Merge branch 'main' into enable-passing-vk-tests

s-perron · web-flow · commit 9cdcd6a47b21 · 2025-08-29T11:12:26.000-04:00
diff --git a/.github/workflows/build-and-test-callable.yaml b/.github/workflows/build-and-test-callable.yaml
@@ -135,11 +135,16 @@ jobs:
           ref: main
           path: golden-images
           fetch-depth: 1
-      - name: Setup Windows
-        if: inputs.OS == 'windows'
+      - name: Setup Windows x64
+        if: inputs.OS == 'windows' && runner.arch != 'ARM64'
         uses: llvm/actions/setup-windows@main
         with:
           arch: amd64
+      - name: Setup Windows
+        if: inputs.OS == 'windows' && runner.arch == 'ARM64'
+        uses: llvm/actions/setup-windows@main
+        with:
+          arch: arm64
       - name: Detect Clang
         id: Test-Clang
         shell: bash
diff --git a/.github/workflows/pr-matrix.yaml b/.github/workflows/pr-matrix.yaml
@@ -49,7 +49,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        SKU: [windows-nvidia, windows-amd]
+        SKU: [windows-nvidia, windows-amd, windows-qc]
         TestTarget: [check-hlsl-d3d12, check-hlsl-vk, check-hlsl-clang-d3d12, check-hlsl-clang-vk]
 
     uses: ./.github/workflows/build-and-test-callable.yaml
diff --git a/.github/workflows/windows-qc-clang-d3d12.yaml b/.github/workflows/windows-qc-clang-d3d12.yaml
@@ -0,0 +1,20 @@
+name: Windows D3D12 AMD Clang
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-D3D12-AMD-Clang:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      TestTarget: check-hlsl-clang-d3d12
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DOFFLOADTEST_USE_CLANG_TIDY=On
diff --git a/.github/workflows/windows-qc-clang-vk.yaml b/.github/workflows/windows-qc-clang-vk.yaml
@@ -0,0 +1,20 @@
+name: Windows Vulkan AMD Clang
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-VK-AMD-Clang:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      TestTarget: check-hlsl-clang-vk
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DOFFLOADTEST_USE_CLANG_TIDY=On
diff --git a/.github/workflows/windows-qc-clang-warp-d3d12.yaml b/.github/workflows/windows-qc-clang-warp-d3d12.yaml
@@ -0,0 +1,20 @@
+name: Windows D3D12 Warp Clang
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-D3D12-Warp-Clang:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      TestTarget: check-hlsl-clang-warp-d3d12
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DOFFLOADTEST_USE_CLANG_TIDY=On
diff --git a/.github/workflows/windows-qc-dxc-d3d12.yaml b/.github/workflows/windows-qc-dxc-d3d12.yaml
@@ -0,0 +1,21 @@
+name: Windows D3D12 AMD DXC
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-D3D12-AMD-DXC:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      BuildType: Debug
+      TestTarget: check-hlsl-d3d12
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl
diff --git a/.github/workflows/windows-qc-dxc-vk.yaml b/.github/workflows/windows-qc-dxc-vk.yaml
@@ -0,0 +1,21 @@
+name: Windows Vulkan AMD DXC
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-VK-AMD-DXC:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      BuildType: Debug
+      TestTarget: check-hlsl-vk
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl
diff --git a/.github/workflows/windows-qc-dxc-warp-d3d12.yaml b/.github/workflows/windows-qc-dxc-warp-d3d12.yaml
@@ -0,0 +1,21 @@
+name: Windows D3D12 Warp DXC
+
+permissions:
+  contents: read
+  checks: write
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 */2 * * *' # run every 2 hours
+
+jobs:
+  Windows-D3D12-Warp-DXC:
+    uses: ./.github/workflows/build-and-test-callable.yaml
+    with:
+      OS: windows
+      SKU: windows-amd
+      BuildType: Debug
+      TestTarget: check-hlsl-warp-d3d12
+      OffloadTest-branch: ${{ github.ref }}
+      LLVM-ExtraCMakeArgs: -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl
diff --git a/README.md b/README.md
@@ -8,11 +8,14 @@ Experimental Runtime test suite for HLSL
 | Windows DirectX12 AMD GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-dxc-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-clang-d3d12.yaml/badge.svg) |
 | Windows DirectX12 Intel GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-intel-dxc-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-intel-clang-d3d12.yaml/badge.svg) |
 | Windows DirectX12 NVIDIA GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-nvidia-dxc-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-nvidia-clang-d3d12.yaml/badge.svg) |
-| Windows DirectX12 Warp (LKG) | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-dxc-warp-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-clang-warp-d3d12.yaml/badge.svg) |
+| Windows DirectX12 Qualcomm GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-dxc-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-clang-d3d12.yaml/badge.svg) |
+| Windows DirectX12 Warp (x64 LKG) | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-dxc-warp-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-clang-warp-d3d12.yaml/badge.svg) |
+| Windows DirectX12 Warp (arm64 LKG) | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-dxc-warp-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-clang-warp-d3d12.yaml/badge.svg) |
 | Windows DirectX12 Warp (1.0.15-preview) | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-dxc-warp-preview-d3d12.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-clang-warp-preview-d3d12.yaml/badge.svg) |
 | Windows Vulkan AMD GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-dxc-vk.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-amd-clang-vk.yaml/badge.svg) |
 | Windows Vulkan Intel GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-intel-dxc-vk.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-intel-clang-vk.yaml/badge.svg) |
 | Windows Vulkan NVIDIA GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-nvidia-dxc-vk.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-nvidia-clang-vk.yaml/badge.svg) |
+| Windows Vulkan Qualcomm GPU | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-dxc-vk.yaml/badge.svg) | ![Clang](https://github.com/llvm/offload-test-suite/actions/workflows/windows-qc-clang-vk.yaml/badge.svg) |
 | macOS Apple M1 | ![DXC](https://github.com/llvm/offload-test-suite/actions/workflows/macos-dxc-mtl.yaml/badge.svg) | ![Clang & DXC](https://github.com/llvm/offload-test-suite/actions/workflows/macos-clang-mtl.yaml/badge.svg) |
 
