RenderingServerDefault::redraw_request();
 	}
 
-	state.current_data_buffer_index = (state.current_data_buffer_index + 1) % state.canvas_instance_data_buffers.size();
+	state.current_data_buffer_index = (state.current_data_buffer_index + 1) % BATCH_DATA_BUFFER_COUNT;
 	state.current_instance_buffer_index = 0;
 }
 
@@ -1649,6 +1649,9 @@ RendererCanvasRenderRD::RendererCanvasRenderRD() {
 		default_samplers.default_repeat = RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED;
 	}
 
+	// preallocate 5 slots for uniform set 3
+	state.batch_texture_uniforms.resize(5);
+
 	{ //shader variants
 
 		String global_defines;
@@ -1976,15 +1979,18 @@ void fragment() {
 	}
 
 	{
+		uint32_t cache_size = uint32_t(GLOBAL_GET("rendering/2d/batching/uniform_set_cache_size"));
+		rid_set_to_uniform_set.set_capacity(cache_size);
+	}
+
+	{
 		state.max_instances_per_buffer = uint32_t(GLOBAL_GET("rendering/2d/batching/item_buffer_size"));
 		state.max_instance_buffer_size = state.max_instances_per_buffer * sizeof(InstanceData);
-		state.canvas_instance_data_buffers.resize(3);
 		state.canvas_instance_batches.reserve(200);
 
-		for (int i = 0; i < 3; i++) {
-			DataBuffer db;
+		for (uint32_t i = 0; i < BATCH_DATA_BUFFER_COUNT; i++) {
+			DataBuffer &db = state.canvas_instance_data_buffers[i];
 			db.instance_buffers.push_back(RD::get_singleton()->storage_buffer_create(state.max_instance_buffer_size));
-			state.canvas_instance_data_buffers[i] = db;
 		}
 		state.instance_data_array = memnew_arr(InstanceData, state.max_instances_per_buffer);
 	}
@@ -2161,7 +2167,7 @@ void RendererCanvasRenderRD::_render_batch_items(RenderTarget p_to_render_target
 
 	RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer);
 
-	RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, clear ? RD::INITIAL_ACTION_CLEAR : RD::INITIAL_ACTION_LOAD, RD::FINAL_ACTION_STORE, RD::INITIAL_ACTION_LOAD, RD::FINAL_ACTION_DISCARD, clear_colors);
+	RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, clear ? RD::INITIAL_ACTION_CLEAR : RD::INITIAL_ACTION_LOAD, RD::FINAL_ACTION_STORE, RD::INITIAL_ACTION_LOAD, RD::FINAL_ACTION_DISCARD, clear_colors, 1, 0, Rect2(), RDD::BreadcrumbMarker::UI_PASS);
 
 	RD::get_singleton()->draw_list_bind_uniform_set(draw_list, fb_uniform_set, BASE_UNIFORM_SET);
 	RD::get_singleton()->draw_list_bind_uniform_set(draw_list, state.default_transforms_uniform_set, TRANSFORMS_UNIFORM_SET);
@@ -2205,14 +2211,47 @@ void RendererCanvasRenderRD::_render_batch_items(RenderTarget p_to_render_target
 
 	RD::get_singleton()->draw_list_end();
 
+	texture_info_map.clear();
 	state.current_batch_index = 0;
 	state.canvas_instance_batches.clear();
 	state.last_instance_index += instance_index;
 }
 
+RendererCanvasRenderRD::InstanceData *RendererCanvasRenderRD::new_instance_data(float *p_world, uint32_t *p_lights, uint32_t p_base_flags, uint32_t p_index, TextureInfo *p_info) {
+	InstanceData *instance_data = &state.instance_data_array[p_index];
+	// Zero out most fields.
+	for (int i = 0; i < 4; i++) {
+		instance_data->modulation[i] = 0.0;
+		instance_data->ninepatch_margins[i] = 0.0;
+		instance_data->src_rect[i] = 0.0;
+		instance_data->dst_rect[i] = 0.0;
+	}
+
+	instance_data->pad[0] = 0.0;
+	instance_data->pad[1] = 0.0;
+
+	instance_data->lights[0] = p_lights[0];
+	instance_data->lights[1] = p_lights[1];
+	instance_data->lights[2] = p_lights[2];
+	instance_data->lights[3] = p_lights[3];
+
+	for (int i = 0; i < 6; i++) {
+		instance_data->world[i] = p_world[i];
+	}
+
+	instance_data->flags = p_base_flags | p_info->flags; // Reset on each command for safety, keep canvas texture binding config.
+
+	instance_data->color_texture_pixel_size[0] = p_info->texpixel_size.width;
+	instance_data->color_texture_pixel_size[1] = p_info->texpixel_size.height;
+
+	instance_data->pad1 = 0;
+
+	return instance_data;
+}
+
 void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTarget p_render_target, const Transform2D &p_base_transform, Item *&r_current_clip, Light *p_lights, uint32_t &r_index, bool &r_batch_broken, bool &r_sdf_used, Batch *&r_current_batch) {
-	RenderingServer::CanvasItemTextureFilter texture_filter = p_item->texture_filter == RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT ? default_filter : p_item->texture_filter;
-	RenderingServer::CanvasItemTextureRepeat texture_repeat = p_item->texture_repeat == RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT ? default_repeat : p_item->texture_repeat;
+	const RenderingServer::CanvasItemTextureFilter texture_filter = p_item->texture_filter == RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT ? default_filter : p_item->texture_filter;
+	const RenderingServer::CanvasItemTextureRepeat texture_repeat = p_item->texture_repeat == RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT ? default_repeat : p_item->texture_repeat;
 
 	Transform2D base_transform = p_base_transform;
 
@@ -2243,7 +2282,7 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 
 				light_count++;
 
-				if (light_count == state.max_lights_per_item - 1) {
+				if (light_count == MAX_LIGHTS_PER_ITEM - 1) {
 					break;
 				}
 			}
@@ -2260,38 +2299,6 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 		r_current_batch->use_lighting = use_lighting;
 	}
 
-	// new_instance_data should be called after the current_batch is set.
-	auto new_instance_data = [&]() -> InstanceData * {
-		InstanceData *instance_data = &state.instance_data_array[r_index];
-		// Zero out most fields.
-		for (int i = 0; i < 4; i++) {
-			instance_data->modulation[i] = 0.0;
-			instance_data->ninepatch_margins[i] = 0.0;
-			instance_data->src_rect[i] = 0.0;
-			instance_data->dst_rect[i] = 0.0;
-		}
-
-		instance_data->pad[0] = 0.0;
-		instance_data->pad[1] = 0.0;
-
-		instance_data->lights[0] = lights[0];
-		instance_data->lights[1] = lights[1];
-		instance_data->lights[2] = lights[2];
-		instance_data->lights[3] = lights[3];
-
-		for (int i = 0; i < 6; i++) {
-			instance_data->world[i] = world[i];
-		}
-
-		instance_data->flags = base_flags | r_current_batch->tex_info.flags; // Reset on each command for safety, keep canvas texture binding config.
-
-		instance_data->color_texture_pixel_size[0] = r_current_batch->tex_info.texpixel_size.width;
-		instance_data->color_texture_pixel_size[1] = r_current_batch->tex_info.texpixel_size.height;
-		instance_data->specular_shininess = r_current_batch->tex_info.specular_shininess;
-
-		return instance_data;
-	};
-
 	const Item::Command *c = p_item->commands;
 	while (c) {
 		if (skipping && c->type != Item::Command::TYPE_ANIMATION_SLICE) {
@@ -2313,17 +2320,9 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 					r_current_batch->render_primitive = RD::RENDER_PRIMITIVE_TRIANGLES;
 				}
 
+				RenderingServer::CanvasItemTextureRepeat rect_repeat = texture_repeat;
 				if (bool(rect->flags & CANVAS_RECT_TILE)) {
-					texture_repeat = RenderingServer::CanvasItemTextureRepeat::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED;
-				}
-
-				bool has_msdf = bool(rect->flags & CANVAS_RECT_MSDF);
-				TextureState tex_state(rect->texture, texture_filter, texture_repeat, has_msdf, use_linear_colors);
-
-				if (tex_state != r_current_batch->tex_info.state) {
-					r_current_batch = _new_batch(r_batch_broken);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, rect->texture);
+					rect_repeat = RenderingServer::CanvasItemTextureRepeat::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED;
 				}
 
 				Color modulated = rect->modulate * base_color;
@@ -2342,12 +2341,25 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 					r_current_batch->render_primitive = RD::RENDER_PRIMITIVE_TRIANGLES;
 				}
 
-				InstanceData *instance_data = new_instance_data();
+				bool has_msdf = bool(rect->flags & CANVAS_RECT_MSDF);
+				TextureState tex_state(rect->texture, texture_filter, rect_repeat, has_msdf, use_linear_colors);
+				TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+				if (!tex_info) {
+					tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+					_prepare_batch_texture_info(rect->texture, tex_state, tex_info);
+				}
+
+				if (r_current_batch->tex_info != tex_info) {
+					r_current_batch = _new_batch(r_batch_broken);
+					r_current_batch->tex_info = tex_info;
+				}
+
+				InstanceData *instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 				Rect2 src_rect;
 				Rect2 dst_rect;
 
 				if (rect->texture.is_valid()) {
-					src_rect = (rect->flags & CANVAS_RECT_REGION) ? Rect2(rect->source.position * r_current_batch->tex_info.texpixel_size, rect->source.size * r_current_batch->tex_info.texpixel_size) : Rect2(0, 0, 1, 1);
+					src_rect = (rect->flags & CANVAS_RECT_REGION) ? Rect2(rect->source.position * tex_info->texpixel_size, rect->source.size * tex_info->texpixel_size) : Rect2(0, 0, 1, 1);
 					dst_rect = Rect2(rect->rect.position, rect->rect.size);
 
 					if (dst_rect.size.width < 0) {
@@ -2433,13 +2445,18 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 				}
 
 				TextureState tex_state(np->texture, texture_filter, texture_repeat, false, use_linear_colors);
-				if (tex_state != r_current_batch->tex_info.state) {
+				TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+				if (!tex_info) {
+					tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+					_prepare_batch_texture_info(np->texture, tex_state, tex_info);
+				}
+
+				if (r_current_batch->tex_info != tex_info) {
 					r_current_batch = _new_batch(r_batch_broken);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, np->texture);
+					r_current_batch->tex_info = tex_info;
 				}
 
