blob: 5a6b600f6a9711bdc891165bf93a7496d9d46187 [file] [log] [blame]
// Copyright 2017 The Dawn Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "dawn_native/d3d12/TextureD3D12.h"
#include "common/Constants.h"
#include "common/Math.h"
#include "dawn_native/DynamicUploader.h"
#include "dawn_native/Error.h"
#include "dawn_native/d3d12/BufferD3D12.h"
#include "dawn_native/d3d12/CommandRecordingContext.h"
#include "dawn_native/d3d12/D3D12Error.h"
#include "dawn_native/d3d12/DeviceD3D12.h"
#include "dawn_native/d3d12/HeapD3D12.h"
#include "dawn_native/d3d12/ResourceAllocatorManagerD3D12.h"
#include "dawn_native/d3d12/StagingBufferD3D12.h"
#include "dawn_native/d3d12/StagingDescriptorAllocatorD3D12.h"
#include "dawn_native/d3d12/TextureCopySplitter.h"
#include "dawn_native/d3d12/UtilsD3D12.h"
namespace dawn_native { namespace d3d12 {
namespace {
D3D12_RESOURCE_STATES D3D12TextureUsage(wgpu::TextureUsage usage, const Format& format) {
D3D12_RESOURCE_STATES resourceState = D3D12_RESOURCE_STATE_COMMON;
if (usage & kPresentTextureUsage) {
// The present usage is only used internally by the swapchain and is never used in
// combination with other usages.
ASSERT(usage == kPresentTextureUsage);
return D3D12_RESOURCE_STATE_PRESENT;
}
if (usage & wgpu::TextureUsage::CopySrc) {
resourceState |= D3D12_RESOURCE_STATE_COPY_SOURCE;
}
if (usage & wgpu::TextureUsage::CopyDst) {
resourceState |= D3D12_RESOURCE_STATE_COPY_DEST;
}
if (usage & (wgpu::TextureUsage::Sampled | kReadonlyStorageTexture)) {
resourceState |= (D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE |
D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE);
}
if (usage & wgpu::TextureUsage::Storage) {
resourceState |= D3D12_RESOURCE_STATE_UNORDERED_ACCESS;
}
if (usage & wgpu::TextureUsage::OutputAttachment) {
if (format.HasDepthOrStencil()) {
resourceState |= D3D12_RESOURCE_STATE_DEPTH_WRITE;
} else {
resourceState |= D3D12_RESOURCE_STATE_RENDER_TARGET;
}
}
return resourceState;
}
D3D12_RESOURCE_FLAGS D3D12ResourceFlags(wgpu::TextureUsage usage,
const Format& format,
bool isMultisampledTexture) {
D3D12_RESOURCE_FLAGS flags = D3D12_RESOURCE_FLAG_NONE;
if (usage & wgpu::TextureUsage::Storage) {
flags |= D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS;
}
// A multisampled resource must have either D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET or
// D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL set in D3D12_RESOURCE_DESC::Flags.
// https://docs.microsoft.com/en-us/windows/desktop/api/d3d12/ns-d3d12-d3d12_resource_desc
// Currently all textures are zero-initialized via the render-target path so always add
// the render target flag, except for compressed textures for which the render-target
// flag is invalid.
// TODO(natlee@microsoft.com, jiawei.shao@intel.com): do not require render target for
// lazy clearing.
if ((usage & wgpu::TextureUsage::OutputAttachment) || isMultisampledTexture ||
!format.isCompressed) {
if (format.HasDepthOrStencil()) {
flags |= D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL;
} else {
flags |= D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET;
}
}
ASSERT(!(flags & D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL) ||
flags == D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL);
return flags;
}
D3D12_RESOURCE_DIMENSION D3D12TextureDimension(wgpu::TextureDimension dimension) {
switch (dimension) {
case wgpu::TextureDimension::e2D:
return D3D12_RESOURCE_DIMENSION_TEXTURE2D;
default:
UNREACHABLE();
}
}
DXGI_FORMAT D3D12TypelessTextureFormat(wgpu::TextureFormat format) {
switch (format) {
case wgpu::TextureFormat::R8Unorm:
case wgpu::TextureFormat::R8Snorm:
case wgpu::TextureFormat::R8Uint:
case wgpu::TextureFormat::R8Sint:
return DXGI_FORMAT_R8_TYPELESS;
case wgpu::TextureFormat::R16Uint:
case wgpu::TextureFormat::R16Sint:
case wgpu::TextureFormat::R16Float:
return DXGI_FORMAT_R16_TYPELESS;
case wgpu::TextureFormat::RG8Unorm:
case wgpu::TextureFormat::RG8Snorm:
case wgpu::TextureFormat::RG8Uint:
case wgpu::TextureFormat::RG8Sint:
return DXGI_FORMAT_R8G8_TYPELESS;
case wgpu::TextureFormat::R32Uint:
case wgpu::TextureFormat::R32Sint:
