This PR implements #126353; In short: keep discard list as part of swap chain images. This allows better determination when resources are actually not in use anymore. ## Resource pool Resource pools keep track of the resources for a swap chain image. In Blender this is a bit more complicated due to the way GPUContext work. A single thread can have multiple contexts. Some of them have a swap chain (GHOST Window) other don't (draw manager). The resource pool should be shared between the contexts running on the same thread. When opening multiple windows there are also multiple swap chains to consider. ### Discard pile Resource handles that are deleted and stored in the discard pile. When we are sure that these resources are not used on the GPU anymore these are destroyed. ### Reusable resources There are other resources as well like: - Descriptor sets - Descriptor pools ## Open issues There are some limitations that require future PRs to fix including: - Background rendering - Handling multiple windows - Improve CPU/GPU synchronization - Reuse staging buffers Pull Request: https://projects.blender.org/blender/blender/pulls/126353
299 lines
8.9 KiB
C++
299 lines
8.9 KiB
C++
/* SPDX-FileCopyrightText: 2022 Blender Authors
|
|
*
|
|
* SPDX-License-Identifier: GPL-2.0-or-later */
|
|
|
|
/** \file
|
|
* \ingroup gpu
|
|
*/
|
|
|
|
#include "GHOST_C-api.h"
|
|
|
|
#include "gpu_capabilities_private.hh"
|
|
#include "gpu_platform_private.hh"
|
|
|
|
#include "vk_batch.hh"
|
|
#include "vk_context.hh"
|
|
#include "vk_drawlist.hh"
|
|
#include "vk_fence.hh"
|
|
#include "vk_framebuffer.hh"
|
|
#include "vk_index_buffer.hh"
|
|
#include "vk_pixel_buffer.hh"
|
|
#include "vk_query.hh"
|
|
#include "vk_shader.hh"
|
|
#include "vk_state_manager.hh"
|
|
#include "vk_storage_buffer.hh"
|
|
#include "vk_texture.hh"
|
|
#include "vk_uniform_buffer.hh"
|
|
#include "vk_vertex_buffer.hh"
|
|
|
|
#include "vk_backend.hh"
|
|
|
|
namespace blender::gpu {
|
|
|
|
static eGPUOSType determine_os_type()
|
|
{
|
|
#ifdef _WIN32
|
|
return GPU_OS_WIN;
|
|
#elif defined(__APPLE__)
|
|
return GPU_OS_MAC;
|
|
#else
|
|
return GPU_OS_UNIX;
|
|
#endif
|
|
}
|
|
|
|
void VKBackend::platform_init()
|
|
{
|
|
GPG.init(GPU_DEVICE_ANY,
|
|
determine_os_type(),
|
|
GPU_DRIVER_ANY,
|
|
GPU_SUPPORT_LEVEL_SUPPORTED,
|
|
GPU_BACKEND_VULKAN,
|
|
"",
|
|
"",
|
|
"",
|
|
GPU_ARCHITECTURE_IMR);
|
|
}
|
|
|
|
void VKBackend::platform_init(const VKDevice &device)
|
|
{
|
|
const VkPhysicalDeviceProperties &properties = device.physical_device_properties_get();
|
|
|
|
eGPUDeviceType device_type = device.device_type();
|
|
eGPUOSType os = determine_os_type();
|
|
eGPUDriverType driver = GPU_DRIVER_ANY;
|
|
eGPUSupportLevel support_level = GPU_SUPPORT_LEVEL_SUPPORTED;
|
|
|
|
std::string vendor_name = device.vendor_name();
|
|
std::string driver_version = device.driver_version();
|
|
|
|
GPG.init(device_type,
|
|
os,
|
|
driver,
|
|
support_level,
|
|
GPU_BACKEND_VULKAN,
|
|
vendor_name.c_str(),
|
|
properties.deviceName,
|
|
driver_version.c_str(),
|
|
GPU_ARCHITECTURE_IMR);
|
|
}
|
|
|
|
void VKBackend::detect_workarounds(VKDevice &device)
|
|
{
|
|
VKWorkarounds workarounds;
|
|
|
|
if (G.debug & G_DEBUG_GPU_FORCE_WORKAROUNDS) {
|
|
printf("\n");
|
|
printf("VK: Forcing workaround usage and disabling features and extensions.\n");
|
|
printf(" Vendor: %s\n", device.vendor_name().c_str());
|
|
printf(" Device: %s\n", device.physical_device_properties_get().deviceName);
|
|
printf(" Driver: %s\n", device.driver_version().c_str());
|
|
/* Force workarounds. */
|
|
workarounds.not_aligned_pixel_formats = true;