-				InstanceData *instance_data = new_instance_data();
+				InstanceData *instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 				Rect2 src_rect;
 				Rect2 dst_rect(np->rect.position.x, np->rect.position.y, np->rect.size.x, np->rect.size.y);
@@ -2448,7 +2465,7 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 					src_rect = Rect2(0, 0, 1, 1);
 				} else {
 					if (np->source != Rect2()) {
-						src_rect = Rect2(np->source.position.x * r_current_batch->tex_info.texpixel_size.width, np->source.position.y * r_current_batch->tex_info.texpixel_size.height, np->source.size.x * r_current_batch->tex_info.texpixel_size.width, np->source.size.y * r_current_batch->tex_info.texpixel_size.height);
+						src_rect = Rect2(np->source.position.x * tex_info->texpixel_size.width, np->source.position.y * tex_info->texpixel_size.height, np->source.size.x * tex_info->texpixel_size.width, np->source.size.y * tex_info->texpixel_size.height);
 						instance_data->color_texture_pixel_size[0] = 1.0 / np->source.size.width;
 						instance_data->color_texture_pixel_size[1] = 1.0 / np->source.size.height;
 					} else {
@@ -2502,10 +2519,15 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 				r_current_batch->command = c;
 
 				TextureState tex_state(polygon->texture, texture_filter, texture_repeat, false, use_linear_colors);
-				if (tex_state != r_current_batch->tex_info.state) {
+				TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+				if (!tex_info) {
+					tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+					_prepare_batch_texture_info(polygon->texture, tex_state, tex_info);
+				}
+
+				if (r_current_batch->tex_info != tex_info) {
 					r_current_batch = _new_batch(r_batch_broken);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, polygon->texture);
+					r_current_batch->tex_info = tex_info;
 				}
 
 				// pipeline variant
@@ -2515,7 +2537,7 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 					r_current_batch->render_primitive = _primitive_type_to_render_primitive(polygon->primitive);
 				}
 
-				InstanceData *instance_data = new_instance_data();
+				InstanceData *instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 				Color color = base_color;
 				if (use_linear_colors) {
@@ -2559,17 +2581,22 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 						default:
 							// Unknown point count.
 							break;
-					};
-
-					TextureState tex_state(primitive->texture, texture_filter, texture_repeat, false, use_linear_colors);
-					if (tex_state != r_current_batch->tex_info.state) {
-						r_current_batch = _new_batch(r_batch_broken);
-						r_current_batch->tex_info.state = tex_state;
-						_prepare_batch_texture_info(r_current_batch, primitive->texture);
 					}
 				}
 
-				InstanceData *instance_data = new_instance_data();
+				TextureState tex_state(primitive->texture, texture_filter, texture_repeat, false, use_linear_colors);
+				TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+				if (!tex_info) {
+					tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+					_prepare_batch_texture_info(primitive->texture, tex_state, tex_info);
+				}
+
+				if (r_current_batch->tex_info != tex_info) {
+					r_current_batch = _new_batch(r_batch_broken);
+					r_current_batch->tex_info = tex_info;
+				}
+
+				InstanceData *instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 				for (uint32_t j = 0; j < MIN(3u, primitive->point_count); j++) {
 					instance_data->points[j * 2 + 0] = primitive->points[j].x;
@@ -2587,7 +2614,7 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 				_add_to_batch(r_index, r_batch_broken, r_current_batch);
 
 				if (primitive->point_count == 4) {
-					instance_data = new_instance_data();
+					instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 					for (uint32_t j = 0; j < 3; j++) {
 						int offset = j == 0 ? 0 : 1;
@@ -2596,7 +2623,7 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 						instance_data->points[j * 2 + 1] = primitive->points[j + offset].y;
 						instance_data->uvs[j * 2 + 0] = primitive->uvs[j + offset].x;
 						instance_data->uvs[j * 2 + 1] = primitive->uvs[j + offset].y;
-						Color col = primitive->colors[j] * base_color;
+						Color col = primitive->colors[j + offset] * base_color;
 						if (use_linear_colors) {
 							col = col.srgb_to_linear();
 						}
@@ -2623,9 +2650,13 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 				if (c->type == Item::Command::TYPE_MESH) {
 					const Item::CommandMesh *m = static_cast<const Item::CommandMesh *>(c);
 					TextureState tex_state(m->texture, texture_filter, texture_repeat, false, use_linear_colors);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, m->texture);
-					instance_data = new_instance_data();
+					TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+					if (!tex_info) {
+						tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+						_prepare_batch_texture_info(m->texture, tex_state, tex_info);
+					}
+					r_current_batch->tex_info = tex_info;
+					instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 					r_current_batch->mesh_instance_count = 1;
 					_update_transform_2d_to_mat2x3(base_transform * draw_transform * m->transform, instance_data->world);
@@ -2646,9 +2677,13 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 					}
 
 					TextureState tex_state(mm->texture, texture_filter, texture_repeat, false, use_linear_colors);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, mm->texture);
-					instance_data = new_instance_data();
+					TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+					if (!tex_info) {
+						tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+						_prepare_batch_texture_info(mm->texture, tex_state, tex_info);
+					}
+					r_current_batch->tex_info = tex_info;
+					instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 					instance_data->flags |= 1; // multimesh, trails disabled
 
@@ -2664,10 +2699,13 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 
 					const Item::CommandParticles *pt = static_cast<const Item::CommandParticles *>(c);
 					TextureState tex_state(pt->texture, texture_filter, texture_repeat, false, use_linear_colors);
-					r_current_batch->tex_info.state = tex_state;
-					_prepare_batch_texture_info(r_current_batch, pt->texture);
-
-					instance_data = new_instance_data();
+					TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+					if (!tex_info) {
+						tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+						_prepare_batch_texture_info(pt->texture, tex_state, tex_info);
+					}
+					r_current_batch->tex_info = tex_info;
+					instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
 
 					uint32_t divisor = 1;
 					r_current_batch->mesh_instance_count = particles_storage->particles_get_amount(pt->particles, divisor);
@@ -2749,36 +2787,137 @@ void RendererCanvasRenderRD::_record_item_commands(const Item *p_item, RenderTar
 		r_batch_broken = false;
 	}
 
+#ifdef DEBUG_ENABLED
+	if (debug_redraw && p_item->debug_redraw_time > 0.0) {
+		Color dc = debug_redraw_color;
+		dc.a *= p_item->debug_redraw_time / debug_redraw_time;
+
+		// 1: If commands are different, start a new batch.
+		if (r_current_batch->command_type != Item::Command::TYPE_RECT) {
+			r_current_batch = _new_batch(r_batch_broken);
+			r_current_batch->command_type = Item::Command::TYPE_RECT;
+			// it is ok to be null for a TYPE_RECT
+			r_current_batch->command = nullptr;
+			// default variant
+			r_current_batch->shader_variant = SHADER_VARIANT_QUAD;
+			r_current_batch->render_primitive = RD::RENDER_PRIMITIVE_TRIANGLES;
+		}
+
+		// 2: If the current batch has lighting, start a new batch.
+		if (r_current_batch->use_lighting) {
+			r_current_batch = _new_batch(r_batch_broken);
+			r_current_batch->use_lighting = false;
+		}
+
+		// 3: If the current batch has blend, start a new batch.
+		if (r_current_batch->has_blend) {
+			r_current_batch = _new_batch(r_batch_broken);
+			r_current_batch->has_blend = false;
+		}
+
+		TextureState tex_state(default_canvas_texture, texture_filter, texture_repeat, false, use_linear_colors);
+		TextureInfo *tex_info = texture_info_map.getptr(tex_state);
+		if (!tex_info) {
+			tex_info = &texture_info_map.insert(tex_state, TextureInfo())->value;
+			_prepare_batch_texture_info(default_canvas_texture, tex_state, tex_info);
+		}
+
+		if (r_current_batch->tex_info != tex_info) {
+			r_current_batch = _new_batch(r_batch_broken);
+			r_current_batch->tex_info = tex_info;
+		}
+
+		InstanceData *instance_data = new_instance_data(world, lights, base_flags, r_index, tex_info);
+
+		Rect2 src_rect;
+		Rect2 dst_rect;
+
+		dst_rect = Rect2(Vector2(), p_item->rect.size);
+		if (dst_rect.size.width < 0) {
+			dst_rect.position.x += dst_rect.size.width;
+			dst_rect.size.width *= -1;
+		}
+		if (dst_rect.size.height < 0) {
+			dst_rect.position.y += dst_rect.size.height;
+			dst_rect.size.height *= -1;
+		}
+
+		src_rect = Rect2(0, 0, 1, 1);
+
+		instance_data->modulation[0] = dc.r;
+		instance_data->modulation[1] = dc.g;
+		instance_data->modulation[2] = dc.b;
+		instance_data->modulation[3] = dc.a;
+
+		instance_data->src_rect[0] = src_rect.position.x;
+		instance_data->src_rect[1] = src_rect.position.y;
+		instance_data->src_rect[2] = src_rect.size.width;
+		instance_data->src_rect[3] = src_rect.size.height;
+
+		instance_data->dst_rect[0] = dst_rect.position.x;
+		instance_data->dst_rect[1] = dst_rect.position.y;
+		instance_data->dst_rect[2] = dst_rect.size.width;
+		instance_data->dst_rect[3] = dst_rect.size.height;
+
+		_add_to_batch(r_index, r_batch_broken, r_current_batch);
+
+		p_item->debug_redraw_time -= RSG::rasterizer->get_frame_delta_time();
+
+		RenderingServerDefault::redraw_request();
+
+		r_batch_broken = false;
+	}
+#endif
+
 	if (r_current_clip && reclip) {
 		// will make it re-enable clipping if needed afterwards
 		r_current_clip = nullptr;
 	}
 }
 
-void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasShaderData *p_shader_data, RenderingDevice::FramebufferFormatID p_framebuffer_format, Light *p_lights, Batch const *p_batch, RenderingMethod::RenderInfo *r_render_info) {
-	UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton();
-	ERR_FAIL_NULL(uniform_set_cache);
+void RendererCanvasRenderRD::_before_evict(RendererCanvasRenderRD::RIDSetKey &p_key, RID &p_rid) {
+	RD::get_singleton()->uniform_set_set_invalidation_callback(p_rid, nullptr, nullptr);
+	RD::get_singleton()->free(p_rid);
+}
 
-	ERR_FAIL_NULL(p_batch->command);
+void RendererCanvasRenderRD::_uniform_set_invalidation_callback(void *p_userdata) {
+	const RIDSetKey *key = static_cast<RIDSetKey *>(p_userdata);
+	static_cast<RendererCanvasRenderRD *>(singleton)->rid_set_to_uniform_set.erase(*key);
+}
 
