During normal execution it isn't possible to switch workarounds. However when running test cases we want to check if shaders and other tests work when work arounds are enabled. Currently shader patches are cached globally. This PR moves the cached shader patch to the device level which is rebuild every time a device needs to be reinitialized. For OpenGL this is also an issue, but harder to solve as the concept device doesn't exist there. Pull Request: https://projects.blender.org/blender/blender/pulls/116042
407 lines
12 KiB
C++
407 lines
12 KiB
C++
/* SPDX-FileCopyrightText: 2023 Blender Authors
|
|
*
|
|
* SPDX-License-Identifier: GPL-2.0-or-later */
|
|
|
|
/** \file
|
|
* \ingroup gpu
|
|
*/
|
|
|
|
#include <sstream>
|
|
|
|
#include "vk_backend.hh"
|
|
#include "vk_context.hh"
|
|
#include "vk_device.hh"
|
|
#include "vk_memory.hh"
|
|
#include "vk_state_manager.hh"
|
|
#include "vk_storage_buffer.hh"
|
|
#include "vk_texture.hh"
|
|
#include "vk_vertex_buffer.hh"
|
|
|
|
#include "GPU_capabilities.h"
|
|
|
|
#include "BLI_math_matrix_types.hh"
|
|
|
|
#include "GHOST_C-api.h"
|
|
|
|
extern "C" char datatoc_glsl_shader_defines_glsl[];
|
|
|
|
namespace blender::gpu {
|
|
|
|
void VKDevice::deinit()
|
|
{
|
|
VK_ALLOCATION_CALLBACKS
|
|
if (!is_initialized()) {
|
|
return;
|
|
}
|
|
|
|
timeline_semaphore_.free(*this);
|
|
dummy_buffer_.free();
|
|
if (dummy_color_attachment_.has_value()) {
|
|
delete &(*dummy_color_attachment_).get();
|
|
dummy_color_attachment_.reset();
|
|
}
|
|
samplers_.free();
|
|
destroy_discarded_resources();
|
|
vkDestroyPipelineCache(vk_device_, vk_pipeline_cache_, vk_allocation_callbacks);
|
|
vmaDestroyAllocator(mem_allocator_);
|
|
mem_allocator_ = VK_NULL_HANDLE;
|
|
|
|
debugging_tools_.deinit(vk_instance_);
|
|
|
|
vk_instance_ = VK_NULL_HANDLE;
|
|
vk_physical_device_ = VK_NULL_HANDLE;
|
|
vk_device_ = VK_NULL_HANDLE;
|
|
vk_queue_family_ = 0;
|
|
vk_queue_ = VK_NULL_HANDLE;
|
|
vk_physical_device_properties_ = {};
|
|
glsl_patch_.clear();
|
|
}
|
|
|
|
bool VKDevice::is_initialized() const
|
|
{
|
|
return vk_device_ != VK_NULL_HANDLE;
|
|
}
|
|
|
|
void VKDevice::init(void *ghost_context)
|
|
{
|
|
BLI_assert(!is_initialized());
|
|
GHOST_GetVulkanHandles((GHOST_ContextHandle)ghost_context,
|
|
&vk_instance_,
|
|
&vk_physical_device_,
|
|
&vk_device_,
|
|
&vk_queue_family_,
|
|
&vk_queue_);
|
|
|
|
init_physical_device_properties();
|
|
init_physical_device_memory_properties();
|
|
init_physical_device_features();
|
|
VKBackend::platform_init(*this);
|
|
VKBackend::capabilities_init(*this);
|
|
init_debug_callbacks();
|
|
init_memory_allocator();
|
|
init_pipeline_cache();
|
|
|
|
samplers_.init();
|
|
timeline_semaphore_.init(*this);
|
|
|
|
debug::object_label(device_get(), "LogicalDevice");
|
|
debug::object_label(queue_get(), "GenericQueue");
|
|
init_glsl_patch();
|
|
}
|
|
|
|
void VKDevice::init_debug_callbacks()
|
|
{
|
|
debugging_tools_.init(vk_instance_);
|
|
}
|
|
|
|
void VKDevice::init_physical_device_properties()
|
|
{
|
|
BLI_assert(vk_physical_device_ != VK_NULL_HANDLE);
|
|
vkGetPhysicalDeviceProperties(vk_physical_device_, &vk_physical_device_properties_);
|
|
}
|
|
|
|
void VKDevice::init_physical_device_memory_properties()
|
|
{
|
|
BLI_assert(vk_physical_device_ != VK_NULL_HANDLE);
|
|
vkGetPhysicalDeviceMemoryProperties(vk_physical_device_, &vk_physical_device_memory_properties_);
|
|
}
|
|
|
|
void VKDevice::init_physical_device_features()
|
|
{
|
|
BLI_assert(vk_physical_device_ != VK_NULL_HANDLE);
|
|
|
|
VkPhysicalDeviceFeatures2 features = {};
|
|
features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
|
|
vk_physical_device_vulkan_11_features_.sType =
|
|
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES;
|
|