|
|
workarounds.shader_output_layer = true;
|
|
workarounds.shader_output_viewport_index = true;
|
|
workarounds.vertex_formats.r8g8b8 = true;
|
|
|
|
device.workarounds_ = workarounds;
|
|
return;
|
|
}
|
|
|
|
workarounds.shader_output_layer =
|
|
!device.physical_device_vulkan_12_features_get().shaderOutputLayer;
|
|
workarounds.shader_output_viewport_index =
|
|
!device.physical_device_vulkan_12_features_get().shaderOutputViewportIndex;
|
|
|
|
/* AMD GPUs don't support texture formats that use are aligned to 24 or 48 bits. */
|
|
if (GPU_type_matches(GPU_DEVICE_ATI, GPU_OS_ANY, GPU_DRIVER_ANY) ||
|
|
GPU_type_matches(GPU_DEVICE_APPLE, GPU_OS_MAC, GPU_DRIVER_ANY))
|
|
{
|
|
workarounds.not_aligned_pixel_formats = true;
|
|
}
|
|
|
|
VkFormatProperties format_properties = {};
|
|
vkGetPhysicalDeviceFormatProperties(
|
|
device.physical_device_get(), VK_FORMAT_R8G8B8_UNORM, &format_properties);
|
|
workarounds.vertex_formats.r8g8b8 = (format_properties.bufferFeatures &
|
|
VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT) == 0;
|
|
|
|
device.workarounds_ = workarounds;
|
|
}
|
|
|
|
void VKBackend::platform_exit()
|
|
{
|
|
GPG.clear();
|
|
VKDevice &device = VKBackend::get().device;
|
|
if (device.is_initialized()) {
|
|
device.deinit();
|
|
}
|
|
}
|
|
|
|
void VKBackend::delete_resources() {}
|
|
|
|
void VKBackend::samplers_update()
|
|
{
|
|
VKDevice &device = VKBackend::get().device;
|
|
if (device.is_initialized()) {
|
|
device.reinit();
|
|
}
|
|
}
|
|
|
|
void VKBackend::compute_dispatch(int groups_x_len, int groups_y_len, int groups_z_len)
|
|
{
|
|
VKContext &context = *VKContext::get();
|
|
render_graph::VKResourceAccessInfo &resources = context.update_and_get_access_info();
|
|
render_graph::VKDispatchNode::CreateInfo dispatch_info(resources);
|
|
context.update_pipeline_data(dispatch_info.dispatch_node.pipeline_data);
|
|
dispatch_info.dispatch_node.group_count_x = groups_x_len;
|
|
dispatch_info.dispatch_node.group_count_y = groups_y_len;
|
|
dispatch_info.dispatch_node.group_count_z = groups_z_len;
|
|
context.render_graph.add_node(dispatch_info);
|
|
}
|
|
|
|
void VKBackend::compute_dispatch_indirect(StorageBuf *indirect_buf)
|
|
{
|
|
BLI_assert(indirect_buf);
|
|
VKContext &context = *VKContext::get();
|
|
VKStorageBuffer &indirect_buffer = *unwrap(indirect_buf);
|
|
render_graph::VKResourceAccessInfo &resources = context.update_and_get_access_info();
|
|
render_graph::VKDispatchIndirectNode::CreateInfo dispatch_indirect_info(resources);
|
|
context.update_pipeline_data(dispatch_indirect_info.dispatch_indirect_node.pipeline_data);
|
|
dispatch_indirect_info.dispatch_indirect_node.buffer = indirect_buffer.vk_handle();
|
|
dispatch_indirect_info.dispatch_indirect_node.offset = 0;
|
|
context.render_graph.add_node(dispatch_indirect_info);
|
|
}
|
|
|
|
Context *VKBackend::context_alloc(void *ghost_window, void *ghost_context)
|
|
{
|
|
if (ghost_window) {
|
|
BLI_assert(ghost_context == nullptr);
|
|
ghost_context = GHOST_GetDrawingContext((GHOST_WindowHandle)ghost_window);
|
|
}
|
|
|
|
BLI_assert(ghost_context != nullptr);
|
|
if (!device.is_initialized()) {
|
|
device.init(ghost_context);
|
|
}
|
|
|
|
VKContext *context = new VKContext(ghost_window, ghost_context, device.current_thread_data());
|
|
device.context_register(*context);
|
|
GHOST_SetVulkanSwapBuffersCallbacks((GHOST_ContextHandle)ghost_context,
|
|
VKContext::swap_buffers_pre_callback,
|
|
VKContext::swap_buffers_post_callback);
|
|
return context;
|
|
}
|
|
|
|
Batch *VKBackend::batch_alloc()
|
|
{
|
|
return new VKBatch();