 
diff --git a/include/API/Device.h b/include/API/Device.h
@@ -30,6 +30,11 @@ namespace offloadtest {
 
 struct Pipeline;
 
+struct DeviceConfig {
+  bool EnableDebugLayer = false;
+  bool EnableValidationLayer = false;
+};
+
 class Device {
 protected:
   std::string Description;
@@ -46,7 +51,7 @@ class Device {
   llvm::StringRef getDescription() const { return Description; }
 
   static void registerDevice(std::shared_ptr<Device> D);
-  static llvm::Error initialize();
+  static llvm::Error initialize(const DeviceConfig Config);
   static void uninitialize();
 
   using DeviceArray = llvm::SmallVector<std::shared_ptr<Device>>;
@@ -58,16 +63,16 @@ class Device {
   static inline DeviceRange devices() { return DeviceRange(begin(), end()); }
 
 #ifdef OFFLOADTEST_ENABLE_D3D12
-  static llvm::Error initializeDXDevices();
+  static llvm::Error initializeDXDevices(const DeviceConfig Config);
 #endif
 
 #ifdef OFFLOADTEST_ENABLE_VULKAN
-  static llvm::Error initializeVKDevices();
+  static llvm::Error initializeVKDevices(const DeviceConfig Config);
   static void cleanupVKDevices();
 #endif
 
 #ifdef OFFLOADTEST_ENABLE_METAL
-  static llvm::Error initializeMtlDevices();
+  static llvm::Error initializeMtlDevices(const DeviceConfig Config);
 #endif
 };
 
diff --git a/lib/API/DX/Device.cpp b/lib/API/DX/Device.cpp
@@ -287,7 +287,8 @@ class DXDevice : public offloadtest::Device {
   llvm::StringRef getAPIName() const override { return "DirectX"; }
   GPUAPI getAPI() const override { return GPUAPI::DirectX; }
 
-  static llvm::Expected<DXDevice> create(ComPtr<IDXCoreAdapter> Adapter) {
+  static llvm::Expected<DXDevice> create(ComPtr<IDXCoreAdapter> Adapter,
+                                         const DeviceConfig &Config) {
     ComPtr<ID3D12Device> Device;
     if (auto Err =
             HR::toError(D3D12CreateDevice(Adapter.Get(), D3D_FEATURE_LEVEL_11_0,
@@ -302,8 +303,9 @@ class DXDevice : public offloadtest::Device {
     std::vector<char> DescVec(BufferSize);
     Adapter->GetProperty(DXCoreAdapterProperty::DriverDescription, BufferSize,
                          (void *)DescVec.data());
-    if (auto Err = configureInfoQueue(Device.Get()))
-      return Err;
+    if (Config.EnableDebugLayer || Config.EnableValidationLayer)
+      if (auto Err = configureInfoQueue(Device.Get()))
+        return Err;
     return DXDevice(Adapter, Device, std::string(DescVec.data()));
   }
 