vk_physical_device_vulkan_12_features_.sType =
|
|
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES;
|
|
|
|
features.pNext = &vk_physical_device_vulkan_11_features_;
|
|
vk_physical_device_vulkan_11_features_.pNext = &vk_physical_device_vulkan_12_features_;
|
|
|
|
vkGetPhysicalDeviceFeatures2(vk_physical_device_, &features);
|
|
vk_physical_device_features_ = features.features;
|
|
}
|
|
|
|
void VKDevice::init_memory_allocator()
|
|
{
|
|
VK_ALLOCATION_CALLBACKS;
|
|
VmaAllocatorCreateInfo info = {};
|
|
info.vulkanApiVersion = VK_API_VERSION_1_2;
|
|
info.physicalDevice = vk_physical_device_;
|
|
info.device = vk_device_;
|
|
info.instance = vk_instance_;
|
|
info.pAllocationCallbacks = vk_allocation_callbacks;
|
|
vmaCreateAllocator(&info, &mem_allocator_);
|
|
}
|
|
|
|
void VKDevice::init_pipeline_cache()
|
|
{
|
|
VK_ALLOCATION_CALLBACKS;
|
|
VkPipelineCacheCreateInfo create_info = {};
|
|
create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;
|
|
vkCreatePipelineCache(vk_device_, &create_info, vk_allocation_callbacks, &vk_pipeline_cache_);
|
|
}
|
|
|
|
void VKDevice::init_dummy_buffer(VKContext &context)
|
|
{
|
|
if (dummy_buffer_.is_allocated()) {
|
|
return;
|
|
}
|
|
|
|
dummy_buffer_.create(sizeof(float4x4),
|
|
GPU_USAGE_DEVICE_ONLY,
|
|
VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT);
|
|
dummy_buffer_.clear(context, 0);
|
|
}
|
|
|
|
void VKDevice::init_dummy_color_attachment()
|
|
{
|
|
if (dummy_color_attachment_.has_value()) {
|
|
return;
|
|
}
|
|
|
|
GPUTexture *texture = GPU_texture_create_2d(
|
|
"dummy_attachment", 1, 1, 1, GPU_R32F, GPU_TEXTURE_USAGE_ATTACHMENT, nullptr);
|
|
BLI_assert(texture);
|
|
VKTexture &vk_texture = *unwrap(unwrap(texture));
|
|
dummy_color_attachment_ = std::make_optional(std::reference_wrapper(vk_texture));
|
|
}
|
|
|
|
void VKDevice::init_glsl_patch()
|
|
{
|
|
std::stringstream ss;
|
|
|
|
ss << "#version 450\n";
|
|
if (GPU_shader_draw_parameters_support()) {
|
|
ss << "#extension GL_ARB_shader_draw_parameters : enable\n";
|
|
ss << "#define GPU_ARB_shader_draw_parameters\n";
|
|
ss << "#define gpu_BaseInstance (gl_BaseInstanceARB)\n";
|
|
}
|
|
|
|
ss << "#define gl_VertexID gl_VertexIndex\n";
|
|
ss << "#define gpu_InstanceIndex (gl_InstanceIndex)\n";
|
|
ss << "#define GPU_ARB_texture_cube_map_array\n";
|
|
ss << "#define gl_InstanceID (gpu_InstanceIndex - gpu_BaseInstance)\n";
|
|
|
|
/* TODO(fclem): This creates a validation error and should be already part of Vulkan 1.2. */
|
|
ss << "#extension GL_ARB_shader_viewport_layer_array: enable\n";
|
|
if (!workarounds_.shader_output_layer) {
|
|
ss << "#define gpu_Layer gl_Layer\n";
|
|
}
|
|
if (!workarounds_.shader_output_viewport_index) {
|
|
ss << "#define gpu_ViewportIndex gl_ViewportIndex\n";
|
|
}
|
|
|
|
ss << "#define DFDX_SIGN 1.0\n";
|
|
ss << "#define DFDY_SIGN 1.0\n";
|
|
|
|
/* GLSL Backend Lib. */
|
|
ss << datatoc_glsl_shader_defines_glsl;
|
|
glsl_patch_ = ss.str();
|
|
}
|
|
|
|
const char *VKDevice::glsl_patch_get() const
|
|
{
|
|
BLI_assert(!glsl_patch_.empty());
|
|
return glsl_patch_.c_str();
|
|
}
|
|
|
|
/* -------------------------------------------------------------------- */
|
|
/** \name Platform/driver/device information
|
|
* \{ */
|
|
|
|
constexpr int32_t PCI_ID_NVIDIA = 0x10de;
|
|
constexpr int32_t PCI_ID_INTEL = 0x8086;
|
|
constexpr int32_t PCI_ID_AMD = 0x1002;
|
|
constexpr int32_t PCI_ID_ATI = 0x1022;
|
|
|
|
eGPUDeviceType VKDevice::device_type() const
|
|
{
|
|
/* According to the vulkan specifications:
|
|
*
|
|
* If the vendor has a PCI vendor ID, the low 16 bits of vendorID must contain that PCI vendor
|
|
* ID, and the remaining bits must be set to zero. Otherwise, the value returned must be a valid
|
|
* Khronos vendor ID.