case wgpu::TextureFormat::R32Float:
return DXGI_FORMAT_R32_TYPELESS;
case wgpu::TextureFormat::RG16Uint:
case wgpu::TextureFormat::RG16Sint:
case wgpu::TextureFormat::RG16Float:
return DXGI_FORMAT_R16G16_TYPELESS;
case wgpu::TextureFormat::RGBA8Unorm:
case wgpu::TextureFormat::RGBA8UnormSrgb:
case wgpu::TextureFormat::RGBA8Snorm:
case wgpu::TextureFormat::RGBA8Uint:
case wgpu::TextureFormat::RGBA8Sint:
return DXGI_FORMAT_R8G8B8A8_TYPELESS;
case wgpu::TextureFormat::BGRA8Unorm:
case wgpu::TextureFormat::BGRA8UnormSrgb:
return DXGI_FORMAT_B8G8R8A8_TYPELESS;
case wgpu::TextureFormat::RGB10A2Unorm:
return DXGI_FORMAT_R10G10B10A2_TYPELESS;
case wgpu::TextureFormat::RG11B10Float:
return DXGI_FORMAT_R11G11B10_FLOAT;
case wgpu::TextureFormat::RG32Uint:
case wgpu::TextureFormat::RG32Sint:
case wgpu::TextureFormat::RG32Float:
return DXGI_FORMAT_R32G32_TYPELESS;
case wgpu::TextureFormat::RGBA16Uint:
case wgpu::TextureFormat::RGBA16Sint:
case wgpu::TextureFormat::RGBA16Float:
return DXGI_FORMAT_R16G16B16A16_TYPELESS;
case wgpu::TextureFormat::RGBA32Uint:
case wgpu::TextureFormat::RGBA32Sint:
case wgpu::TextureFormat::RGBA32Float:
return DXGI_FORMAT_R32G32B32A32_TYPELESS;
case wgpu::TextureFormat::Depth32Float:
case wgpu::TextureFormat::Depth24Plus:
return DXGI_FORMAT_R32_TYPELESS;
case wgpu::TextureFormat::Depth24PlusStencil8:
return DXGI_FORMAT_X32_TYPELESS_G8X24_UINT;
case wgpu::TextureFormat::BC1RGBAUnorm:
case wgpu::TextureFormat::BC1RGBAUnormSrgb:
return DXGI_FORMAT_BC1_TYPELESS;
case wgpu::TextureFormat::BC2RGBAUnorm:
case wgpu::TextureFormat::BC2RGBAUnormSrgb:
return DXGI_FORMAT_BC2_TYPELESS;
case wgpu::TextureFormat::BC3RGBAUnorm:
case wgpu::TextureFormat::BC3RGBAUnormSrgb:
return DXGI_FORMAT_BC3_TYPELESS;
case wgpu::TextureFormat::BC4RSnorm:
case wgpu::TextureFormat::BC4RUnorm:
return DXGI_FORMAT_BC4_TYPELESS;
case wgpu::TextureFormat::BC5RGSnorm:
case wgpu::TextureFormat::BC5RGUnorm:
return DXGI_FORMAT_BC5_TYPELESS;
case wgpu::TextureFormat::BC6HRGBSfloat:
case wgpu::TextureFormat::BC6HRGBUfloat:
return DXGI_FORMAT_BC6H_TYPELESS;
case wgpu::TextureFormat::BC7RGBAUnorm:
case wgpu::TextureFormat::BC7RGBAUnormSrgb:
return DXGI_FORMAT_BC7_TYPELESS;
default:
UNREACHABLE();
}
}
} // namespace
DXGI_FORMAT D3D12TextureFormat(wgpu::TextureFormat format) {
switch (format) {
case wgpu::TextureFormat::R8Unorm:
return DXGI_FORMAT_R8_UNORM;
case wgpu::TextureFormat::R8Snorm:
return DXGI_FORMAT_R8_SNORM;
case wgpu::TextureFormat::R8Uint:
return DXGI_FORMAT_R8_UINT;
case wgpu::TextureFormat::R8Sint:
return DXGI_FORMAT_R8_SINT;
case wgpu::TextureFormat::R16Uint:
return DXGI_FORMAT_R16_UINT;
case wgpu::TextureFormat::R16Sint:
return DXGI_FORMAT_R16_SINT;
case wgpu::TextureFormat::R16Float:
return DXGI_FORMAT_R16_FLOAT;
case wgpu::TextureFormat::RG8Unorm:
return DXGI_FORMAT_R8G8_UNORM;
case wgpu::TextureFormat::RG8Snorm:
return DXGI_FORMAT_R8G8_SNORM;
case wgpu::TextureFormat::RG8Uint:
return DXGI_FORMAT_R8G8_UINT;
case wgpu::TextureFormat::RG8Sint:
return DXGI_FORMAT_R8G8_SINT;
case wgpu::TextureFormat::R32Uint:
return DXGI_FORMAT_R32_UINT;
case wgpu::TextureFormat::R32Sint:
return DXGI_FORMAT_R32_SINT;
case wgpu::TextureFormat::R32Float:
return DXGI_FORMAT_R32_FLOAT;
case wgpu::TextureFormat::RG16Uint:
return DXGI_FORMAT_R16G16_UINT;
case wgpu::TextureFormat::RG16Sint:
return DXGI_FORMAT_R16G16_SINT;
case wgpu::TextureFormat::RG16Float:
return DXGI_FORMAT_R16G16_FLOAT;
case wgpu::TextureFormat::RGBA8Unorm:
return DXGI_FORMAT_R8G8B8A8_UNORM;
case wgpu::TextureFormat::RGBA8UnormSrgb:
return DXGI_FORMAT_R8G8B8A8_UNORM_SRGB;
case wgpu::TextureFormat::RGBA8Snorm:
return DXGI_FORMAT_R8G8B8A8_SNORM;
case wgpu::TextureFormat::RGBA8Uint:
return DXGI_FORMAT_R8G8B8A8_UINT;
case wgpu::TextureFormat::RGBA8Sint:
return DXGI_FORMAT_R8G8B8A8_SINT;
case wgpu::TextureFormat::BGRA8Unorm:
return DXGI_FORMAT_B8G8R8A8_UNORM;
case wgpu::TextureFormat::BGRA8UnormSrgb:
return DXGI_FORMAT_B8G8R8A8_UNORM_SRGB;
case wgpu::TextureFormat::RGB10A2Unorm:
return DXGI_FORMAT_R10G10B10A2_UNORM;
case wgpu::TextureFormat::RG11B10Float:
return DXGI_FORMAT_R11G11B10_FLOAT;
case wgpu::TextureFormat::RG32Uint:
return DXGI_FORMAT_R32G32_UINT;