+void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasShaderData *p_shader_data, RenderingDevice::FramebufferFormatID p_framebuffer_format, Light *p_lights, Batch const *p_batch, RenderingMethod::RenderInfo *r_render_info) {
 	{
-		RD::Uniform u_diffuse(RD::UNIFORM_TYPE_TEXTURE, 0, p_batch->tex_info.diffuse);
-		RD::Uniform u_normal(RD::UNIFORM_TYPE_TEXTURE, 1, p_batch->tex_info.normal);
-		RD::Uniform u_specular(RD::UNIFORM_TYPE_TEXTURE, 2, p_batch->tex_info.specular);
-		RD::Uniform u_sampler(RD::UNIFORM_TYPE_SAMPLER, 3, p_batch->tex_info.sampler);
-		RD::Uniform u_instance_data(RD::UNIFORM_TYPE_STORAGE_BUFFER, 4, state.canvas_instance_data_buffers[state.current_data_buffer_index].instance_buffers[p_batch->instance_buffer_index]);
-
-		RID uniform_set = uniform_set_cache->get_cache(shader.default_version_rd_shader, BATCH_UNIFORM_SET, u_diffuse, u_normal, u_specular, u_sampler, u_instance_data);
+		RIDSetKey key(
+				p_batch->tex_info->state,
+				state.canvas_instance_data_buffers[state.current_data_buffer_index].instance_buffers[p_batch->instance_buffer_index]);
+
+		const RID *uniform_set = rid_set_to_uniform_set.getptr(key);
+		if (uniform_set == nullptr) {
+			state.batch_texture_uniforms.write[0] = RD::Uniform(RD::UNIFORM_TYPE_TEXTURE, 0, p_batch->tex_info->diffuse);
+			state.batch_texture_uniforms.write[1] = RD::Uniform(RD::UNIFORM_TYPE_TEXTURE, 1, p_batch->tex_info->normal);
+			state.batch_texture_uniforms.write[2] = RD::Uniform(RD::UNIFORM_TYPE_TEXTURE, 2, p_batch->tex_info->specular);
+			state.batch_texture_uniforms.write[3] = RD::Uniform(RD::UNIFORM_TYPE_SAMPLER, 3, p_batch->tex_info->sampler);
+			state.batch_texture_uniforms.write[4] = RD::Uniform(RD::UNIFORM_TYPE_STORAGE_BUFFER, 4, state.canvas_instance_data_buffers[state.current_data_buffer_index].instance_buffers[p_batch->instance_buffer_index]);
+
+			RID rid = RD::get_singleton()->uniform_set_create(state.batch_texture_uniforms, shader.default_version_rd_shader, BATCH_UNIFORM_SET);
+			ERR_FAIL_COND_MSG(rid.is_null(), "Failed to create uniform set for batch.");
+
+			const RIDCache::Pair *iter = rid_set_to_uniform_set.insert(key, rid);
+			uniform_set = &iter->data;
+			RD::get_singleton()->uniform_set_set_invalidation_callback(rid, RendererCanvasRenderRD::_uniform_set_invalidation_callback, (void *)&iter->key);
+		}
 
-		if (state.current_batch_uniform_set != uniform_set) {
-			state.current_batch_uniform_set = uniform_set;
-			RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, BATCH_UNIFORM_SET);
+		if (state.current_batch_uniform_set != *uniform_set) {
+			state.current_batch_uniform_set = *uniform_set;
+			RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, *uniform_set, BATCH_UNIFORM_SET);
 		}
 	}
+	PushConstant push_constant;
+	push_constant.base_instance_index = p_batch->start;
+	push_constant.specular_shininess = p_batch->tex_info->specular_shininess;
 
 	RID pipeline;
 	PipelineKey pipeline_key;
-	PushConstant push_constant;
 	pipeline_key.framebuffer_format_id = p_framebuffer_format;
 	pipeline_key.variant = p_batch->shader_variant;
 	pipeline_key.render_primitive = p_batch->render_primitive;
@@ -2794,7 +2933,6 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 				RD::get_singleton()->draw_list_set_blend_constants(p_draw_list, p_batch->modulate);
 			}
 
-			push_constant.base_instance_index = p_batch->start;
 			RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant));
 			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, p_batch->instance_count);
@@ -2807,6 +2945,8 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 		} break;
 
 		case Item::Command::TYPE_POLYGON: {
+			ERR_FAIL_NULL(p_batch->command);
+
 			const Item::CommandPolygon *polygon = static_cast<const Item::CommandPolygon *>(p_batch->command);
 
 			PolygonBuffers *pb = polygon_buffers.polygons.getptr(polygon->polygon.polygon_id);
@@ -2816,7 +2956,6 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 			pipeline = _get_pipeline_specialization_or_ubershader(p_shader_data, pipeline_key, push_constant);
 			RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline);
 
-			push_constant.base_instance_index = p_batch->start;
 			RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant));
 			RD::get_singleton()->draw_list_bind_vertex_array(p_draw_list, pb->vertex_array);
 			if (pb->indices.is_valid()) {
@@ -2832,12 +2971,13 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 		} break;
 
 		case Item::Command::TYPE_PRIMITIVE: {
+			ERR_FAIL_NULL(p_batch->command);
+
 			const Item::CommandPrimitive *primitive = static_cast<const Item::CommandPrimitive *>(p_batch->command);
 
 			pipeline = _get_pipeline_specialization_or_ubershader(p_shader_data, pipeline_key, push_constant);
 			RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline);
 
-			push_constant.base_instance_index = p_batch->start;
 			RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant));
 			RD::get_singleton()->draw_list_bind_index_array(p_draw_list, primitive_arrays.index_array[MIN(3u, primitive->point_count) - 1]);
 			uint32_t instance_count = p_batch->instance_count;
@@ -2854,6 +2994,8 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 		case Item::Command::TYPE_MESH:
 		case Item::Command::TYPE_MULTIMESH:
 		case Item::Command::TYPE_PARTICLES: {
+			ERR_FAIL_NULL(p_batch->command);
+
 			RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton();
 			RendererRD::ParticlesStorage *particles_storage = RendererRD::ParticlesStorage::get_singleton();
 
@@ -2914,7 +3056,6 @@ void RendererCanvasRenderRD::_render_batch(RD::DrawListID p_draw_list, CanvasSha
 				pipeline = _get_pipeline_specialization_or_ubershader(p_shader_data, pipeline_key, push_constant, mesh_instance, surface, j, &vertex_array);
 				RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline);
 
-				push_constant.base_instance_index = p_batch->start;
 				RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant));
 
 				RID index_array = mesh_storage->mesh_surface_get_index_array(surface, 0);
@@ -2996,7 +3137,7 @@ void RendererCanvasRenderRD::_allocate_instance_buffer() {
 	state.canvas_instance_data_buffers[state.current_data_buffer_index].instance_buffers.push_back(buf);
 }
 
-void RendererCanvasRenderRD::_prepare_batch_texture_info(Batch *p_current_batch, RID p_texture) const {
+void RendererCanvasRenderRD::_prepare_batch_texture_info(RID p_texture, TextureState &p_state, TextureInfo *p_info) {
 	if (p_texture.is_null()) {
 		p_texture = default_canvas_texture;
 	}
@@ -3004,38 +3145,38 @@ void RendererCanvasRenderRD::_prepare_batch_texture_info(Batch *p_current_batch,
 	RendererRD::TextureStorage::CanvasTextureInfo info =
 			RendererRD::TextureStorage::get_singleton()->canvas_texture_get_info(
 					p_texture,
-					p_current_batch->tex_info.state.texture_filter(),
-					p_current_batch->tex_info.state.texture_repeat(),
-					p_current_batch->tex_info.state.linear_colors(),
-					p_current_batch->tex_info.state.texture_is_data());
-
+					p_state.texture_filter(),
+					p_state.texture_repeat(),
+					p_state.linear_colors(),
+					p_state.texture_is_data());
 	// something odd happened
 	if (info.is_null()) {
-		_prepare_batch_texture_info(p_current_batch, default_canvas_texture);
+		_prepare_batch_texture_info(default_canvas_texture, p_state, p_info);
 		return;
 	}
 
-	p_current_batch->tex_info.diffuse = info.diffuse;
-	p_current_batch->tex_info.normal = info.normal;
-	p_current_batch->tex_info.specular = info.specular;
-	p_current_batch->tex_info.sampler = info.sampler;
+	p_info->state = p_state;
+	p_info->diffuse = info.diffuse;
+	p_info->normal = info.normal;
+	p_info->specular = info.specular;
+	p_info->sampler = info.sampler;
 
 	// cache values to be copied to instance data
 	if (info.specular_color.a < 0.999) {
-		p_current_batch->tex_info.flags |= FLAGS_DEFAULT_SPECULAR_MAP_USED;
+		p_info->flags |= FLAGS_DEFAULT_SPECULAR_MAP_USED;
 	}
 
 	if (info.use_normal) {
-		p_current_batch->tex_info.flags |= FLAGS_DEFAULT_NORMAL_MAP_USED;
+		p_info->flags |= FLAGS_DEFAULT_NORMAL_MAP_USED;
 	}
 
 	uint8_t a = uint8_t(CLAMP(info.specular_color.a * 255.0, 0.0, 255.0));
 	uint8_t b = uint8_t(CLAMP(info.specular_color.b * 255.0, 0.0, 255.0));
 	uint8_t g = uint8_t(CLAMP(info.specular_color.g * 255.0, 0.0, 255.0));
 	uint8_t r = uint8_t(CLAMP(info.specular_color.r * 255.0, 0.0, 255.0));
-	p_current_batch->tex_info.specular_shininess = uint32_t(a) << 24 | uint32_t(b) << 16 | uint32_t(g) << 8 | uint32_t(r);
+	p_info->specular_shininess = uint32_t(a) << 24 | uint32_t(b) << 16 | uint32_t(g) << 8 | uint32_t(r);
 
-	p_current_batch->tex_info.texpixel_size = Vector2(1.0 / float(info.size.width), 1.0 / float(info.size.height));
+	p_info->texpixel_size = Vector2(1.0 / float(info.size.width), 1.0 / float(info.size.height));
 }
 
 RendererCanvasRenderRD::~RendererCanvasRenderRD() {
@@ -3077,7 +3218,7 @@ RendererCanvasRenderRD::~RendererCanvasRenderRD() {
 	RD::get_singleton()->free(state.shadow_texture);
 
 	memdelete_arr(state.instance_data_array);
-	for (uint32_t i = 0; i < state.canvas_instance_data_buffers.size(); i++) {
+	for (uint32_t i = 0; i < BATCH_DATA_BUFFER_COUNT; i++) {
 		for (uint32_t j = 0; j < state.canvas_instance_data_buffers[i].instance_buffers.size(); j++) {
 			RD::get_singleton()->free(state.canvas_instance_data_buffers[i].instance_buffers[j]);
 		}
diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h
index 8ef1221108..a2c39d7181 100644
--- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h
+++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h
@@ -33,6 +33,7 @@
 #ifndef RENDERER_CANVAS_RENDER_RD_H
 #define RENDERER_CANVAS_RENDER_RD_H
 