|
|
*/
|
|
switch (vk_physical_device_properties_.vendorID) {
|
|
case PCI_ID_NVIDIA:
|
|
return GPU_DEVICE_NVIDIA;
|
|
case PCI_ID_INTEL:
|
|
return GPU_DEVICE_INTEL;
|
|
case PCI_ID_AMD:
|
|
case PCI_ID_ATI:
|
|
return GPU_DEVICE_ATI;
|
|
default:
|
|
break;
|
|
}
|
|
|
|
return GPU_DEVICE_UNKNOWN;
|
|
}
|
|
|
|
eGPUDriverType VKDevice::driver_type() const
|
|
{
|
|
/* It is unclear how to determine the driver type, but it is required to extract the correct
|
|
* driver version. */
|
|
return GPU_DRIVER_ANY;
|
|
}
|
|
|
|
std::string VKDevice::vendor_name() const
|
|
{
|
|
/* Below 0x10000 are the PCI vendor IDs (https://pcisig.com/membership/member-companies) */
|
|
if (vk_physical_device_properties_.vendorID < 0x10000) {
|
|
switch (vk_physical_device_properties_.vendorID) {
|
|
case 0x1022:
|
|
return "Advanced Micro Devices";
|
|
case 0x10DE:
|
|
return "NVIDIA Corporation";
|
|
case 0x8086:
|
|
return "Intel Corporation";
|
|
default:
|
|
return std::to_string(vk_physical_device_properties_.vendorID);
|
|
}
|
|
}
|
|
else {
|
|
/* above 0x10000 should be vkVendorIDs
|
|
* NOTE: When debug_messaging landed we can use something similar to
|
|
* vk::to_string(vk::VendorId(properties.vendorID));
|
|
*/
|
|
return std::to_string(vk_physical_device_properties_.vendorID);
|
|
}
|
|
}
|
|
|
|
std::string VKDevice::driver_version() const
|
|
{
|
|
/*
|
|
* NOTE: this depends on the driver type and is currently incorrect. Idea is to use a default per
|
|
* OS.