case wgpu::TextureFormat::RG32Sint:
return DXGI_FORMAT_R32G32_SINT;
case wgpu::TextureFormat::RG32Float:
return DXGI_FORMAT_R32G32_FLOAT;
case wgpu::TextureFormat::RGBA16Uint:
return DXGI_FORMAT_R16G16B16A16_UINT;
case wgpu::TextureFormat::RGBA16Sint:
return DXGI_FORMAT_R16G16B16A16_SINT;
case wgpu::TextureFormat::RGBA16Float:
return DXGI_FORMAT_R16G16B16A16_FLOAT;
case wgpu::TextureFormat::RGBA32Uint:
return DXGI_FORMAT_R32G32B32A32_UINT;
case wgpu::TextureFormat::RGBA32Sint:
return DXGI_FORMAT_R32G32B32A32_SINT;
case wgpu::TextureFormat::RGBA32Float:
return DXGI_FORMAT_R32G32B32A32_FLOAT;
case wgpu::TextureFormat::Depth32Float:
return DXGI_FORMAT_D32_FLOAT;
case wgpu::TextureFormat::Depth24Plus:
return DXGI_FORMAT_D32_FLOAT;
case wgpu::TextureFormat::Depth24PlusStencil8:
return DXGI_FORMAT_D32_FLOAT_S8X24_UINT;
case wgpu::TextureFormat::BC1RGBAUnorm:
return DXGI_FORMAT_BC1_UNORM;
case wgpu::TextureFormat::BC1RGBAUnormSrgb:
return DXGI_FORMAT_BC1_UNORM_SRGB;
case wgpu::TextureFormat::BC2RGBAUnorm:
return DXGI_FORMAT_BC2_UNORM;
case wgpu::TextureFormat::BC2RGBAUnormSrgb:
return DXGI_FORMAT_BC2_UNORM_SRGB;
case wgpu::TextureFormat::BC3RGBAUnorm:
return DXGI_FORMAT_BC3_UNORM;
case wgpu::TextureFormat::BC3RGBAUnormSrgb:
return DXGI_FORMAT_BC3_UNORM_SRGB;
case wgpu::TextureFormat::BC4RSnorm:
return DXGI_FORMAT_BC4_SNORM;
case wgpu::TextureFormat::BC4RUnorm:
return DXGI_FORMAT_BC4_UNORM;
case wgpu::TextureFormat::BC5RGSnorm:
return DXGI_FORMAT_BC5_SNORM;
case wgpu::TextureFormat::BC5RGUnorm:
return DXGI_FORMAT_BC5_UNORM;
case wgpu::TextureFormat::BC6HRGBSfloat:
return DXGI_FORMAT_BC6H_SF16;
case wgpu::TextureFormat::BC6HRGBUfloat:
return DXGI_FORMAT_BC6H_UF16;
case wgpu::TextureFormat::BC7RGBAUnorm:
return DXGI_FORMAT_BC7_UNORM;
case wgpu::TextureFormat::BC7RGBAUnormSrgb:
return DXGI_FORMAT_BC7_UNORM_SRGB;
default:
UNREACHABLE();
}
}
MaybeError ValidateTextureDescriptorCanBeWrapped(const TextureDescriptor* descriptor) {
if (descriptor->dimension != wgpu::TextureDimension::e2D) {
return DAWN_VALIDATION_ERROR("Texture must be 2D");
}
if (descriptor->mipLevelCount != 1) {
return DAWN_VALIDATION_ERROR("Mip level count must be 1");
}
if (descriptor->size.depth != 1) {
return DAWN_VALIDATION_ERROR("Depth must be 1");
}
if (descriptor->sampleCount != 1) {
return DAWN_VALIDATION_ERROR("Sample count must be 1");
}
return {};
}
MaybeError ValidateD3D12TextureCanBeWrapped(ID3D12Resource* d3d12Resource,
const TextureDescriptor* dawnDescriptor) {
const D3D12_RESOURCE_DESC d3dDescriptor = d3d12Resource->GetDesc();
if ((dawnDescriptor->size.width != d3dDescriptor.Width) ||
(dawnDescriptor->size.height != d3dDescriptor.Height) ||
(dawnDescriptor->size.depth != 1)) {
return DAWN_VALIDATION_ERROR("D3D12 texture size doesn't match descriptor");
}
const DXGI_FORMAT dxgiFormatFromDescriptor = D3D12TextureFormat(dawnDescriptor->format);
if (dxgiFormatFromDescriptor != d3dDescriptor.Format) {
return DAWN_VALIDATION_ERROR(
"D3D12 texture format must be compatible with descriptor format.");
}
if (d3dDescriptor.MipLevels != 1) {
return DAWN_VALIDATION_ERROR("D3D12 texture number of miplevels must be 1.");
}
if (d3dDescriptor.DepthOrArraySize != 1) {
return DAWN_VALIDATION_ERROR("D3D12 texture array size must be 1.");
}
// Shared textures cannot be multi-sample so no need to check those.
ASSERT(d3dDescriptor.SampleDesc.Count == 1);
ASSERT(d3dDescriptor.SampleDesc.Quality == 0);
return {};
}
ResultOrError<Ref<TextureBase>> Texture::Create(Device* device,
const TextureDescriptor* descriptor) {
Ref<Texture> dawnTexture =
AcquireRef(new Texture(device, descriptor, TextureState::OwnedInternal));
DAWN_TRY(dawnTexture->InitializeAsInternalTexture());
return std::move(dawnTexture);
}
ResultOrError<Ref<TextureBase>> Texture::Create(Device* device,
const ExternalImageDescriptor* descriptor,
HANDLE sharedHandle,
uint64_t acquireMutexKey,
bool isSwapChainTexture) {
const TextureDescriptor* textureDescriptor =
reinterpret_cast<const TextureDescriptor*>(descriptor->cTextureDescriptor);
// TODO(dawn:22): Remove once migration from GPUTextureDescriptor.arrayLayerCount to
// GPUTextureDescriptor.size.depth is done.