+#include "core/templates/lru.h"
 #include "servers/rendering/renderer_canvas_render.h"
 #include "servers/rendering/renderer_compositor.h"
 #include "servers/rendering/renderer_rd/pipeline_cache_rd.h"
@@ -52,6 +53,8 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 	};
 
 	const int SAMPLERS_BINDING_FIRST_INDEX = 10;
+	// The size of the ring buffer to store GPU buffers. Triple-buffering the max expected frames in flight.
+	static const uint32_t BATCH_DATA_BUFFER_COUNT = 3;
 
 	enum ShaderVariant {
 		SHADER_VARIANT_QUAD,
@@ -341,7 +344,7 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 	struct InstanceData {
 		float world[6];
 		uint32_t flags;
-		uint32_t specular_shininess;
+		uint32_t pad1;
 		union {
 			//rect
 			struct {
@@ -368,8 +371,8 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 	struct PushConstant {
 		uint32_t base_instance_index;
 		ShaderSpecialization shader_specialization;
-		uint32_t pad2;
-		uint32_t pad3;
+		uint32_t specular_shininess;
+		uint32_t pad;
 	};
 
 	// TextureState is used to determine when a new batch is required due to a change of texture state.
@@ -400,50 +403,91 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 					(((uint32_t)p_use_linear_colors & LINEAR_COLORS_MASK) << LINEAR_COLORS_SHIFT);
 		}
 
-		_FORCE_INLINE_ RS::CanvasItemTextureFilter texture_filter() const {
+		_ALWAYS_INLINE_ RS::CanvasItemTextureFilter texture_filter() const {
 			return (RS::CanvasItemTextureFilter)((other >> FILTER_SHIFT) & FILTER_MASK);
 		}
 
-		_FORCE_INLINE_ RS::CanvasItemTextureRepeat texture_repeat() const {
+		_ALWAYS_INLINE_ RS::CanvasItemTextureRepeat texture_repeat() const {
 			return (RS::CanvasItemTextureRepeat)((other >> REPEAT_SHIFT) & REPEAT_MASK);
 		}
 
-		_FORCE_INLINE_ bool linear_colors() const {
+		_ALWAYS_INLINE_ bool linear_colors() const {
 			return (other >> LINEAR_COLORS_SHIFT) & LINEAR_COLORS_MASK;
 		}
 
-		_FORCE_INLINE_ bool texture_is_data() const {
+		_ALWAYS_INLINE_ bool texture_is_data() const {
 			return (other >> TEXTURE_IS_DATA_SHIFT) & TEXTURE_IS_DATA_MASK;
 		}
 
-		bool operator==(const TextureState &p_val) const {
+		_ALWAYS_INLINE_ bool operator==(const TextureState &p_val) const {
 			return (texture == p_val.texture) && (other == p_val.other);
 		}
 
-		bool operator!=(const TextureState &p_val) const {
+		_ALWAYS_INLINE_ bool operator!=(const TextureState &p_val) const {
 			return (texture != p_val.texture) || (other != p_val.other);
 		}
+
+		_ALWAYS_INLINE_ bool is_valid() const { return texture.is_valid(); }
+		_ALWAYS_INLINE_ bool is_null() const { return texture.is_null(); }
+
+		uint32_t hash() const {
+			uint32_t hash = hash_murmur3_one_64(texture.get_id());
+			return hash_murmur3_one_32(other, hash);
+		}
 	};
 
 	struct TextureInfo {
 		TextureState state;
-		uint32_t specular_shininess = 0;
-		uint32_t flags = 0;
-		Vector2 texpixel_size;
-
 		RID diffuse;
 		RID normal;
 		RID specular;
 		RID sampler;
+		Vector2 texpixel_size;
+		uint32_t specular_shininess = 0;
+		uint32_t flags = 0;
+	};
+
+	/// A key used to uniquely identify a distinct BATCH_UNIFORM_SET
+	struct RIDSetKey {
+		TextureState state;
+		RID instance_data;
+
+		RIDSetKey() {
+		}
+
+		RIDSetKey(TextureState p_state, RID p_instance_data) :
+				state(p_state),
+				instance_data(p_instance_data) {
+		}
+
+		_ALWAYS_INLINE_ bool operator==(const RIDSetKey &p_val) const {
+			return state == p_val.state && instance_data == p_val.instance_data;
+		}
+
+		_ALWAYS_INLINE_ bool operator!=(const RIDSetKey &p_val) const {
+			return !(*this == p_val);
+		}
+
+		_ALWAYS_INLINE_ uint32_t hash() const {
+			uint32_t h = state.hash();
+			h = hash_murmur3_one_64(instance_data.get_id(), h);
+			return hash_fmix32(h);
+		}
 	};
 
+	static void _before_evict(RendererCanvasRenderRD::RIDSetKey &p_key, RID &p_rid);
+	static void _uniform_set_invalidation_callback(void *p_userdata);
+
+	typedef LRUCache<RIDSetKey, RID, HashableHasher<RIDSetKey>, HashMapComparatorDefault<RIDSetKey>, _before_evict> RIDCache;
+	RIDCache rid_set_to_uniform_set;
+
 	struct Batch {
 		// Position in the UBO measured in bytes
 		uint32_t start = 0;
 		uint32_t instance_count = 0;
 		uint32_t instance_buffer_index = 0;
 
-		TextureInfo tex_info;
+		TextureInfo *tex_info;
 
 		Color modulate = Color(1.0, 1.0, 1.0, 1.0);
 
@@ -468,6 +512,9 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 		bool has_blend = false;
 	};
 
+	HashMap<TextureState, TextureInfo, HashableHasher<TextureState>> texture_info_map;
+
+	// per-frame buffers
 	struct DataBuffer {
 		LocalVector<RID> instance_buffers;
 	};
@@ -494,7 +541,7 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 			uint32_t pad2;
 		};
 
-		LocalVector<DataBuffer> canvas_instance_data_buffers;
+		DataBuffer canvas_instance_data_buffers[BATCH_DATA_BUFFER_COUNT];
 		LocalVector<Batch> canvas_instance_batches;
 		uint32_t current_data_buffer_index = 0;
 		uint32_t current_instance_buffer_index = 0;
@@ -505,6 +552,7 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 		uint32_t max_instances_per_buffer = 16384;
 		uint32_t max_instance_buffer_size = 16384 * sizeof(InstanceData);
 
+		Vector<RD::Uniform> batch_texture_uniforms;
 		RID current_batch_uniform_set;
 
 		LightUniform *light_uniforms = nullptr;
@@ -520,7 +568,6 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 		RID default_transforms_uniform_set;
 
 		uint32_t max_lights_per_render;
-		uint32_t max_lights_per_item;
 
 		double time;
 
@@ -560,7 +607,8 @@ class RendererCanvasRenderRD : public RendererCanvasRender {
 	void _render_batch_items(RenderTarget p_to_render_target, int p_item_count, const Transform2D &p_canvas_transform_inverse, Light *p_lights, bool &r_sdf_used, bool p_to_backbuffer = false, RenderingMethod::RenderInfo *r_render_info = nullptr);
 	void _record_item_commands(const Item *p_item, RenderTarget p_render_target, const Transform2D &p_base_transform, Item *&r_current_clip, Light *p_lights, uint32_t &r_index, bool &r_batch_broken, bool &r_sdf_used, Batch *&r_current_batch);
 	void _render_batch(RD::DrawListID p_draw_list, CanvasShaderData *p_shader_data, RenderingDevice::FramebufferFormatID p_framebuffer_format, Light *p_lights, Batch const *p_batch, RenderingMethod::RenderInfo *r_render_info = nullptr);
-	void _prepare_batch_texture_info(Batch *p_current_batch, RID p_texture) const;
+	void _prepare_batch_texture_info(RID p_texture, TextureState &p_state, TextureInfo *p_info);
+	InstanceData *new_instance_data(float *p_world, uint32_t *p_lights, uint32_t p_base_flags, uint32_t p_index, TextureInfo *p_info);
 	[[nodiscard]] Batch *_new_batch(bool &r_batch_broken);
 	void _add_to_batch(uint32_t &r_index, bool &r_batch_broken, Batch *&r_current_batch);
 	void _allocate_instance_buffer();
diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp
index 9df54e372e..7514e1d5af 100644
--- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp
+++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp
@@ -68,6 +68,16 @@ void RendererCompositorRD::blit_render_targets_to_screen(DisplayServer::WindowID
 		RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array);
 		RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_target_descriptors[rd_texture], 0);
 
+		// We need to invert the phone rotation.
+		int screen_rotation_degrees = -DisplayServer::get_singleton()->screen_get_internal_current_rotation();
+		float screen_rotation = Math::deg_to_rad((float)screen_rotation_degrees);
+
+		blit.push_constant.rotation_cos = Math::cos(screen_rotation);
+		blit.push_constant.rotation_sin = Math::sin(screen_rotation);
+		// Swap width and height when the orientation is not the native one.
+		if (screen_rotation_degrees % 180 != 0) {
+			SWAP(screen_size.width, screen_size.height);
+		}
 		blit.push_constant.src_rect[0] = p_render_targets[i].src_rect.position.x;
 		blit.push_constant.src_rect[1] = p_render_targets[i].src_rect.position.y;
 		blit.push_constant.src_rect[2] = p_render_targets[i].src_rect.size.width;
@@ -230,6 +240,10 @@ void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color
 	RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array);
 	RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uset, 0);
 
+	int screen_rotation_degrees = DisplayServer::get_singleton()->screen_get_internal_current_rotation();
+	float screen_rotation = Math::deg_to_rad((float)screen_rotation_degrees);
+	blit.push_constant.rotation_cos = Math::cos(screen_rotation);
+	blit.push_constant.rotation_sin = Math::sin(screen_rotation);
 	blit.push_constant.src_rect[0] = 0.0;
 	blit.push_constant.src_rect[1] = 0.0;
 	blit.push_constant.src_rect[2] = 1.0;
diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h
index 9a4ad87bf0..728879134f 100644
--- a/servers/rendering/renderer_rd/renderer_compositor_rd.h
+++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h
@@ -76,6 +76,10 @@ protected:
 		float src_rect[4];
 		float dst_rect[4];
 
