Add draw indirect to Rendering Device

doc/classes/RenderingDevice.xml

@@ -324,6 +324,18 @@
 				Submits [param draw_list] for rendering on the GPU. This is the raster equivalent to [method compute_list_dispatch].
 			</description>
 		</method>
+		<method name="draw_list_draw_indirect">
+			<return type="void" />
+			<param index="0" name="draw_list" type="int" />
+			<param index="1" name="use_indices" type="bool" />
+			<param index="2" name="buffer" type="RID" />
+			<param index="3" name="offset" type="int" default="0" />
+			<param index="4" name="draw_count" type="int" default="1" />
+			<param index="5" name="stride" type="int" default="0" />
+			<description>
+				Submits [param draw_list] for rendering on the GPU with the given parameters stored in the [param buffer] at [param offset]. Parameters being integers: vertex count, instance count, first vertex, first instance. And when using indices: index count, instance count, first index, vertex offset, first instance. Buffer must have been created with [constant STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT] flag.
+			</description>
+		</method>
 		<method name="draw_list_enable_scissor">
 			<return type="void" />
 			<param index="0" name="draw_list" type="int" />

servers/rendering/rendering_device.cpp

@@ -4191,6 +4191,117 @@ 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) {
 	DrawList *dl = _get_draw_list_ptr(p_list);
 
@@ -6023,6 +6134,7 @@ 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);

servers/rendering/rendering_device.h

@@ -1172,6 +1172,7 @@ 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);

servers/rendering/rendering_device_graph.cpp

@@ -699,6 +699,16 @@ 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);
@@ -1178,6 +1188,16 @@ 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);
@@ -1596,6 +1616,26 @@ 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;

servers/rendering/rendering_device_graph.h

@@ -69,6 +69,8 @@ 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,
@@ -463,6 +465,20 @@ 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.
 	};
@@ -675,6 +691,8 @@ 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);