TextureDescriptor fixedDescriptor;
DAWN_TRY_ASSIGN(fixedDescriptor, FixTextureDescriptor(device, textureDescriptor));
textureDescriptor = &fixedDescriptor;
Ref<Texture> dawnTexture =
AcquireRef(new Texture(device, textureDescriptor, TextureState::OwnedExternal));
DAWN_TRY(dawnTexture->InitializeAsExternalTexture(textureDescriptor, sharedHandle,
acquireMutexKey, isSwapChainTexture));
dawnTexture->SetIsSubresourceContentInitialized(descriptor->isCleared,
dawnTexture->GetAllSubresources());
return std::move(dawnTexture);
}
MaybeError Texture::InitializeAsExternalTexture(const TextureDescriptor* descriptor,
HANDLE sharedHandle,
uint64_t acquireMutexKey,
bool isSwapChainTexture) {
Device* dawnDevice = ToBackend(GetDevice());
DAWN_TRY(ValidateTextureDescriptor(dawnDevice, descriptor));
DAWN_TRY(ValidateTextureDescriptorCanBeWrapped(descriptor));
ComPtr<ID3D12Resource> d3d12Resource;
DAWN_TRY(CheckHRESULT(dawnDevice->GetD3D12Device()->OpenSharedHandle(
sharedHandle, IID_PPV_ARGS(&d3d12Resource)),
"D3D12 opening shared handle"));
DAWN_TRY(ValidateD3D12TextureCanBeWrapped(d3d12Resource.Get(), descriptor));
ComPtr<IDXGIKeyedMutex> dxgiKeyedMutex;
DAWN_TRY_ASSIGN(dxgiKeyedMutex,
dawnDevice->CreateKeyedMutexForTexture(d3d12Resource.Get()));
DAWN_TRY(CheckHRESULT(dxgiKeyedMutex->AcquireSync(acquireMutexKey, INFINITE),
"D3D12 acquiring shared mutex"));
mAcquireMutexKey = acquireMutexKey;
mDxgiKeyedMutex = std::move(dxgiKeyedMutex);
mSwapChainTexture = isSwapChainTexture;
AllocationInfo info;
info.mMethod = AllocationMethod::kExternal;
// When creating the ResourceHeapAllocation, the resource heap is set to nullptr because the
// texture is owned externally. The texture's owning entity must remain responsible for
// memory management.
mResourceAllocation = {info, 0, std::move(d3d12Resource), nullptr};
return {};
}
MaybeError Texture::InitializeAsInternalTexture() {
D3D12_RESOURCE_DESC resourceDescriptor;
resourceDescriptor.Dimension = D3D12TextureDimension(GetDimension());
resourceDescriptor.Alignment = 0;
const Extent3D& size = GetSize();
resourceDescriptor.Width = size.width;
resourceDescriptor.Height = size.height;
resourceDescriptor.DepthOrArraySize = size.depth;
// This will need to be much more nuanced when WebGPU has
// texture view compatibility rules.
bool needsTypelessFormat = GetFormat().format == wgpu::TextureFormat::Depth32Float &&
(GetUsage() & wgpu::TextureUsage::Sampled) != 0;
DXGI_FORMAT dxgiFormat = needsTypelessFormat
? D3D12TypelessTextureFormat(GetFormat().format)
: D3D12TextureFormat(GetFormat().format);
resourceDescriptor.MipLevels = static_cast<UINT16>(GetNumMipLevels());
resourceDescriptor.Format = dxgiFormat;
resourceDescriptor.SampleDesc.Count = GetSampleCount();
// TODO(bryan.bernhart@intel.com): investigate how to specify standard MSAA sample pattern.
resourceDescriptor.SampleDesc.Quality = 0;
resourceDescriptor.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
resourceDescriptor.Flags =
D3D12ResourceFlags(GetUsage(), GetFormat(), IsMultisampledTexture());
DAWN_TRY_ASSIGN(mResourceAllocation,
ToBackend(GetDevice())
->AllocateMemory(D3D12_HEAP_TYPE_DEFAULT, resourceDescriptor,
D3D12_RESOURCE_STATE_COMMON));
Device* device = ToBackend(GetDevice());
if (device->IsToggleEnabled(Toggle::NonzeroClearResourcesOnCreationForTesting)) {
CommandRecordingContext* commandContext;
DAWN_TRY_ASSIGN(commandContext, device->GetPendingCommandContext());
DAWN_TRY(ClearTexture(commandContext, GetAllSubresources(),
TextureBase::ClearValue::NonZero));
}
return {};
}
Texture::Texture(Device* device, const TextureDescriptor* descriptor, TextureState state)
: TextureBase(device, descriptor, state),
mSubresourceStateAndDecay(
GetSubresourceCount(),
{D3D12_RESOURCE_STATES::D3D12_RESOURCE_STATE_COMMON, UINT64_MAX, false}) {
}
Texture::Texture(Device* device,
const TextureDescriptor* descriptor,
ComPtr<ID3D12Resource> nativeTexture)
: Texture(device, descriptor, TextureState::OwnedExternal) {
AllocationInfo info;
info.mMethod = AllocationMethod::kExternal;
// When creating the ResourceHeapAllocation, the resource heap is set to nullptr because the
// texture is owned externally. The texture's owning entity must remain responsible for
// memory management.
mResourceAllocation = {info, 0, std::move(nativeTexture), nullptr};
SetIsSubresourceContentInitialized(true, GetAllSubresources());
}
Texture::~Texture() {
DestroyInternal();
}
void Texture::DestroyImpl() {
Device* device = ToBackend(GetDevice());
// In PIX's D3D12-only mode, there is no way to determine frame boundaries
// for WebGPU since Dawn does not manage DXGI swap chains. Without assistance,
// PIX will wait forever for a present that never happens.
