D3D12: Add RENDER_ATTACHMENT usage on depth stencil textures on Intel GPUs

This patch adds RENDER_ATTACHMENT to the internal usage of D3D12 depth
stencil textures on Intel GPU because due to a driver bug we have to
initialize all the depth stencil textures with clear on Intel GPUs
before copying with them. Otherwise, if a depth stencil texture is created
without RENDER_ATTACHMENT usage, it will be initialized by copies, which
will still trigger the driver bug mentioned above.

Bug: 1487
Test: dawn_end2end_test
Change-Id: I78b9a3e2bc4098d6f3f2619644c80fd54dafd4e8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/96985
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>

src/dawn/native/Texture.cpp

@@ -650,6 +650,10 @@ wgpu::TextureUsage TextureBase::GetInternalUsage() const {
     ASSERT(!IsError());
     return mInternalUsage;
 }
+void TextureBase::AddInternalUsage(wgpu::TextureUsage usage) {
+    ASSERT(!IsError());
+    mInternalUsage |= usage;
+}
 
 TextureBase::TextureState TextureBase::GetTextureState() const {
     ASSERT(!IsError());

src/dawn/native/Texture.h

@@ -112,6 +112,7 @@ class TextureBase : public ApiObjectBase {
     ~TextureBase() override;
 
     void DestroyImpl() override;
+    void AddInternalUsage(wgpu::TextureUsage usage);
 
   private:
     TextureBase(DeviceBase* device, const TextureDescriptor* descriptor, ObjectBase::ErrorTag tag);

src/dawn/native/Toggles.cpp

@@ -268,11 +268,11 @@ static constexpr ToggleEnumAndInfoList kToggleNameAndInfoList = {{
       "Enables usage of the blob cache (backed by the platform cache if set/passed). Necessary for "
       "any persistent caching capabilities, i.e. pipeline caching.",
       "https://crbug.com/dawn/549"}},
-    {Toggle::D3D12ForceInitializeCopyableDepthStencilTextureOnCreation,
+    {Toggle::D3D12ForceClearCopyableDepthStencilTextureOnCreation,
-     {"d3d12_force_initialize_copyable_depth_stencil_texture_on_creation",
+     {"d3d12_force_clear_copyable_depth_stencil_texture_on_creation",
-      "Always initializing copyable depth stencil textures when creating them instead of skipping "
+      "Always clearing copyable depth stencil textures when creating them instead of skipping the "
-      "the initialization when the entire subresource is the copy destination as a workaround on "
+      "initialization when the entire subresource is the copy destination as a workaround on Intel "
-      "Intel D3D12 drivers.",
+      "D3D12 drivers.",
       "https://crbug.com/dawn/1487"}},
     {Toggle::D3D12DontSetClearValueOnDepthTextureCreation,
      {"d3d12_dont_set_clear_value_on_depth_texture_creation",

src/dawn/native/Toggles.h

@@ -72,7 +72,7 @@ enum class Toggle {
     D3D12SplitBufferTextureCopyForRowsPerImagePaddings,
     MetalRenderR8RG8UnormSmallMipToTempTexture,
     EnableBlobCache,
-    D3D12ForceInitializeCopyableDepthStencilTextureOnCreation,
+    D3D12ForceClearCopyableDepthStencilTextureOnCreation,
     D3D12DontSetClearValueOnDepthTextureCreation,
     D3D12AlwaysUseTypelessFormatsForCastableTexture,
 

src/dawn/native/d3d12/DeviceD3D12.cpp

@@ -660,7 +660,7 @@ void Device::InitTogglesFromDriver() {
     // Currently this workaround is only needed on Intel GPUs.
     // See http://crbug.com/dawn/1487 for more information.
     if (gpu_info::IsIntel(vendorId)) {
-        SetToggle(Toggle::D3D12ForceInitializeCopyableDepthStencilTextureOnCreation, true);
+        SetToggle(Toggle::D3D12ForceClearCopyableDepthStencilTextureOnCreation, true);
     }
 
     // Currently this workaround is only needed on Intel Gen12 GPUs.

src/dawn/native/d3d12/TextureD3D12.cpp

@@ -581,6 +581,14 @@ MaybeError Texture::InitializeAsInternalTexture() {
     resourceDescriptor.Height = size.height;
     resourceDescriptor.DepthOrArraySize = size.depthOrArrayLayers;
 