+		float rotation_sin;
+		float rotation_cos;
+		float pad[2];
+
 		float eye_center[2];
 		float k1;
 		float k2;
@@ -105,7 +109,7 @@ protected:
 	static RendererCompositorRD *singleton;
 
 public:
-	RendererUtilities *get_utilities() { return utilities; };
+	RendererUtilities *get_utilities() { return utilities; }
 	RendererLightStorage *get_light_storage() { return light_storage; }
 	RendererMaterialStorage *get_material_storage() { return material_storage; }
 	RendererMeshStorage *get_mesh_storage() { return mesh_storage; }
diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h
index 7c1b014af1..25a55debdd 100644
--- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h
+++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h
@@ -80,7 +80,7 @@ protected:
 
 	////////////////////////////////
 
-	virtual RendererRD::ForwardIDStorage *create_forward_id_storage() { return memnew(RendererRD::ForwardIDStorage); };
+	virtual RendererRD::ForwardIDStorage *create_forward_id_storage() { return memnew(RendererRD::ForwardIDStorage); }
 
 	void _update_vrs(Ref<RenderSceneBuffersRD> p_render_buffers);
 
diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl
index d451647bec..fe6416f03c 100644
--- a/servers/rendering/renderer_rd/shaders/blit.glsl
+++ b/servers/rendering/renderer_rd/shaders/blit.glsl
@@ -8,6 +8,10 @@ layout(push_constant, std140) uniform Pos {
 	vec4 src_rect;
 	vec4 dst_rect;
 
+	float rotation_sin;
+	float rotation_cos;
+	vec2 pad;
+
 	vec2 eye_center;
 	float k1;
 	float k2;
@@ -15,17 +19,23 @@ layout(push_constant, std140) uniform Pos {
 	float upscale;
 	float aspect_ratio;
 	uint layer;
-	uint pad1;
+	bool convert_to_srgb;
 }
 data;
 
 layout(location = 0) out vec2 uv;
 
 void main() {
+	mat4 swapchain_transform = mat4(1.0);
+	swapchain_transform[0][0] = data.rotation_cos;
+	swapchain_transform[0][1] = -data.rotation_sin;
+	swapchain_transform[1][0] = data.rotation_sin;
+	swapchain_transform[1][1] = data.rotation_cos;
+
 	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 = data.src_rect.xy + base_arr[gl_VertexIndex] * data.src_rect.zw;
 	vec2 vtx = data.dst_rect.xy + base_arr[gl_VertexIndex] * data.dst_rect.zw;
-	gl_Position = vec4(vtx * 2.0 - 1.0, 0.0, 1.0);
+	gl_Position = swapchain_transform * vec4(vtx * 2.0 - 1.0, 0.0, 1.0);
 }
 
 #[fragment]
@@ -38,6 +48,10 @@ layout(push_constant, std140) uniform Pos {
 	vec4 src_rect;
 	vec4 dst_rect;
 
+	float rotation_sin;
+	float rotation_cos;
+	vec2 pad;
+
 	vec2 eye_center;
 	float k1;
 	float k2;
diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl
index dafcce37ad..f665bc24a4 100644
--- a/servers/rendering/renderer_rd/shaders/canvas.glsl
+++ b/servers/rendering/renderer_rd/shaders/canvas.glsl
@@ -28,7 +28,7 @@ layout(location = 11) in vec4 weight_attrib;
 
 layout(location = 4) out flat uint instance_index_interp;
 
-#endif // USE_ATTRIBUTES
+#endif // !USE_ATTRIBUTES
 
 layout(location = 0) out vec2 uv_interp;
 layout(location = 1) out vec4 color_interp;
@@ -322,11 +322,7 @@ vec4 light_compute(
 #ifdef USE_NINEPATCH
 
 float map_ninepatch_axis(float pixel, float draw_size, float tex_pixel_size, float margin_begin, float margin_end, int np_repeat, inout int draw_center) {
-#ifdef USE_ATTRIBUTES
-	const InstanceData draw_data = instances.data[params.base_instance_index];
-#else
 	const InstanceData draw_data = instances.data[instance_index];
-#endif // USE_ATTRIBUTES
 
 	float tex_size = 1.0 / tex_pixel_size;
 
@@ -567,7 +563,7 @@ void main() {
 
 	if (specular_shininess_used || (using_light && normal_used && bool(draw_data.flags & FLAGS_DEFAULT_SPECULAR_MAP_USED))) {
 		specular_shininess = texture(sampler2D(specular_texture, texture_sampler), uv);
-		specular_shininess *= unpackUnorm4x8(draw_data.specular_shininess);
+		specular_shininess *= unpackUnorm4x8(params.specular_shininess);
 		specular_shininess_used = true;
 	} else {
 		specular_shininess = vec4(1.0);
diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl
index ead8c459a4..84017a1fe1 100644
--- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl
@@ -37,7 +37,7 @@ struct InstanceData {
 	vec2 world_y;
 	vec2 world_ofs;
 	uint flags;
-	uint specular_shininess;
+	uint pad2;
 #ifdef USE_PRIMITIVE
 	vec2 points[3];
 	vec2 uvs[3];
@@ -57,8 +57,8 @@ struct InstanceData {
 layout(push_constant, std430) uniform Params {
 	uint base_instance_index; // base index to instance data
 	uint sc_packed_0;
-	uint pad2;
-	uint pad3;
+	uint specular_shininess;
+	uint pad;
 }
 params;
 
@@ -68,7 +68,7 @@ params;
 
 // Pull the constants from the draw call's push constants.
 uint sc_packed_0() {
-	return draw_call.sc_packed_0;
+	return params.sc_packed_0;
 }
 
 #else
diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl
index 9f68d59be2..7bfcb2fb12 100644
--- a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl
@@ -107,19 +107,19 @@ bool sc_use_lightmap_bicubic_filter() {
 }
 
 uint sc_soft_shadow_samples() {
-	return (sc_packed_0() >> 8) & 15U;
+	return (sc_packed_0() >> 8) & 63U;
 }
 
 uint sc_penumbra_shadow_samples() {
-	return (sc_packed_0() >> 12) & 15U;
+	return (sc_packed_0() >> 14) & 63U;
 }
 
 uint sc_directional_soft_shadow_samples() {
-	return (sc_packed_0() >> 16) & 15U;
+	return (sc_packed_0() >> 20) & 63U;
 }
 
 uint sc_directional_penumbra_shadow_samples() {
-	return (sc_packed_0() >> 20) & 15U;
+	return (sc_packed_0() >> 26) & 63U;
 }
 
 float sc_luminance_multiplier() {
diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl
index 495e52a29e..df528973da 100644
--- a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl
@@ -22,8 +22,8 @@ layout(push_constant, std430) uniform DrawCall {
 	uint pad;
 #ifdef UBERSHADER
 	uint sc_packed_0;
-	float sc_packed_1;
-	uint sc_packed_2;
+	uint sc_packed_1;
+	float sc_packed_2;
 	uint uc_packed_0;
 #endif
 }
@@ -42,10 +42,14 @@ uint sc_packed_0() {
 	return draw_call.sc_packed_0;
 }
 
-float sc_packed_1() {
+uint sc_packed_1() {
 	return draw_call.sc_packed_1;
 }
 
+float sc_packed_2() {
+	return draw_call.sc_packed_2;
+}
+
 uint uc_cull_mode() {
 	return (draw_call.uc_packed_0 >> 0) & 3U;
 }
@@ -54,16 +58,21 @@ uint uc_cull_mode() {
 
 // Pull the constants from the pipeline's specialization constants.
 layout(constant_id = 0) const uint pso_sc_packed_0 = 0;
-layout(constant_id = 1) const float pso_sc_packed_1 = 2.0;
+layout(constant_id = 1) const uint pso_sc_packed_1 = 0;
+layout(constant_id = 2) const float pso_sc_packed_2 = 2.0;
 
 uint sc_packed_0() {
 	return pso_sc_packed_0;
 }
 
-float sc_packed_1() {
+uint sc_packed_1() {
 	return pso_sc_packed_1;
 }
 
+float sc_packed_2() {
+	return pso_sc_packed_2;
+}
+
 #endif
 
 bool sc_use_light_projector() {
@@ -123,23 +132,23 @@ bool sc_use_lightmap_bicubic_filter() {
 }
 
 uint sc_soft_shadow_samples() {
-	return (sc_packed_0() >> 16) & 15U;
+	return (sc_packed_0() >> 14) & 63U;
 }
 
 uint sc_penumbra_shadow_samples() {
-	return (sc_packed_0() >> 20) & 15U;
+	return (sc_packed_0() >> 20) & 63U;
 }
 
 uint sc_directional_soft_shadow_samples() {
-	return (sc_packed_0() >> 24) & 15U;
+	return (sc_packed_0() >> 26) & 63U;
 }
 
 uint sc_directional_penumbra_shadow_samples() {
-	return (sc_packed_0() >> 28) & 15U;
+	return (sc_packed_1() >> 0) & 63U;
 }
 
 float sc_luminance_multiplier() {
-	return sc_packed_1();
+	return sc_packed_2();
 }
 
 /* Set 0: Base Pass (never changes) */
diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp
index d219047cba..410a6ae5c6 100644
--- a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp
+++ b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp
@@ -1051,7 +1051,7 @@ void LightStorage::reflection_probe_free(RID p_rid) {
 	ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_rid);
 	reflection_probe->dependency.deleted_notify(p_rid);
 	reflection_probe_owner.free(p_rid);
-};
+}
 
 void LightStorage::reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) {
 	ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe);
diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.h b/servers/rendering/renderer_rd/storage_rd/light_storage.h
index 210e4c3035..7cd7d3e363 100644
--- a/servers/rendering/renderer_rd/storage_rd/light_storage.h
+++ b/servers/rendering/renderer_rd/storage_rd/light_storage.h
@@ -592,7 +592,7 @@ public:
 
 	/* LIGHT INSTANCE API */
 
-	bool owns_light_instance(RID p_rid) { return light_instance_owner.owns(p_rid); };
+	bool owns_light_instance(RID p_rid) { return light_instance_owner.owns(p_rid); }
 
 	virtual RID light_instance_create(RID p_light) override;
 	virtual void light_instance_free(RID p_light) override;
@@ -813,7 +813,7 @@ public:
 
 	/* REFLECTION PROBE */
 
-	bool owns_reflection_probe(RID p_rid) { return reflection_probe_owner.owns(p_rid); };
+	bool owns_reflection_probe(RID p_rid) { return reflection_probe_owner.owns(p_rid); }
 
 	virtual RID reflection_probe_allocate() override;
 	virtual void reflection_probe_initialize(RID p_reflection_probe) override;
@@ -955,7 +955,7 @@ public:
 
 	/* LIGHTMAP */
 
-	bool owns_lightmap(RID p_rid) { return lightmap_owner.owns(p_rid); };
+	bool owns_lightmap(RID p_rid) { return lightmap_owner.owns(p_rid); }
 
 	virtual RID lightmap_allocate() override;
 	virtual void lightmap_initialize(RID p_lightmap) override;
@@ -1021,7 +1021,7 @@ public:
 