// If we know we're dealing with a swapbuffer texture, inform PIX we've
// "presented" the texture so it can determine frame boundaries and use its
// contents for the UI.
if (mSwapChainTexture) {
ID3D12SharingContract* d3dSharingContract = device->GetSharingContract();
if (d3dSharingContract != nullptr) {
d3dSharingContract->Present(mResourceAllocation.GetD3D12Resource().Get(), 0, 0);
}
}
device->DeallocateMemory(mResourceAllocation);
if (mDxgiKeyedMutex != nullptr) {
mDxgiKeyedMutex->ReleaseSync(mAcquireMutexKey + 1);
device->ReleaseKeyedMutexForTexture(std::move(mDxgiKeyedMutex));
}
}
DXGI_FORMAT Texture::GetD3D12Format() const {
return D3D12TextureFormat(GetFormat().format);
}
ID3D12Resource* Texture::GetD3D12Resource() const {
return mResourceAllocation.GetD3D12Resource().Get();
}
void Texture::TrackUsageAndTransitionNow(CommandRecordingContext* commandContext,
wgpu::TextureUsage usage,
const SubresourceRange& range) {
TrackUsageAndTransitionNow(commandContext, D3D12TextureUsage(usage, GetFormat()), range);
}
void Texture::TrackAllUsageAndTransitionNow(CommandRecordingContext* commandContext,
wgpu::TextureUsage usage) {
TrackUsageAndTransitionNow(commandContext, D3D12TextureUsage(usage, GetFormat()),
GetAllSubresources());
}
void Texture::TrackAllUsageAndTransitionNow(CommandRecordingContext* commandContext,
D3D12_RESOURCE_STATES newState) {
TrackUsageAndTransitionNow(commandContext, newState, GetAllSubresources());
}
void Texture::TrackUsageAndTransitionNow(CommandRecordingContext* commandContext,
D3D12_RESOURCE_STATES newState,
const SubresourceRange& range) {
if (mResourceAllocation.GetInfo().mMethod != AllocationMethod::kExternal) {
// Track the underlying heap to ensure residency.
Heap* heap = ToBackend(mResourceAllocation.GetResourceHeap());
commandContext->TrackHeapUsage(heap, GetDevice()->GetPendingCommandSerial());
}
std::vector<D3D12_RESOURCE_BARRIER> barriers;
barriers.reserve(range.levelCount * range.layerCount);
TransitionUsageAndGetResourceBarrier(commandContext, &barriers, newState, range);
if (barriers.size()) {
commandContext->GetCommandList()->ResourceBarrier(barriers.size(), barriers.data());
}
}
void Texture::TransitionSingleOrAllSubresources(std::vector<D3D12_RESOURCE_BARRIER>* barriers,
uint32_t index,
D3D12_RESOURCE_STATES newState,
const Serial pendingCommandSerial,
bool allSubresources) {
StateAndDecay* state = &mSubresourceStateAndDecay[index];
// Reuse the subresource(s) directly and avoid transition when it isn't needed, and
// return false.
// TODO(cwallez@chromium.org): Need some form of UAV barriers at some point.
if (state->lastState == newState) {
return;
}
D3D12_RESOURCE_STATES lastState = state->lastState;
// The COMMON state represents a state where no write operations can be pending, and
// where all pixels are uncompressed. This makes it possible to transition to and
// from some states without synchronization (i.e. without an explicit
// ResourceBarrier call). Textures can be implicitly promoted to 1) a single write
// state, or 2) multiple read states. Textures will implicitly decay to the COMMON
// state when all of the following are true: 1) the texture is accessed on a command
// list, 2) the ExecuteCommandLists call that uses that command list has ended, and
// 3) the texture was promoted implicitly to a read-only state and is still in that
// state.
// https://docs.microsoft.com/en-us/windows/desktop/direct3d12/using-resource-barriers-to-synchronize-resource-states-in-direct3d-12#implicit-state-transitions
// To track implicit decays, we must record the pending serial on which that
// transition will occur. When that texture is used again, the previously recorded
// serial must be compared to the last completed serial to determine if the texture
// has implicity decayed to the common state.
if (state->isValidToDecay && pendingCommandSerial > state->lastDecaySerial) {
lastState = D3D12_RESOURCE_STATE_COMMON;
}
// Update the tracked state.
state->lastState = newState;
// Destination states that qualify for an implicit promotion for a
// non-simultaneous-access texture: NON_PIXEL_SHADER_RESOURCE,
// PIXEL_SHADER_RESOURCE, COPY_SRC, COPY_DEST.
{
static constexpr D3D12_RESOURCE_STATES kD3D12PromotableReadOnlyStates =
D3D12_RESOURCE_STATE_COPY_SOURCE | D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE |
D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE;
if (lastState == D3D12_RESOURCE_STATE_COMMON) {
if (newState == (newState & kD3D12PromotableReadOnlyStates)) {
// Implicit texture state decays can only occur when the texture was implicitly
// transitioned to a read-only state. isValidToDecay is needed to differentiate
// between resources that were implictly or explicitly transitioned to a
// read-only state.
state->isValidToDecay = true;
state->lastDecaySerial = pendingCommandSerial;
return;
} else if (newState == D3D12_RESOURCE_STATE_COPY_DEST) {
state->isValidToDecay = false;
return;
}
}
}
D3D12_RESOURCE_BARRIER barrier;
barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_TRANSITION;
barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE;
barrier.Transition.pResource = GetD3D12Resource();
barrier.Transition.StateBefore = lastState;
barrier.Transition.StateAfter = newState;
barrier.Transition.Subresource =
allSubresources ? D3D12_RESOURCE_BARRIER_ALL_SUBRESOURCES : index;
barriers->push_back(barrier);
// TODO(yunchao.he@intel.com): support subresource for depth/stencil. Depth stencil
// texture has different plane slices. While the current implementation only has differernt
// mip slices and array slices for subresources.
// This is a hack because Dawn doesn't handle subresource of multiplanar resources
// correctly. We force the transition to be the same for all planes to match what the
// frontend validation checks for. This hack might be incorrect for stencil-only texture
// because we always set transition barrier for depth plane.
if (!allSubresources && newState == D3D12_RESOURCE_STATE_DEPTH_WRITE &&
GetFormat().HasStencil()) {
D3D12_RESOURCE_BARRIER barrierStencil = barrier;
barrierStencil.Transition.Subresource += GetArrayLayers() * GetNumMipLevels();
barriers->push_back(barrierStencil);
}
state->isValidToDecay = false;
}
void Texture::HandleTransitionSpecialCases(CommandRecordingContext* commandContext) {
// Textures with keyed mutexes can be written from other graphics queues. Hence, they
// must be acquired before command list submission to ensure work from the other queues
// has finished. See Device::ExecuteCommandContext.
if (mDxgiKeyedMutex != nullptr) {
commandContext->AddToSharedTextureList(this);
}
}
void Texture::TransitionUsageAndGetResourceBarrier(
CommandRecordingContext* commandContext,
std::vector<D3D12_RESOURCE_BARRIER>* barriers,
D3D12_RESOURCE_STATES newState,
const SubresourceRange& range) {
HandleTransitionSpecialCases(commandContext);
const Serial pendingCommandSerial = ToBackend(GetDevice())->GetPendingCommandSerial();
uint32_t subresourceCount = GetSubresourceCount();
// This transitions assume it is a 2D texture
ASSERT(GetDimension() == wgpu::TextureDimension::e2D);
// If the usages transitions can cover all subresources, and old usages of all subresources
// are the same, then we can use one barrier to do state transition for all subresources.