|
|
}
|
|
|
|
DrawList *VKBackend::drawlist_alloc(int list_length)
|
|
{
|
|
return new VKDrawList(list_length);
|
|
}
|
|
|
|
Fence *VKBackend::fence_alloc()
|
|
{
|
|
return new VKFence();
|
|
}
|
|
|
|
FrameBuffer *VKBackend::framebuffer_alloc(const char *name)
|
|
{
|
|
return new VKFrameBuffer(name);
|
|
}
|
|
|
|
IndexBuf *VKBackend::indexbuf_alloc()
|
|
{
|
|
return new VKIndexBuffer();
|
|
}
|
|
|
|
PixelBuffer *VKBackend::pixelbuf_alloc(size_t size)
|
|
{
|
|
return new VKPixelBuffer(size);
|
|
}
|
|
|
|
QueryPool *VKBackend::querypool_alloc()
|
|
{
|
|
return new VKQueryPool();
|
|
}
|
|
|
|
Shader *VKBackend::shader_alloc(const char *name)
|
|
{
|
|
return new VKShader(name);
|
|
}
|
|
|
|
Texture *VKBackend::texture_alloc(const char *name)
|
|
{
|
|
return new VKTexture(name);
|
|
}
|
|
|
|
UniformBuf *VKBackend::uniformbuf_alloc(size_t size, const char *name)
|
|
{
|
|
return new VKUniformBuffer(size, name);
|
|
}
|
|
|
|
StorageBuf *VKBackend::storagebuf_alloc(size_t size, GPUUsageType usage, const char *name)
|
|
{
|
|
return new VKStorageBuffer(size, usage, name);
|
|
}
|
|
|
|
VertBuf *VKBackend::vertbuf_alloc()
|
|
{
|
|
return new VKVertexBuffer();
|
|
}
|
|
|
|
void VKBackend::render_begin() {}
|
|
|
|
void VKBackend::render_end() {}
|
|
|
|
void VKBackend::render_step() {}
|
|
|
|
shaderc::Compiler &VKBackend::get_shaderc_compiler()
|
|
{
|
|
return shaderc_compiler_;
|
|
}
|
|
|
|
void VKBackend::capabilities_init(VKDevice &device)
|
|
{
|
|
const VkPhysicalDeviceProperties &properties = device.physical_device_properties_get();
|
|
const VkPhysicalDeviceLimits &limits = properties.limits;
|
|
|
|
/* Reset all capabilities from previous context. */
|
|
GCaps = {};
|
|
GCaps.geometry_shader_support = true;
|
|
GCaps.texture_view_support = true;
|
|
GCaps.stencil_export_support = device.supports_extension(
|
|
VK_EXT_SHADER_STENCIL_EXPORT_EXTENSION_NAME);
|
|
GCaps.shader_draw_parameters_support =
|
|
device.physical_device_vulkan_11_features_get().shaderDrawParameters;
|
|
|
|
GCaps.max_texture_size = max_ii(limits.maxImageDimension1D, limits.maxImageDimension2D);
|
|
GCaps.max_texture_3d_size = limits.maxImageDimension3D;
|
|
GCaps.max_texture_layers = limits.maxImageArrayLayers;
|
|
GCaps.max_textures = limits.maxDescriptorSetSampledImages;
|
|
GCaps.max_textures_vert = limits.maxPerStageDescriptorSampledImages;
|
|
GCaps.max_textures_geom = limits.maxPerStageDescriptorSampledImages;
|
|
GCaps.max_textures_frag = limits.maxPerStageDescriptorSampledImages;
|
|
GCaps.max_samplers = limits.maxSamplerAllocationCount;
|
|
GCaps.max_images = limits.maxPerStageDescriptorStorageImages;
|
|
for (int i = 0; i < 3; i++) {
|
|
GCaps.max_work_group_count[i] = limits.maxComputeWorkGroupCount[i];
|
|
GCaps.max_work_group_size[i] = limits.maxComputeWorkGroupSize[i];
|
|
}
|
|
GCaps.max_uniforms_vert = limits.maxPerStageDescriptorUniformBuffers;
|
|
GCaps.max_uniforms_frag = limits.maxPerStageDescriptorUniformBuffers;
|
|
GCaps.max_batch_indices = limits.maxDrawIndirectCount;
|
|
GCaps.max_batch_vertices = limits.maxDrawIndexedIndexValue;
|
|
GCaps.max_vertex_attribs = limits.maxVertexInputAttributes;
|
|
GCaps.max_varying_floats = limits.maxVertexOutputComponents;
|
|
GCaps.max_shader_storage_buffer_bindings = limits.maxPerStageDescriptorStorageBuffers;
|
|
GCaps.max_compute_shader_storage_blocks = limits.maxPerStageDescriptorStorageBuffers;
|
|
GCaps.max_storage_buffer_size = size_t(limits.maxStorageBufferRange);
|
|
|
|
GCaps.mem_stats_support = true;
|
|
|
|
detect_workarounds(device);
|
|
}
|
|
|
|
} // namespace blender::gpu
|