 	/* LIGHTMAP INSTANCE */
 
-	bool owns_lightmap_instance(RID p_rid) { return lightmap_instance_owner.owns(p_rid); };
+	bool owns_lightmap_instance(RID p_rid) { return lightmap_instance_owner.owns(p_rid); }
 
 	virtual RID lightmap_instance_create(RID p_lightmap) override;
 	virtual void lightmap_instance_free(RID p_lightmap) override;
@@ -1041,7 +1041,7 @@ public:
 
 	/* SHADOW ATLAS API */
 
-	bool owns_shadow_atlas(RID p_rid) { return shadow_atlas_owner.owns(p_rid); };
+	bool owns_shadow_atlas(RID p_rid) { return shadow_atlas_owner.owns(p_rid); }
 
 	virtual RID shadow_atlas_create() override;
 	virtual void shadow_atlas_free(RID p_atlas) override;
diff --git a/servers/rendering/renderer_rd/storage_rd/material_storage.h b/servers/rendering/renderer_rd/storage_rd/material_storage.h
index 2f477c4196..e6a25c547a 100644
--- a/servers/rendering/renderer_rd/storage_rd/material_storage.h
+++ b/servers/rendering/renderer_rd/storage_rd/material_storage.h
@@ -256,7 +256,7 @@ private:
 
 	MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX];
 	mutable RID_Owner<Material, true> material_owner;
-	Material *get_material(RID p_rid) { return material_owner.get_or_null(p_rid); };
+	Material *get_material(RID p_rid) { return material_owner.get_or_null(p_rid); }
 
 	SelfList<Material>::List material_update_list;
 	Mutex material_update_list_mutex;
@@ -405,7 +405,7 @@ public:
 
 	/* SHADER API */
 
-	bool owns_shader(RID p_rid) { return shader_owner.owns(p_rid); };
+	bool owns_shader(RID p_rid) { return shader_owner.owns(p_rid); }
 
 	virtual RID shader_allocate() override;
 	virtual void shader_initialize(RID p_shader) override;
@@ -425,7 +425,7 @@ public:
 
 	/* MATERIAL API */
 
-	bool owns_material(RID p_rid) { return material_owner.owns(p_rid); };
+	bool owns_material(RID p_rid) { return material_owner.owns(p_rid); }
 
 	void _material_queue_update(Material *material, bool p_uniform, bool p_texture);
 	void _update_queued_materials();
diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h
index 79bd8d0be2..e70b7568e0 100644
--- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h
+++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h
@@ -353,7 +353,7 @@ public:
 
 	/* MESH API */
 
-	bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); };
+	bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); }
 
 	virtual RID mesh_allocate() override;
 	virtual void mesh_initialize(RID p_mesh) override;
@@ -618,7 +618,7 @@ public:
 
 	/* MESH INSTANCE API */
 
-	bool owns_mesh_instance(RID p_rid) const { return mesh_instance_owner.owns(p_rid); };
+	bool owns_mesh_instance(RID p_rid) const { return mesh_instance_owner.owns(p_rid); }
 
 	virtual RID mesh_instance_create(RID p_base) override;
 	virtual void mesh_instance_free(RID p_rid) override;
@@ -630,7 +630,7 @@ public:
 
 	/* MULTIMESH API */
 
-	bool owns_multimesh(RID p_rid) { return multimesh_owner.owns(p_rid); };
+	bool owns_multimesh(RID p_rid) { return multimesh_owner.owns(p_rid); }
 
 	virtual RID _multimesh_allocate() override;
 	virtual void _multimesh_initialize(RID p_multimesh) override;
@@ -739,7 +739,7 @@ public:
 
 	/* SKELETON API */
 
-	bool owns_skeleton(RID p_rid) const { return skeleton_owner.owns(p_rid); };
+	bool owns_skeleton(RID p_rid) const { return skeleton_owner.owns(p_rid); }
 
 	virtual RID skeleton_allocate() override;
 	virtual void skeleton_initialize(RID p_skeleton) override;
diff --git a/servers/rendering/renderer_rd/storage_rd/texture_storage.h b/servers/rendering/renderer_rd/storage_rd/texture_storage.h
index 4e257f0e04..2d59b822a0 100644
--- a/servers/rendering/renderer_rd/storage_rd/texture_storage.h
+++ b/servers/rendering/renderer_rd/storage_rd/texture_storage.h
@@ -124,7 +124,6 @@ private:
 		Size2i size_cache = Size2i(1, 1);
 		bool use_normal_cache = false;
 		bool use_specular_cache = false;
-		bool cleared_cache = true;
 
 		void clear_cache();
 		~CanvasTexture();
@@ -197,7 +196,7 @@ private:
 
 	// Textures can be created from threads, so this RID_Owner is thread safe.
 	mutable RID_Owner<Texture, true> texture_owner;
-	Texture *get_texture(RID p_rid) { return texture_owner.get_or_null(p_rid); };
+	Texture *get_texture(RID p_rid) { return texture_owner.get_or_null(p_rid); }
 
 	struct TextureToRDFormat {
 		RD::DataFormat format;
@@ -441,7 +440,7 @@ private:
 	};
 
 	mutable RID_Owner<RenderTarget> render_target_owner;
-	RenderTarget *get_render_target(RID p_rid) const { return render_target_owner.get_or_null(p_rid); };
+	RenderTarget *get_render_target(RID p_rid) const { return render_target_owner.get_or_null(p_rid); }
 
 	void _clear_render_target(RenderTarget *rt);
 	void _update_render_target(RenderTarget *rt);
@@ -488,7 +487,7 @@ public:
 
 	/* Canvas Texture API */
 
-	bool owns_canvas_texture(RID p_rid) { return canvas_texture_owner.owns(p_rid); };
+	bool owns_canvas_texture(RID p_rid) { return canvas_texture_owner.owns(p_rid); }
 
 	virtual RID canvas_texture_allocate() override;
 	virtual void canvas_texture_initialize(RID p_rid) override;
@@ -504,7 +503,7 @@ public:
 
 	/* Texture API */
 
-	bool owns_texture(RID p_rid) const { return texture_owner.owns(p_rid); };
+	bool owns_texture(RID p_rid) const { return texture_owner.owns(p_rid); }
 
 	virtual RID texture_allocate() override;
 	virtual void texture_free(RID p_rid) override;
@@ -593,7 +592,7 @@ public:
 
 	void update_decal_atlas();
 
-	bool owns_decal(RID p_rid) const { return decal_owner.owns(p_rid); };
+	bool owns_decal(RID p_rid) const { return decal_owner.owns(p_rid); }
 
 	RID decal_atlas_get_texture() const;
 	RID decal_atlas_get_texture_srgb() const;
@@ -733,7 +732,7 @@ public:
 
 	/* RENDER TARGET API */
 
-	bool owns_render_target(RID p_rid) const { return render_target_owner.owns(p_rid); };
+	bool owns_render_target(RID p_rid) const { return render_target_owner.owns(p_rid); }
 
 	virtual RID render_target_create() override;
 	virtual void render_target_free(RID p_rid) override;
diff --git a/servers/rendering/renderer_rd/storage_rd/utilities.h b/servers/rendering/renderer_rd/storage_rd/utilities.h
index 5fafe2752d..bdbe7af803 100644
--- a/servers/rendering/renderer_rd/storage_rd/utilities.h
+++ b/servers/rendering/renderer_rd/storage_rd/utilities.h
@@ -79,8 +79,8 @@ public:
 
 	/* VISIBILITY NOTIFIER */
 
-	VisibilityNotifier *get_visibility_notifier(RID p_rid) { return visibility_notifier_owner.get_or_null(p_rid); };
-	bool owns_visibility_notifier(RID p_rid) const { return visibility_notifier_owner.owns(p_rid); };
+	VisibilityNotifier *get_visibility_notifier(RID p_rid) { return visibility_notifier_owner.get_or_null(p_rid); }
+	bool owns_visibility_notifier(RID p_rid) const { return visibility_notifier_owner.owns(p_rid); }
 
 	virtual RID visibility_notifier_allocate() override;
 	virtual void visibility_notifier_initialize(RID p_notifier) override;
diff --git a/servers/rendering/renderer_scene_occlusion_cull.h b/servers/rendering/renderer_scene_occlusion_cull.h
index 51272d803c..94159f72bb 100644
--- a/servers/rendering/renderer_scene_occlusion_cull.h
+++ b/servers/rendering/renderer_scene_occlusion_cull.h
@@ -86,7 +86,7 @@ public:
 				Vector3 view = p_cam_inv_transform.xform(corner);
 
 				if (p_cam_projection.is_orthogonal()) {
-					min_depth = MIN(min_depth, view.z);
+					min_depth = MIN(min_depth, -view.z);
 				}
 
 				Plane vp = Plane(view, 1.0);
@@ -236,11 +236,11 @@ public:
 
 	RendererSceneOcclusionCull() {
 		singleton = this;
-	};
+	}
 
 	virtual ~RendererSceneOcclusionCull() {
 		singleton = nullptr;
-	};
+	}
 };
 
 #endif // RENDERER_SCENE_OCCLUSION_CULL_H
diff --git a/servers/rendering/rendering_device.cpp b/servers/rendering/rendering_device.cpp
index 3c328ab34d..df94d320c3 100644
--- a/servers/rendering/rendering_device.cpp
+++ b/servers/rendering/rendering_device.cpp
@@ -1262,8 +1262,6 @@ Error RenderingDevice::_texture_initialize(RID p_texture, uint32_t p_layer, cons
 			write_ptr = driver->buffer_map(transfer_worker->staging_buffer);
 			ERR_FAIL_NULL_V(write_ptr, ERR_CANT_CREATE);
 
-			write_ptr += staging_worker_offset;
-
 			if (driver->api_trait_get(RDD::API_TRAIT_HONORS_PIPELINE_BARRIERS)) {
 				// Transition the texture to the optimal layout.
 				RDD::TextureBarrier tb;
@@ -1305,11 +1303,12 @@ Error RenderingDevice::_texture_initialize(RID p_texture, uint32_t p_layer, cons
 
 				if (copy_pass) {
 					const uint8_t *read_ptr_mipmap_layer = read_ptr_mipmap + (tight_mip_size / depth) * z;
-					_copy_region_block_or_regular(read_ptr_mipmap_layer, write_ptr, 0, 0, width, width, height, block_w, block_h, pitch, pixel_size, block_size);
-					write_ptr += to_allocate;
+					uint64_t staging_buffer_offset = staging_worker_offset + staging_local_offset;
+					uint8_t *write_ptr_mipmap_layer = write_ptr + staging_buffer_offset;
+					_copy_region_block_or_regular(read_ptr_mipmap_layer, write_ptr_mipmap_layer, 0, 0, width, width, height, block_w, block_h, pitch, pixel_size, block_size);
 