// Note that if the texture has only one mip level and one array slice, it will fall into
// this category.
bool areAllSubresourcesCovered = range.levelCount * range.layerCount == subresourceCount;
if (mSameLastUsagesAcrossSubresources && areAllSubresourcesCovered) {
TransitionSingleOrAllSubresources(barriers, 0, newState, pendingCommandSerial, true);
// TODO(yunchao.he@intel.com): compress and decompress if all subresources have the
// same states. We may need to retain mSubresourceStateAndDecay[0] only.
for (uint32_t i = 1; i < subresourceCount; ++i) {
mSubresourceStateAndDecay[i] = mSubresourceStateAndDecay[0];
}
return;
}
for (uint32_t arrayLayer = 0; arrayLayer < range.layerCount; ++arrayLayer) {
for (uint32_t mipLevel = 0; mipLevel < range.levelCount; ++mipLevel) {
uint32_t index = GetSubresourceIndex(range.baseMipLevel + mipLevel,
range.baseArrayLayer + arrayLayer);
TransitionSingleOrAllSubresources(barriers, index, newState, pendingCommandSerial,
false);
}
}
mSameLastUsagesAcrossSubresources = areAllSubresourcesCovered;
}
void Texture::TrackUsageAndGetResourceBarrierForPass(
CommandRecordingContext* commandContext,
std::vector<D3D12_RESOURCE_BARRIER>* barriers,
const PassTextureUsage& textureUsages) {
HandleTransitionSpecialCases(commandContext);
const Serial pendingCommandSerial = ToBackend(GetDevice())->GetPendingCommandSerial();
uint32_t subresourceCount = GetSubresourceCount();
ASSERT(textureUsages.subresourceUsages.size() == subresourceCount);
// This transitions assume it is a 2D texture
ASSERT(GetDimension() == wgpu::TextureDimension::e2D);
// If new usages of all subresources are the same and old usages of all subresources are
// the same too, we can use one barrier to do state transition for all subresources.
// Note that if the texture has only one mip level and one array slice, it will fall into
// this category.
if (textureUsages.sameUsagesAcrossSubresources && mSameLastUsagesAcrossSubresources) {
D3D12_RESOURCE_STATES newState = D3D12TextureUsage(textureUsages.usage, GetFormat());
TransitionSingleOrAllSubresources(barriers, 0, newState, pendingCommandSerial, true);
// TODO(yunchao.he@intel.com): compress and decompress if all subresources have the
// same states. We may need to retain mSubresourceStateAndDecay[0] only.
for (uint32_t i = 1; i < subresourceCount; ++i) {
mSubresourceStateAndDecay[i] = mSubresourceStateAndDecay[0];
}
return;
}
for (uint32_t arrayLayer = 0; arrayLayer < GetArrayLayers(); ++arrayLayer) {
for (uint32_t mipLevel = 0; mipLevel < GetNumMipLevels(); ++mipLevel) {
uint32_t index = GetSubresourceIndex(mipLevel, arrayLayer);
// Skip if this subresource is not used during the current pass
if (textureUsages.subresourceUsages[index] == wgpu::TextureUsage::None) {
continue;
}
D3D12_RESOURCE_STATES newState =
D3D12TextureUsage(textureUsages.subresourceUsages[index], GetFormat());
TransitionSingleOrAllSubresources(barriers, index, newState, pendingCommandSerial,
false);
}
}
mSameLastUsagesAcrossSubresources = textureUsages.sameUsagesAcrossSubresources;
}
D3D12_RENDER_TARGET_VIEW_DESC Texture::GetRTVDescriptor(uint32_t mipLevel,
uint32_t baseArrayLayer,
uint32_t layerCount) const {
ASSERT(GetDimension() == wgpu::TextureDimension::e2D);
D3D12_RENDER_TARGET_VIEW_DESC rtvDesc;
rtvDesc.Format = GetD3D12Format();
if (IsMultisampledTexture()) {
ASSERT(GetNumMipLevels() == 1);
ASSERT(layerCount == 1);
ASSERT(baseArrayLayer == 0);
ASSERT(mipLevel == 0);
rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DMS;
} else {
// Currently we always use D3D12_TEX2D_ARRAY_RTV because we cannot specify base array
// layer and layer count in D3D12_TEX2D_RTV. For 2D texture views, we treat them as
// 1-layer 2D array textures. (Just like how we treat SRVs)
// https://docs.microsoft.com/en-us/windows/desktop/api/d3d12/ns-d3d12-d3d12_tex2d_rtv
// https://docs.microsoft.com/en-us/windows/desktop/api/d3d12/ns-d3d12-d3d12_tex2d_array
// _rtv
rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY;
rtvDesc.Texture2DArray.FirstArraySlice = baseArrayLayer;
rtvDesc.Texture2DArray.ArraySize = layerCount;
rtvDesc.Texture2DArray.MipSlice = mipLevel;
rtvDesc.Texture2DArray.PlaneSlice = 0;
}
return rtvDesc;
}
D3D12_DEPTH_STENCIL_VIEW_DESC Texture::GetDSVDescriptor(uint32_t mipLevel,
uint32_t baseArrayLayer,
uint32_t layerCount) const {
D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc;
dsvDesc.Format = GetD3D12Format();
dsvDesc.Flags = D3D12_DSV_FLAG_NONE;
if (IsMultisampledTexture()) {
ASSERT(GetNumMipLevels() == 1);
ASSERT(layerCount == 1);
ASSERT(baseArrayLayer == 0);
ASSERT(mipLevel == 0);
dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DMS;
} else {
dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DARRAY;
dsvDesc.Texture2DArray.FirstArraySlice = baseArrayLayer;
dsvDesc.Texture2DArray.ArraySize = layerCount;
dsvDesc.Texture2DArray.MipSlice = mipLevel;
}
return dsvDesc;
}
MaybeError Texture::ClearTexture(CommandRecordingContext* commandContext,
const SubresourceRange& range,
TextureBase::ClearValue clearValue) {
// TODO(jiawei.shao@intel.com): initialize the textures in compressed formats with copies.