+    Device* device = ToBackend(GetDevice());
+    bool applyForceClearCopyableDepthStencilTextureOnCreationToggle =
+        device->IsToggleEnabled(Toggle::D3D12ForceClearCopyableDepthStencilTextureOnCreation) &&
+        GetFormat().HasDepthOrStencil() && (GetInternalUsage() & wgpu::TextureUsage::CopyDst);
+    if (applyForceClearCopyableDepthStencilTextureOnCreationToggle) {
+        AddInternalUsage(wgpu::TextureUsage::RenderAttachment);
+    }
+
     // This will need to be much more nuanced when WebGPU has
     // texture view compatibility rules.
     const bool needsTypelessFormat =
@@ -601,17 +609,12 @@ MaybeError Texture::InitializeAsInternalTexture() {
     mD3D12ResourceFlags = resourceDescriptor.Flags;
 
     DAWN_TRY_ASSIGN(mResourceAllocation,
-                    ToBackend(GetDevice())
+                    device->AllocateMemory(D3D12_HEAP_TYPE_DEFAULT, resourceDescriptor,
-                        ->AllocateMemory(D3D12_HEAP_TYPE_DEFAULT, resourceDescriptor,
+                                           D3D12_RESOURCE_STATE_COMMON));
-                                         D3D12_RESOURCE_STATE_COMMON));
 
     SetLabelImpl();
 
-    Device* device = ToBackend(GetDevice());
+    if (applyForceClearCopyableDepthStencilTextureOnCreationToggle) {
-
-    if (device->IsToggleEnabled(
-            Toggle::D3D12ForceInitializeCopyableDepthStencilTextureOnCreation) &&
-        GetFormat().HasDepthOrStencil() && (GetInternalUsage() & wgpu::TextureUsage::CopyDst)) {
         CommandRecordingContext* commandContext;
         DAWN_TRY_ASSIGN(commandContext, device->GetPendingCommandContext());
         DAWN_TRY(ClearTexture(commandContext, GetAllSubresources(), TextureBase::ClearValue::Zero));

src/dawn/tests/end2end/CopyTests.cpp

@@ -2582,9 +2582,12 @@ std::ostream& operator<<(std::ostream& o, InitializationMethod method) {
     return o;
 }
 
+using AddRenderAttachmentUsage = bool;
+
 DAWN_TEST_PARAM_STRUCT(CopyToDepthStencilTextureAfterDestroyingBigBufferTestsParams,
                        TextureFormat,
-                       InitializationMethod);
+                       InitializationMethod,
+                       AddRenderAttachmentUsage);
 
 }  // anonymous namespace
 
@@ -2636,8 +2639,10 @@ TEST_P(CopyToDepthStencilTextureAfterDestroyingBigBufferTests, DoTest) {
     wgpu::TextureDescriptor textureDescriptor = {};
     textureDescriptor.format = format;
     textureDescriptor.size = {1, 1, 1};
-    textureDescriptor.usage = wgpu::TextureUsage::CopySrc | wgpu::TextureUsage::CopyDst |
+    textureDescriptor.usage = wgpu::TextureUsage::CopySrc | wgpu::TextureUsage::CopyDst;
-                              wgpu::TextureUsage::RenderAttachment;
+    if (GetParam().mAddRenderAttachmentUsage) {
+        textureDescriptor.usage |= wgpu::TextureUsage::RenderAttachment;
+    }
     wgpu::Texture texture = device.CreateTexture(&textureDescriptor);
 
     // Finally, upload valid data into the texture and validate its contents.
@@ -2723,9 +2728,9 @@ TEST_P(CopyToDepthStencilTextureAfterDestroyingBigBufferTests, DoTest) {
 
 DAWN_INSTANTIATE_TEST_P(
     CopyToDepthStencilTextureAfterDestroyingBigBufferTests,
-    {D3D12Backend(),
+    {D3D12Backend(), D3D12Backend({"d3d12_force_clear_copyable_depth_stencil_texture_on_creation"}),
-     D3D12Backend({"d3d12_force_initialize_copyable_depth_stencil_texture_on_creation"}),
      MetalBackend(), OpenGLBackend(), OpenGLESBackend(), VulkanBackend()},
     {wgpu::TextureFormat::Depth16Unorm, wgpu::TextureFormat::Stencil8},
     {InitializationMethod::CopyBufferToTexture, InitializationMethod::WriteTexture,
-     InitializationMethod::CopyTextureToTexture});
+     InitializationMethod::CopyTextureToTexture},
+    {true, false});