Files
test2/source/blender/gpu/vulkan/vk_pipeline.cc
Jeroen Bakker f35b0373d6 Vulkan: Separate DataTransfer, Compute, Graphics Commands
Goal is to reduce the number of command buffer flushes by tracking what is
happening in the different command queues. This is an initial step towards
advanced queue-ing strategies.

The new (intermediate) strategy records commands to different command
buffers based on what they do. There is a command buffer for data transfers,
compute pipelines and graphics pipelines.

When a compute command is recorded it ensures that all graphic commands
are finished. When a graphic command is recorded it ensures all compute
commands are finished. When a graphic or compute command is scheduled
all recorded data transfer commands are scheduled as well.

Some improvements are expected as multiple compute and data transfers
commands can now be scheduled at the same time and don't need to unbind
and rebind render passes. Especially when using EEVEE-Next which is
compute centric the performance change is visible for the user.

Pull Request: https://projects.blender.org/blender/blender/pulls/114104
2023-10-30 14:21:14 +01:00

219 lines
8.7 KiB
C++

/* SPDX-FileCopyrightText: 2023 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
/** \file
* \ingroup gpu
*/
#include "vk_pipeline.hh"
#include "vk_backend.hh"
#include "vk_batch.hh"
#include "vk_context.hh"
#include "vk_framebuffer.hh"
#include "vk_memory.hh"
#include "vk_state_manager.hh"
#include "vk_vertex_attribute_object.hh"
namespace blender::gpu {
VKPipeline::VKPipeline(VkDescriptorSetLayout vk_descriptor_set_layout,
VKPushConstants &&push_constants)
: active_vk_pipeline_(VK_NULL_HANDLE),
descriptor_set_(vk_descriptor_set_layout),
push_constants_(std::move(push_constants))
{
}
VKPipeline::VKPipeline(VkPipeline vk_pipeline,
VkDescriptorSetLayout vk_descriptor_set_layout,
VKPushConstants &&push_constants)
: active_vk_pipeline_(vk_pipeline),
descriptor_set_(vk_descriptor_set_layout),
push_constants_(std::move(push_constants))
{
vk_pipelines_.append(vk_pipeline);
}
VKPipeline::~VKPipeline()
{
VK_ALLOCATION_CALLBACKS
const VKDevice &device = VKBackend::get().device_get();
for (VkPipeline vk_pipeline : vk_pipelines_) {
vkDestroyPipeline(device.device_get(), vk_pipeline, vk_allocation_callbacks);
}
}
VKPipeline VKPipeline::create_compute_pipeline(
VkShaderModule compute_module,
VkDescriptorSetLayout &descriptor_set_layout,
VkPipelineLayout &pipeline_layout,
const VKPushConstants::Layout &push_constants_layout)
{
VK_ALLOCATION_CALLBACKS
const VKDevice &device = VKBackend::get().device_get();
VkComputePipelineCreateInfo pipeline_info = {};
pipeline_info.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
pipeline_info.flags = 0;
pipeline_info.stage = {};
pipeline_info.stage.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
pipeline_info.stage.flags = 0;
pipeline_info.stage.stage = VK_SHADER_STAGE_COMPUTE_BIT;
pipeline_info.stage.module = compute_module;
pipeline_info.layout = pipeline_layout;
pipeline_info.stage.pName = "main";
VkPipeline vk_pipeline;
if (vkCreateComputePipelines(device.device_get(),
nullptr,
1,
&pipeline_info,
vk_allocation_callbacks,
&vk_pipeline) != VK_SUCCESS)
{
return VKPipeline();
}
VKPushConstants push_constants(&push_constants_layout);
return VKPipeline(vk_pipeline, descriptor_set_layout, std::move(push_constants));
}
VKPipeline VKPipeline::create_graphics_pipeline(
VkDescriptorSetLayout &descriptor_set_layout,
const VKPushConstants::Layout &push_constants_layout)
{
VKPushConstants push_constants(&push_constants_layout);
return VKPipeline(descriptor_set_layout, std::move(push_constants));
}
VkPipeline VKPipeline::vk_handle() const
{
return active_vk_pipeline_;
}
bool VKPipeline::is_valid() const
{
return active_vk_pipeline_ != VK_NULL_HANDLE;
}
void VKPipeline::finalize(VKContext &context,
VkShaderModule vertex_module,
VkShaderModule geometry_module,
VkShaderModule fragment_module,
VkPipelineLayout &pipeline_layout,
const GPUPrimType prim_type,
const VKVertexAttributeObject &vertex_attribute_object)
{
BLI_assert(vertex_module != VK_NULL_HANDLE);
VK_ALLOCATION_CALLBACKS
Vector<VkPipelineShaderStageCreateInfo> pipeline_stages;
VkPipelineShaderStageCreateInfo vertex_stage_info = {};
vertex_stage_info.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
vertex_stage_info.stage = VK_SHADER_STAGE_VERTEX_BIT;
vertex_stage_info.module = vertex_module;
vertex_stage_info.pName = "main";
pipeline_stages.append(vertex_stage_info);
if (geometry_module != VK_NULL_HANDLE) {