if (GetFormat().isCompressed) {
SetIsSubresourceContentInitialized(true, range);
return {};
}
ID3D12GraphicsCommandList* commandList = commandContext->GetCommandList();
Device* device = ToBackend(GetDevice());
uint8_t clearColor = (clearValue == TextureBase::ClearValue::Zero) ? 0 : 1;
float fClearColor = (clearValue == TextureBase::ClearValue::Zero) ? 0.f : 1.f;
if (GetFormat().isRenderable) {
if (GetFormat().HasDepthOrStencil()) {
TrackUsageAndTransitionNow(commandContext, D3D12_RESOURCE_STATE_DEPTH_WRITE, range);
D3D12_CLEAR_FLAGS clearFlags = {};
for (uint32_t level = range.baseMipLevel;
level < range.baseMipLevel + range.levelCount; ++level) {
for (uint32_t layer = range.baseArrayLayer;
layer < range.baseArrayLayer + range.layerCount; ++layer) {
if (clearValue == TextureBase::ClearValue::Zero &&
IsSubresourceContentInitialized(
SubresourceRange::SingleSubresource(level, layer))) {
// Skip lazy clears if already initialized.
continue;
}
CPUDescriptorHeapAllocation dsvHandle;
DAWN_TRY_ASSIGN(dsvHandle, device->GetDepthStencilViewAllocator()
->AllocateTransientCPUDescriptors());
const D3D12_CPU_DESCRIPTOR_HANDLE baseDescriptor =
dsvHandle.GetBaseDescriptor();
D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc = GetDSVDescriptor(level, layer, 1);
device->GetD3D12Device()->CreateDepthStencilView(GetD3D12Resource(),
&dsvDesc, baseDescriptor);
if (GetFormat().HasDepth()) {
clearFlags |= D3D12_CLEAR_FLAG_DEPTH;
}
if (GetFormat().HasStencil()) {
clearFlags |= D3D12_CLEAR_FLAG_STENCIL;
}
commandList->ClearDepthStencilView(baseDescriptor, clearFlags, fClearColor,
clearColor, 0, nullptr);
}
}
} else {
TrackUsageAndTransitionNow(commandContext, D3D12_RESOURCE_STATE_RENDER_TARGET,
range);
const float clearColorRGBA[4] = {fClearColor, fClearColor, fClearColor,
fClearColor};
for (uint32_t level = range.baseMipLevel;
level < range.baseMipLevel + range.levelCount; ++level) {
for (uint32_t layer = range.baseArrayLayer;
layer < range.baseArrayLayer + range.layerCount; ++layer) {
if (clearValue == TextureBase::ClearValue::Zero &&
IsSubresourceContentInitialized(
SubresourceRange::SingleSubresource(level, layer))) {
// Skip lazy clears if already initialized.
continue;
}
CPUDescriptorHeapAllocation rtvHeap;
DAWN_TRY_ASSIGN(rtvHeap, device->GetRenderTargetViewAllocator()
->AllocateTransientCPUDescriptors());
const D3D12_CPU_DESCRIPTOR_HANDLE rtvHandle = rtvHeap.GetBaseDescriptor();
D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = GetRTVDescriptor(level, layer, 1);
device->GetD3D12Device()->CreateRenderTargetView(GetD3D12Resource(),
&rtvDesc, rtvHandle);
commandList->ClearRenderTargetView(rtvHandle, clearColorRGBA, 0, nullptr);
}
}
}
} else {
// TODO(natlee@microsoft.com): test compressed textures are cleared
// create temp buffer with clear color to copy to the texture image
uint32_t bytesPerRow =
Align((GetWidth() / GetFormat().blockWidth) * GetFormat().blockByteSize,
kTextureBytesPerRowAlignment);
uint64_t bufferSize64 = bytesPerRow * (GetHeight() / GetFormat().blockHeight);
if (bufferSize64 > std::numeric_limits<uint32_t>::max()) {
return DAWN_OUT_OF_MEMORY_ERROR("Unable to allocate buffer.");
}
uint32_t bufferSize = static_cast<uint32_t>(bufferSize64);
DynamicUploader* uploader = device->GetDynamicUploader();
UploadHandle uploadHandle;
DAWN_TRY_ASSIGN(uploadHandle,
uploader->Allocate(bufferSize, device->GetPendingCommandSerial()));
memset(uploadHandle.mappedBuffer, clearColor, bufferSize);
TrackUsageAndTransitionNow(commandContext, D3D12_RESOURCE_STATE_COPY_DEST, range);
for (uint32_t level = range.baseMipLevel; level < range.baseMipLevel + range.levelCount;
++level) {
// compute d3d12 texture copy locations for texture and buffer
Extent3D copySize = GetMipLevelVirtualSize(level);
uint32_t rowsPerImage = GetHeight();
TextureCopySplit copySplit =
ComputeTextureCopySplit({0, 0, 0}, copySize, GetFormat(),
uploadHandle.startOffset, bytesPerRow, rowsPerImage);
for (uint32_t layer = range.baseArrayLayer;
layer < range.baseArrayLayer + range.layerCount; ++layer) {
if (clearValue == TextureBase::ClearValue::Zero &&
IsSubresourceContentInitialized(
SubresourceRange::SingleSubresource(level, layer))) {
// Skip lazy clears if already initialized.