|
|
*/
|
|
const uint32_t driver_version = vk_physical_device_properties_.driverVersion;
|
|
switch (vk_physical_device_properties_.vendorID) {
|
|
case PCI_ID_NVIDIA:
|
|
return std::to_string((driver_version >> 22) & 0x3FF) + "." +
|
|
std::to_string((driver_version >> 14) & 0xFF) + "." +
|
|
std::to_string((driver_version >> 6) & 0xFF) + "." +
|
|
std::to_string(driver_version & 0x3F);
|
|
case PCI_ID_INTEL: {
|
|
const uint32_t major = VK_VERSION_MAJOR(driver_version);
|
|
/* When using Mesa driver we should use VK_VERSION_*. */
|
|
if (major > 30) {
|
|
return std::to_string((driver_version >> 14) & 0x3FFFF) + "." +
|
|
std::to_string((driver_version & 0x3FFF));
|
|
}
|
|
break;
|
|
}
|
|
default:
|
|
break;
|
|
}
|
|
|
|
return std::to_string(VK_VERSION_MAJOR(driver_version)) + "." +
|
|
std::to_string(VK_VERSION_MINOR(driver_version)) + "." +
|
|
std::to_string(VK_VERSION_PATCH(driver_version));
|
|
}
|
|
|
|
/** \} */
|
|
|
|
/* -------------------------------------------------------------------- */
|
|
/** \name Resource management
|
|
* \{ */
|
|
|
|
void VKDevice::context_register(VKContext &context)
|
|
{
|
|
contexts_.append(std::reference_wrapper(context));
|
|
}
|
|
|
|
void VKDevice::context_unregister(VKContext &context)
|
|
{
|
|
contexts_.remove(contexts_.first_index_of(std::reference_wrapper(context)));
|
|
}
|
|
const Vector<std::reference_wrapper<VKContext>> &VKDevice::contexts_get() const
|
|
{
|
|
return contexts_;
|
|
};
|
|
|
|
void VKDevice::discard_image(VkImage vk_image, VmaAllocation vma_allocation)
|
|
{
|
|
discarded_images_.append(std::pair(vk_image, vma_allocation));
|
|
}
|
|
|
|
void VKDevice::discard_image_view(VkImageView vk_image_view)
|
|
{
|
|
discarded_image_views_.append(vk_image_view);
|
|
}
|
|
|
|
void VKDevice::discard_buffer(VkBuffer vk_buffer, VmaAllocation vma_allocation)
|
|
{
|
|
discarded_buffers_.append(std::pair(vk_buffer, vma_allocation));
|
|
}
|
|
|
|
void VKDevice::discard_render_pass(VkRenderPass vk_render_pass)
|
|
{
|
|
discarded_render_passes_.append(vk_render_pass);
|
|
}
|
|
void VKDevice::discard_frame_buffer(VkFramebuffer vk_frame_buffer)
|
|
{
|
|
discarded_frame_buffers_.append(vk_frame_buffer);
|
|
}
|
|
|
|
void VKDevice::destroy_discarded_resources()
|
|
{
|
|
VK_ALLOCATION_CALLBACKS
|
|
|
|
while (!discarded_image_views_.is_empty()) {
|
|
VkImageView vk_image_view = discarded_image_views_.pop_last();
|
|
vkDestroyImageView(vk_device_, vk_image_view, vk_allocation_callbacks);
|
|
}
|
|
|
|
while (!discarded_images_.is_empty()) {
|
|
std::pair<VkImage, VmaAllocation> image_allocation = discarded_images_.pop_last();
|
|
vmaDestroyImage(mem_allocator_get(), image_allocation.first, image_allocation.second);
|
|
}
|
|
|
|
while (!discarded_buffers_.is_empty()) {
|
|
std::pair<VkBuffer, VmaAllocation> buffer_allocation = discarded_buffers_.pop_last();
|
|
vmaDestroyBuffer(mem_allocator_get(), buffer_allocation.first, buffer_allocation.second);
|
|
}
|
|
|
|
while (!discarded_render_passes_.is_empty()) {
|
|
VkRenderPass vk_render_pass = discarded_render_passes_.pop_last();
|
|
vkDestroyRenderPass(vk_device_, vk_render_pass, vk_allocation_callbacks);
|
|
}
|
|
|
|
while (!discarded_frame_buffers_.is_empty()) {
|
|
VkFramebuffer vk_frame_buffer = discarded_frame_buffers_.pop_last();
|
|
vkDestroyFramebuffer(vk_device_, vk_frame_buffer, vk_allocation_callbacks);
|
|
}
|
|
}
|
|
|
|
void VKDevice::memory_statistics_get(int *r_total_mem_kb, int *r_free_mem_kb) const
|
|
{
|
|
VmaBudget budgets[VK_MAX_MEMORY_HEAPS];
|
|
vmaGetHeapBudgets(mem_allocator_get(), budgets);
|
|
VkDeviceSize total_mem = 0;
|
|
VkDeviceSize used_mem = 0;
|
|
|
|
for (int memory_heap_index : IndexRange(vk_physical_device_memory_properties_.memoryHeapCount)) {
|
|
const VkMemoryHeap &memory_heap =
|
|
vk_physical_device_memory_properties_.memoryHeaps[memory_heap_index];
|
|
const VmaBudget &budget = budgets[memory_heap_index];
|
|
|
|
/* Skip host memory-heaps. */
|
|
if (!bool(memory_heap.flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT)) {
|
|
continue;
|
|
}
|
|
|
|
total_mem += memory_heap.size;
|
|
used_mem += budget.usage;
|
|
}
|
|
|
|
*r_total_mem_kb = int(total_mem / 1024);
|
|
*r_free_mem_kb = int((total_mem - used_mem) / 1024);
|
|
}
|
|
|
|
/** \} */
|
|
|
|
} // namespace blender::gpu
|