 					RDD::BufferTextureCopyRegion copy_region;
-					copy_region.buffer_offset = staging_worker_offset + staging_local_offset;
+					copy_region.buffer_offset = staging_buffer_offset;
 					copy_region.texture_subresources.aspect = texture->read_aspect_flags;
 					copy_region.texture_subresources.mipmap = mm_i;
 					copy_region.texture_subresources.base_layer = p_layer;
@@ -4443,6 +4442,117 @@ void RenderingDevice::draw_list_draw(DrawListID p_list, bool p_use_indices, uint
 	dl->state.draw_count++;
 }
 
+void RenderingDevice::draw_list_draw_indirect(DrawListID p_list, bool p_use_indices, RID p_buffer, uint32_t p_offset, uint32_t p_draw_count, uint32_t p_stride) {
+	ERR_RENDER_THREAD_GUARD();
+
+	DrawList *dl = _get_draw_list_ptr(p_list);
+	ERR_FAIL_NULL(dl);
+
+	Buffer *buffer = storage_buffer_owner.get_or_null(p_buffer);
+	ERR_FAIL_NULL(buffer);
+
+	ERR_FAIL_COND_MSG(!buffer->usage.has_flag(RDD::BUFFER_USAGE_INDIRECT_BIT), "Buffer provided was not created to do indirect dispatch.");
+
+#ifdef DEBUG_ENABLED
+	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
+#endif
+
+#ifdef DEBUG_ENABLED
+	ERR_FAIL_COND_MSG(!dl->validation.pipeline_active,
+			"No render pipeline was set before attempting to draw.");
+	if (dl->validation.pipeline_vertex_format != INVALID_ID) {
+		// Pipeline uses vertices, validate format.
+		ERR_FAIL_COND_MSG(dl->validation.vertex_format == INVALID_ID,
+				"No vertex array was bound, and render pipeline expects vertices.");
+		// Make sure format is right.
+		ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != dl->validation.vertex_format,
+				"The vertex format used to create the pipeline does not match the vertex format bound.");
+	}
+
+	if (dl->validation.pipeline_push_constant_size > 0) {
+		// Using push constants, check that they were supplied.
+		ERR_FAIL_COND_MSG(!dl->validation.pipeline_push_constant_supplied,
+				"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
+	}
+#endif
+
+#ifdef DEBUG_ENABLED
+	for (uint32_t i = 0; i < dl->state.set_count; i++) {
+		if (dl->state.sets[i].pipeline_expected_format == 0) {
+			// Nothing expected by this pipeline.
+			continue;
+		}
+
+		if (dl->state.sets[i].pipeline_expected_format != dl->state.sets[i].uniform_set_format) {
+			if (dl->state.sets[i].uniform_set_format == 0) {
+				ERR_FAIL_MSG(vformat("Uniforms were never supplied for set (%d) at the time of drawing, which are required by the pipeline.", i));
+			} else if (uniform_set_owner.owns(dl->state.sets[i].uniform_set)) {
+				UniformSet *us = uniform_set_owner.get_or_null(dl->state.sets[i].uniform_set);
+				ERR_FAIL_MSG(vformat("Uniforms supplied for set (%d):\n%s\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n%s", i, _shader_uniform_debug(us->shader_id, us->shader_set), _shader_uniform_debug(dl->state.pipeline_shader)));
+			} else {
+				ERR_FAIL_MSG(vformat("Uniforms supplied for set (%s, which was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n%s", i, _shader_uniform_debug(dl->state.pipeline_shader)));
+			}
+		}
+	}
+#endif
+
+	// Prepare descriptor sets if the API doesn't use pipeline barriers.
+	if (!driver->api_trait_get(RDD::API_TRAIT_HONORS_PIPELINE_BARRIERS)) {
+		for (uint32_t i = 0; i < dl->state.set_count; i++) {
+			if (dl->state.sets[i].pipeline_expected_format == 0) {
+				// Nothing expected by this pipeline.
+				continue;
+			}
+
+			draw_graph.add_draw_list_uniform_set_prepare_for_use(dl->state.pipeline_shader_driver_id, dl->state.sets[i].uniform_set_driver_id, i);
+		}
+	}
+
+	// Bind descriptor sets.
+	for (uint32_t i = 0; i < dl->state.set_count; i++) {
+		if (dl->state.sets[i].pipeline_expected_format == 0) {
+			continue; // Nothing expected by this pipeline.
+		}
+		if (!dl->state.sets[i].bound) {
+			// All good, see if this requires re-binding.
+			draw_graph.add_draw_list_bind_uniform_set(dl->state.pipeline_shader_driver_id, dl->state.sets[i].uniform_set_driver_id, i);
+
+			UniformSet *uniform_set = uniform_set_owner.get_or_null(dl->state.sets[i].uniform_set);
+			_uniform_set_update_shared(uniform_set);
+
+			draw_graph.add_draw_list_usages(uniform_set->draw_trackers, uniform_set->draw_trackers_usage);
+
+			dl->state.sets[i].bound = true;
+		}
+	}
+
+	if (p_use_indices) {
+#ifdef DEBUG_ENABLED
+		ERR_FAIL_COND_MSG(!dl->validation.index_array_count,
+				"Draw command requested indices, but no index buffer was set.");
+
+		ERR_FAIL_COND_MSG(dl->validation.pipeline_uses_restart_indices != dl->validation.index_buffer_uses_restart_indices,
+				"The usage of restart indices in index buffer does not match the render primitive in the pipeline.");
+#endif
+
+		ERR_FAIL_COND_MSG(p_offset + 20 > buffer->size, "Offset provided (+20) is past the end of buffer.");
+
+		draw_graph.add_draw_list_draw_indexed_indirect(buffer->driver_id, p_offset, p_draw_count, p_stride);
+	} else {
+		ERR_FAIL_COND_MSG(p_offset + 16 > buffer->size, "Offset provided (+16) is past the end of buffer.");
+
+		draw_graph.add_draw_list_draw_indirect(buffer->driver_id, p_offset, p_draw_count, p_stride);
+	}
+
+	dl->state.draw_count++;
+
+	if (buffer->draw_tracker != nullptr) {
+		draw_graph.add_draw_list_usage(buffer->draw_tracker, RDG::RESOURCE_USAGE_INDIRECT_BUFFER_READ);
+	}
+
+	_check_transfer_worker_buffer(buffer);
+}
+
 void RenderingDevice::draw_list_enable_scissor(DrawListID p_list, const Rect2 &p_rect) {
 	ERR_RENDER_THREAD_GUARD();
 
@@ -6555,6 +6665,12 @@ void RenderingDevice::finalize() {
 	ERR_FAIL_COND(reverse_dependency_map.size());
 }
 
+void RenderingDevice::_set_max_fps(int p_max_fps) {
+	for (const KeyValue<DisplayServer::WindowID, RDD::SwapChainID> &it : screen_swap_chains) {
+		driver->swap_chain_set_max_fps(it.value, p_max_fps);
+	}
+}
+
 RenderingDevice *RenderingDevice::create_local_device() {
 	RenderingDevice *rd = memnew(RenderingDevice);
 	rd->initialize(context);
@@ -6653,6 +6769,7 @@ void RenderingDevice::_bind_methods() {
 	ClassDB::bind_method(D_METHOD("draw_list_set_push_constant", "draw_list", "buffer", "size_bytes"), &RenderingDevice::_draw_list_set_push_constant);
 
 	ClassDB::bind_method(D_METHOD("draw_list_draw", "draw_list", "use_indices", "instances", "procedural_vertex_count"), &RenderingDevice::draw_list_draw, DEFVAL(0));
+	ClassDB::bind_method(D_METHOD("draw_list_draw_indirect", "draw_list", "use_indices", "buffer", "offset", "draw_count", "stride"), &RenderingDevice::draw_list_draw_indirect, DEFVAL(0), DEFVAL(1), DEFVAL(0));
 
 	ClassDB::bind_method(D_METHOD("draw_list_enable_scissor", "draw_list", "rect"), &RenderingDevice::draw_list_enable_scissor, DEFVAL(Rect2()));
 	ClassDB::bind_method(D_METHOD("draw_list_disable_scissor", "draw_list"), &RenderingDevice::draw_list_disable_scissor);
diff --git a/servers/rendering/rendering_device.h b/servers/rendering/rendering_device.h
index 0a3663523e..6831f3df2b 100644
--- a/servers/rendering/rendering_device.h
+++ b/servers/rendering/rendering_device.h
@@ -1188,6 +1188,7 @@ public:
 	void draw_list_set_push_constant(DrawListID p_list, const void *p_data, uint32_t p_data_size);
 
 	void draw_list_draw(DrawListID p_list, bool p_use_indices, uint32_t p_instances = 1, uint32_t p_procedural_vertices = 0);
+	void draw_list_draw_indirect(DrawListID p_list, bool p_use_indices, RID p_buffer, uint32_t p_offset = 0, uint32_t p_draw_count = 1, uint32_t p_stride = 0);
 
 	void draw_list_enable_scissor(DrawListID p_list, const Rect2 &p_rect);
 	void draw_list_disable_scissor(DrawListID p_list);
@@ -1429,6 +1430,8 @@ public:
 	Error initialize(RenderingContextDriver *p_context, DisplayServer::WindowID p_main_window = DisplayServer::INVALID_WINDOW_ID);
 	void finalize();
 
+	void _set_max_fps(int p_max_fps);
+
 	void free(RID p_id);
 
 	/****************/
diff --git a/servers/rendering/rendering_device_driver.h b/servers/rendering/rendering_device_driver.h
index 22ef46a1c8..258717091d 100644
--- a/servers/rendering/rendering_device_driver.h
+++ b/servers/rendering/rendering_device_driver.h
@@ -459,6 +459,10 @@ public:
 	// Retrieve the format used by the swap chain's framebuffers.
 	virtual DataFormat swap_chain_get_format(SwapChainID p_swap_chain) = 0;
 