continue;
}
D3D12_TEXTURE_COPY_LOCATION textureLocation =
ComputeTextureCopyLocationForTexture(this, level, layer);
for (uint32_t i = 0; i < copySplit.count; ++i) {
TextureCopySplit::CopyInfo& info = copySplit.copies[i];
D3D12_TEXTURE_COPY_LOCATION bufferLocation =
ComputeBufferLocationForCopyTextureRegion(
this, ToBackend(uploadHandle.stagingBuffer)->GetResource(),
info.bufferSize, copySplit.offset, bytesPerRow);
D3D12_BOX sourceRegion =
ComputeD3D12BoxFromOffsetAndSize(info.bufferOffset, info.copySize);
// copy the buffer filled with clear color to the texture
commandList->CopyTextureRegion(&textureLocation, info.textureOffset.x,
info.textureOffset.y, info.textureOffset.z,
&bufferLocation, &sourceRegion);
}
}
}
}
if (clearValue == TextureBase::ClearValue::Zero) {
SetIsSubresourceContentInitialized(true, range);
GetDevice()->IncrementLazyClearCountForTesting();
}
return {};
}
void Texture::EnsureSubresourceContentInitialized(CommandRecordingContext* commandContext,
const SubresourceRange& range) {
if (!ToBackend(GetDevice())->IsToggleEnabled(Toggle::LazyClearResourceOnFirstUse)) {
return;
}
if (!IsSubresourceContentInitialized(range)) {
// If subresource has not been initialized, clear it to black as it could contain
// dirty bits from recycled memory
GetDevice()->ConsumedError(
ClearTexture(commandContext, range, TextureBase::ClearValue::Zero));
}
}
TextureView::TextureView(TextureBase* texture, const TextureViewDescriptor* descriptor)
: TextureViewBase(texture, descriptor) {
mSrvDesc.Format = D3D12TextureFormat(descriptor->format);
if (descriptor->format == wgpu::TextureFormat::Depth32Float) {
// TODO(enga): This will need to be much more nuanced when WebGPU has
// texture view compatibility rules.
mSrvDesc.Format = DXGI_FORMAT_R32_FLOAT;
}
mSrvDesc.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING;
// Currently we always use D3D12_TEX2D_ARRAY_SRV because we cannot specify base array layer
// and layer count in D3D12_TEX2D_SRV. For 2D texture views, we treat them as 1-layer 2D
// array textures.
// https://docs.microsoft.com/en-us/windows/desktop/api/d3d12/ns-d3d12-d3d12_tex2d_srv
// https://docs.microsoft.com/en-us/windows/desktop/api/d3d12/ns-d3d12-d3d12_tex2d_array_srv
// TODO(jiawei.shao@intel.com): support more texture view dimensions.
// TODO(jiawei.shao@intel.com): support creating SRV on multisampled textures.
switch (descriptor->dimension) {
case wgpu::TextureViewDimension::e2D:
case wgpu::TextureViewDimension::e2DArray:
ASSERT(texture->GetDimension() == wgpu::TextureDimension::e2D);
mSrvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY;
mSrvDesc.Texture2DArray.ArraySize = descriptor->arrayLayerCount;
mSrvDesc.Texture2DArray.FirstArraySlice = descriptor->baseArrayLayer;
mSrvDesc.Texture2DArray.MipLevels = descriptor->mipLevelCount;
mSrvDesc.Texture2DArray.MostDetailedMip = descriptor->baseMipLevel;
mSrvDesc.Texture2DArray.PlaneSlice = 0;
mSrvDesc.Texture2DArray.ResourceMinLODClamp = 0;
break;
case wgpu::TextureViewDimension::Cube:
case wgpu::TextureViewDimension::CubeArray:
ASSERT(texture->GetDimension() == wgpu::TextureDimension::e2D);
ASSERT(descriptor->arrayLayerCount % 6 == 0);
mSrvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURECUBEARRAY;
mSrvDesc.TextureCubeArray.First2DArrayFace = descriptor->baseArrayLayer;
mSrvDesc.TextureCubeArray.NumCubes = descriptor->arrayLayerCount / 6;
mSrvDesc.TextureCubeArray.MostDetailedMip = descriptor->baseMipLevel;
mSrvDesc.TextureCubeArray.MipLevels = descriptor->mipLevelCount;
mSrvDesc.TextureCubeArray.ResourceMinLODClamp = 0;
break;
default:
UNREACHABLE();
}
}
DXGI_FORMAT TextureView::GetD3D12Format() const {
return D3D12TextureFormat(GetFormat().format);
}
const D3D12_SHADER_RESOURCE_VIEW_DESC& TextureView::GetSRVDescriptor() const {
return mSrvDesc;
}
D3D12_RENDER_TARGET_VIEW_DESC TextureView::GetRTVDescriptor() const {
return ToBackend(GetTexture())
->GetRTVDescriptor(GetBaseMipLevel(), GetBaseArrayLayer(), GetLayerCount());
}
D3D12_DEPTH_STENCIL_VIEW_DESC TextureView::GetDSVDescriptor() const {
ASSERT(GetLevelCount() == 1);
return ToBackend(GetTexture())
->GetDSVDescriptor(GetBaseMipLevel(), GetBaseArrayLayer(), GetLayerCount());
}
D3D12_UNORDERED_ACCESS_VIEW_DESC TextureView::GetUAVDescriptor() const {
D3D12_UNORDERED_ACCESS_VIEW_DESC uavDesc;
uavDesc.Format = GetD3D12Format();
ASSERT(!GetTexture()->IsMultisampledTexture());
uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY;
uavDesc.Texture2DArray.FirstArraySlice = GetBaseArrayLayer();
uavDesc.Texture2DArray.ArraySize = GetLayerCount();
uavDesc.Texture2DArray.MipSlice = GetBaseMipLevel();
uavDesc.Texture2DArray.PlaneSlice = 0;
return uavDesc;
}
}} // namespace dawn_native::d3d12