VkPipelineShaderStageCreateInfo geometry_stage_info = {};
geometry_stage_info.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
geometry_stage_info.stage = VK_SHADER_STAGE_GEOMETRY_BIT;
geometry_stage_info.module = geometry_module;
geometry_stage_info.pName = "main";
pipeline_stages.append(geometry_stage_info);
}
if (fragment_module != VK_NULL_HANDLE) {
VkPipelineShaderStageCreateInfo fragment_stage_info = {};
fragment_stage_info.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
fragment_stage_info.stage = VK_SHADER_STAGE_FRAGMENT_BIT;
fragment_stage_info.module = fragment_module;
fragment_stage_info.pName = "main";
pipeline_stages.append(fragment_stage_info);
}
VKFrameBuffer &framebuffer = *context.active_framebuffer_get();
framebuffer.vk_render_pass_ensure();
VkGraphicsPipelineCreateInfo pipeline_create_info = {};
pipeline_create_info.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
pipeline_create_info.stageCount = pipeline_stages.size();
pipeline_create_info.pStages = pipeline_stages.data();
pipeline_create_info.layout = pipeline_layout;
pipeline_create_info.renderPass = framebuffer.vk_render_pass_get();
pipeline_create_info.subpass = 0;
/* Vertex input state. */
VkPipelineVertexInputStateCreateInfo vertex_input_state = {};
vertex_input_state.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
vertex_input_state.vertexBindingDescriptionCount = vertex_attribute_object.bindings.size();
vertex_input_state.pVertexBindingDescriptions = vertex_attribute_object.bindings.data();
vertex_input_state.vertexAttributeDescriptionCount = vertex_attribute_object.attributes.size();
vertex_input_state.pVertexAttributeDescriptions = vertex_attribute_object.attributes.data();
pipeline_create_info.pVertexInputState = &vertex_input_state;
/* Input assembly state. */
VkPipelineInputAssemblyStateCreateInfo pipeline_input_assembly = {};
pipeline_input_assembly.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
pipeline_input_assembly.topology = to_vk_primitive_topology(prim_type);
pipeline_input_assembly.primitiveRestartEnable =
ELEM(prim_type, GPU_PRIM_TRIS, GPU_PRIM_LINES, GPU_PRIM_POINTS) ? VK_FALSE : VK_TRUE;
pipeline_create_info.pInputAssemblyState = &pipeline_input_assembly;
/* Viewport state. */
VkPipelineViewportStateCreateInfo viewport_state = {};
viewport_state.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
Array<VkViewport, 16> viewports = framebuffer.vk_viewports_get();
viewport_state.pViewports = &viewports[0];
viewport_state.viewportCount = viewports.size();
Array<VkRect2D, 16> scissors = framebuffer.vk_render_areas_get();
viewport_state.pScissors = &scissors[0];
viewport_state.scissorCount = scissors.size();
pipeline_create_info.pViewportState = &viewport_state;
/* Multi-sample state. */
VkPipelineMultisampleStateCreateInfo multisample_state = {};
multisample_state.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
multisample_state.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
multisample_state.minSampleShading = 1.0f;
pipeline_create_info.pMultisampleState = &multisample_state;
/* States from the state manager. */
VKPipelineStateManager &state_manager = state_manager_get();
state_manager.finalize_color_blend_state(framebuffer);
pipeline_create_info.pColorBlendState = &state_manager.pipeline_color_blend_state;
pipeline_create_info.pRasterizationState = &state_manager.rasterization_state;
pipeline_create_info.pDepthStencilState = &state_manager.depth_stencil_state;
const VKDevice &device = VKBackend::get().device_get();
vkCreateGraphicsPipelines(device.device_get(),
VK_NULL_HANDLE,
1,
&pipeline_create_info,
vk_allocation_callbacks,
&active_vk_pipeline_);
/* TODO: we should cache several pipeline instances and detect pipelines we can reuse. This might
* also be done using a VkPipelineCache. For now we just destroy any available pipeline so it
* won't be overwritten by the newly created one. */
vk_pipelines_.append(active_vk_pipeline_);
debug::object_label(active_vk_pipeline_, "GraphicsPipeline");
}
void VKPipeline::update_and_bind(VKContext &context,
VkPipelineLayout vk_pipeline_layout,
VkPipelineBindPoint vk_pipeline_bind_point)
{
VKCommandBuffers &command_buffers = context.command_buffers_get();
command_buffers.bind(*this, vk_pipeline_bind_point);
push_constants_.update(context);
if (descriptor_set_.has_layout()) {
descriptor_set_.update(context);
command_buffers.bind(
*descriptor_set_.active_descriptor_set(), vk_pipeline_layout, vk_pipeline_bind_point);
}
}
} // namespace blender::gpu