+	// Tells the swapchain the max_fps so it can use the proper frame pacing.
+	// Android uses this with Swappy library. Some implementations or platforms may ignore this hint.
+	virtual void swap_chain_set_max_fps(SwapChainID p_swap_chain, int p_max_fps) {}
+
 	// Wait until all rendering associated to the swap chain is finished before deleting it.
 	virtual void swap_chain_free(SwapChainID p_swap_chain) = 0;
 
diff --git a/servers/rendering/rendering_device_graph.cpp b/servers/rendering/rendering_device_graph.cpp
index fa26fdd79d..2e394df0bb 100644
--- a/servers/rendering/rendering_device_graph.cpp
+++ b/servers/rendering/rendering_device_graph.cpp
@@ -710,6 +710,16 @@ void RenderingDeviceGraph::_run_draw_list_command(RDD::CommandBufferID p_command
 				driver->command_render_draw_indexed(p_command_buffer, draw_indexed_instruction->index_count, draw_indexed_instruction->instance_count, draw_indexed_instruction->first_index, 0, 0);
 				instruction_data_cursor += sizeof(DrawListDrawIndexedInstruction);
 			} break;
+			case DrawListInstruction::TYPE_DRAW_INDIRECT: {
+				const DrawListDrawIndirectInstruction *draw_indirect_instruction = reinterpret_cast<const DrawListDrawIndirectInstruction *>(instruction);
+				driver->command_render_draw_indirect(p_command_buffer, draw_indirect_instruction->buffer, draw_indirect_instruction->offset, draw_indirect_instruction->draw_count, draw_indirect_instruction->stride);
+				instruction_data_cursor += sizeof(DrawListDrawIndirectInstruction);
+			} break;
+			case DrawListInstruction::TYPE_DRAW_INDEXED_INDIRECT: {
+				const DrawListDrawIndexedIndirectInstruction *draw_indexed_indirect_instruction = reinterpret_cast<const DrawListDrawIndexedIndirectInstruction *>(instruction);
+				driver->command_render_draw_indexed_indirect(p_command_buffer, draw_indexed_indirect_instruction->buffer, draw_indexed_indirect_instruction->offset, draw_indexed_indirect_instruction->draw_count, draw_indexed_indirect_instruction->stride);
+				instruction_data_cursor += sizeof(DrawListDrawIndexedIndirectInstruction);
+			} break;
 			case DrawListInstruction::TYPE_EXECUTE_COMMANDS: {
 				const DrawListExecuteCommandsInstruction *execute_commands_instruction = reinterpret_cast<const DrawListExecuteCommandsInstruction *>(instruction);
 				driver->command_buffer_execute_secondary(p_command_buffer, execute_commands_instruction->command_buffer);
@@ -1191,6 +1201,16 @@ void RenderingDeviceGraph::_print_draw_list(const uint8_t *p_instruction_data, u
 				print_line("\tDRAW INDICES", draw_indexed_instruction->index_count, "INSTANCES", draw_indexed_instruction->instance_count, "FIRST INDEX", draw_indexed_instruction->first_index);
 				instruction_data_cursor += sizeof(DrawListDrawIndexedInstruction);
 			} break;
+			case DrawListInstruction::TYPE_DRAW_INDIRECT: {
+				const DrawListDrawIndirectInstruction *draw_indirect_instruction = reinterpret_cast<const DrawListDrawIndirectInstruction *>(instruction);
+				print_line("\tDRAW INDIRECT BUFFER ID", itos(draw_indirect_instruction->buffer.id), "OFFSET", draw_indirect_instruction->offset, "DRAW COUNT", draw_indirect_instruction->draw_count, "STRIDE", draw_indirect_instruction->stride);
+				instruction_data_cursor += sizeof(DrawListDrawIndirectInstruction);
+			} break;
+			case DrawListInstruction::TYPE_DRAW_INDEXED_INDIRECT: {
+				const DrawListDrawIndexedIndirectInstruction *draw_indexed_indirect_instruction = reinterpret_cast<const DrawListDrawIndexedIndirectInstruction *>(instruction);
+				print_line("\tDRAW INDEXED INDIRECT BUFFER ID", itos(draw_indexed_indirect_instruction->buffer.id), "OFFSET", draw_indexed_indirect_instruction->offset, "DRAW COUNT", draw_indexed_indirect_instruction->draw_count, "STRIDE", draw_indexed_indirect_instruction->stride);
+				instruction_data_cursor += sizeof(DrawListDrawIndexedIndirectInstruction);
+			} break;
 			case DrawListInstruction::TYPE_EXECUTE_COMMANDS: {
 				print_line("\tEXECUTE COMMANDS");
 				instruction_data_cursor += sizeof(DrawListExecuteCommandsInstruction);
@@ -1613,6 +1633,26 @@ void RenderingDeviceGraph::add_draw_list_draw_indexed(uint32_t p_index_count, ui
 	instruction->first_index = p_first_index;
 }
 
+void RenderingDeviceGraph::add_draw_list_draw_indirect(RDD::BufferID p_buffer, uint32_t p_offset, uint32_t p_draw_count, uint32_t p_stride) {
+	DrawListDrawIndirectInstruction *instruction = reinterpret_cast<DrawListDrawIndirectInstruction *>(_allocate_draw_list_instruction(sizeof(DrawListDrawIndirectInstruction)));
+	instruction->type = DrawListInstruction::TYPE_DRAW_INDIRECT;
+	instruction->buffer = p_buffer;
+	instruction->offset = p_offset;
+	instruction->draw_count = p_draw_count;
+	instruction->stride = p_stride;
+	draw_instruction_list.stages.set_flag(RDD::PIPELINE_STAGE_DRAW_INDIRECT_BIT);
+}
+
+void RenderingDeviceGraph::add_draw_list_draw_indexed_indirect(RDD::BufferID p_buffer, uint32_t p_offset, uint32_t p_draw_count, uint32_t p_stride) {
+	DrawListDrawIndexedIndirectInstruction *instruction = reinterpret_cast<DrawListDrawIndexedIndirectInstruction *>(_allocate_draw_list_instruction(sizeof(DrawListDrawIndexedIndirectInstruction)));
+	instruction->type = DrawListInstruction::TYPE_DRAW_INDEXED_INDIRECT;
+	instruction->buffer = p_buffer;
+	instruction->offset = p_offset;
+	instruction->draw_count = p_draw_count;
+	instruction->stride = p_stride;
+	draw_instruction_list.stages.set_flag(RDD::PIPELINE_STAGE_DRAW_INDIRECT_BIT);
+}
+
 void RenderingDeviceGraph::add_draw_list_execute_commands(RDD::CommandBufferID p_command_buffer) {
 	DrawListExecuteCommandsInstruction *instruction = reinterpret_cast<DrawListExecuteCommandsInstruction *>(_allocate_draw_list_instruction(sizeof(DrawListExecuteCommandsInstruction)));
 	instruction->type = DrawListInstruction::TYPE_EXECUTE_COMMANDS;
diff --git a/servers/rendering/rendering_device_graph.h b/servers/rendering/rendering_device_graph.h
index f4d2e8534f..0148c6b78d 100644
--- a/servers/rendering/rendering_device_graph.h
+++ b/servers/rendering/rendering_device_graph.h
@@ -71,6 +71,8 @@ public:
 			TYPE_CLEAR_ATTACHMENTS,
 			TYPE_DRAW,
 			TYPE_DRAW_INDEXED,
+			TYPE_DRAW_INDIRECT,
+			TYPE_DRAW_INDEXED_INDIRECT,
 			TYPE_EXECUTE_COMMANDS,
 			TYPE_NEXT_SUBPASS,
 			TYPE_SET_BLEND_CONSTANTS,
@@ -474,6 +476,20 @@ private:
 		uint32_t first_index = 0;
 	};
 
+	struct DrawListDrawIndirectInstruction : DrawListInstruction {
+		RDD::BufferID buffer;
+		uint32_t offset = 0;
+		uint32_t draw_count = 0;
+		uint32_t stride = 0;
+	};
+
+	struct DrawListDrawIndexedIndirectInstruction : DrawListInstruction {
+		RDD::BufferID buffer;
+		uint32_t offset = 0;
+		uint32_t draw_count = 0;
+		uint32_t stride = 0;
+	};
+
 	struct DrawListEndRenderPassInstruction : DrawListInstruction {
 		// No contents.
 	};
@@ -686,6 +702,8 @@ public:
 	void add_draw_list_clear_attachments(VectorView<RDD::AttachmentClear> p_attachments_clear, VectorView<Rect2i> p_attachments_clear_rect);
 	void add_draw_list_draw(uint32_t p_vertex_count, uint32_t p_instance_count);
 	void add_draw_list_draw_indexed(uint32_t p_index_count, uint32_t p_instance_count, uint32_t p_first_index);
+	void add_draw_list_draw_indirect(RDD::BufferID p_buffer, uint32_t p_offset, uint32_t p_draw_count, uint32_t p_stride);
+	void add_draw_list_draw_indexed_indirect(RDD::BufferID p_buffer, uint32_t p_offset, uint32_t p_draw_count, uint32_t p_stride);
 	void add_draw_list_execute_commands(RDD::CommandBufferID p_command_buffer);
 	void add_draw_list_next_subpass(RDD::CommandBufferType p_command_buffer_type);
 	void add_draw_list_set_blend_constants(const Color &p_color);
diff --git a/servers/rendering/storage/camera_attributes_storage.h b/servers/rendering/storage/camera_attributes_storage.h
index 05a44c38e7..0fff7cf72f 100644
--- a/servers/rendering/storage/camera_attributes_storage.h
+++ b/servers/rendering/storage/camera_attributes_storage.h
@@ -74,8 +74,8 @@ public:
 	RendererCameraAttributes();
 	~RendererCameraAttributes();
 
-	CameraAttributes *get_camera_attributes(RID p_rid) { return camera_attributes_owner.get_or_null(p_rid); };
-	bool owns_camera_attributes(RID p_rid) { return camera_attributes_owner.owns(p_rid); };
+	CameraAttributes *get_camera_attributes(RID p_rid) { return camera_attributes_owner.get_or_null(p_rid); }
+	bool owns_camera_attributes(RID p_rid) { return camera_attributes_owner.owns(p_rid); }
 
 	RID camera_attributes_allocate();
 	void camera_attributes_initialize(RID p_rid);
diff --git a/servers/rendering/storage/render_scene_buffers.cpp b/servers/rendering/storage/render_scene_buffers.cpp
index 3894559dc9..fd4131e1fa 100644
--- a/servers/rendering/storage/render_scene_buffers.cpp
+++ b/servers/rendering/storage/render_scene_buffers.cpp
@@ -83,7 +83,7 @@ void RenderSceneBuffersExtension::_bind_methods() {
 
 void RenderSceneBuffersExtension::configure(const RenderSceneBuffersConfiguration *p_config) {
 	GDVIRTUAL_CALL(_configure, p_config);
-};
+}
 
 void RenderSceneBuffersExtension::set_fsr_sharpness(float p_fsr_sharpness) {
 	GDVIRTUAL_CALL(_set_fsr_sharpness, p_fsr_sharpness);