@@ -335,15 +337,13 @@ class DXDevice : public offloadtest::Device {
 
   static llvm::Error configureInfoQueue(ID3D12Device *Device) {
 #ifdef _WIN32
-#ifndef NDEBUG
     ComPtr<ID3D12InfoQueue> InfoQueue;
     if (auto Err = HR::toError(Device->QueryInterface(InfoQueue.GetAddressOf()),
                                "Error initializing info queue"))
       return Err;
     InfoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_CORRUPTION, TRUE);
     InfoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_ERROR, TRUE);
     InfoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_WARNING, TRUE);
-#endif
 #endif
     return llvm::Error::success();
   }
@@ -1169,18 +1169,18 @@ class DXDevice : public offloadtest::Device {
 };
 } // namespace
 
-llvm::Error Device::initializeDXDevices() {
+llvm::Error Device::initializeDXDevices(const DeviceConfig Config) {
 #ifdef _WIN32
-#ifndef NDEBUG
-  ComPtr<ID3D12Debug1> Debug1;
+  if (Config.EnableDebugLayer || Config.EnableValidationLayer) {
+    ComPtr<ID3D12Debug1> Debug1;
 
-  if (auto Err = HR::toError(D3D12GetDebugInterface(IID_PPV_ARGS(&Debug1)),
-                             "failed to create D3D12 Debug Interface"))
-    return Err;
+    if (auto Err = HR::toError(D3D12GetDebugInterface(IID_PPV_ARGS(&Debug1)),
+                               "failed to create D3D12 Debug Interface"))
+      return Err;
 
-  Debug1->EnableDebugLayer();
-  Debug1->SetEnableGPUBasedValidation(true);
-#endif
+    Debug1->EnableDebugLayer();
+    Debug1->SetEnableGPUBasedValidation(Config.EnableValidationLayer);
+  }
 #endif
 
   ComPtr<IDXCoreAdapterFactory> Factory;
@@ -1207,7 +1207,7 @@ llvm::Error Device::initializeDXDevices() {
             "Failed to acquire adapter")) {
       return Err;
     }
-    auto ExDevice = DXDevice::create(Adapter);
+    auto ExDevice = DXDevice::create(Adapter, Config);
     if (!ExDevice)
       return ExDevice.takeError();
     auto ShPtr = std::make_shared<DXDevice>(*ExDevice);
diff --git a/lib/API/Device.cpp b/lib/API/Device.cpp
@@ -54,13 +54,13 @@ void Device::registerDevice(std::shared_ptr<Device> D) {
   DeviceContext::instance().registerDevice(D);
 }
 
-llvm::Error Device::initialize() {
+llvm::Error Device::initialize(const DeviceConfig Config) {
 #ifdef OFFLOADTEST_ENABLE_D3D12
-  if (auto Err = initializeDXDevices())
+  if (auto Err = initializeDXDevices(Config))
     return Err;
 #endif
 #ifdef OFFLOADTEST_ENABLE_VULKAN
-  if (auto Err = initializeVKDevices())
+  if (auto Err = initializeVKDevices(Config))
     return Err;
   // Validation layers have internal state which require a specific destruction
   // ordering. Relying on the global dtor call for this is unreliable and can
@@ -70,7 +70,7 @@ llvm::Error Device::initialize() {
   atexit(Device::cleanupVKDevices);
 #endif
 #ifdef OFFLOADTEST_ENABLE_METAL
-  if (auto Err = initializeMtlDevices())
+  if (auto Err = initializeMtlDevices(Config))
     return Err;
 #endif
   return llvm::Error::success();
diff --git a/lib/API/MTL/MTLDevice.cpp b/lib/API/MTL/MTLDevice.cpp
@@ -304,6 +304,6 @@ class MTLContext {
 
 } // namespace
 
-llvm::Error Device::initializeMtlDevices() {
+llvm::Error Device::initializeMtlDevices(const DeviceConfig /*Config*/) {
   return MTLContext::instance().initialize();
 }
diff --git a/lib/API/VK/Device.cpp b/lib/API/VK/Device.cpp
@@ -1238,7 +1238,7 @@ class VKContext {
     Instance = VK_NULL_HANDLE;
   }
 
-  llvm::Error initialize() {
+  llvm::Error initialize(const DeviceConfig Config) {
     // Create a Vulkan 1.1 instance to determine the API version
     VkApplicationInfo AppInfo = {};
     AppInfo.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
@@ -1286,15 +1286,18 @@ class VKContext {
       auto TmpDev = std::make_shared<VKDevice>(PhysicalDevicesTmp[0]);
       AppInfo.apiVersion = TmpDev->getProps().apiVersion;
 
-#ifndef NDEBUG
-      const llvm::StringRef ValidationLayer = "VK_LAYER_KHRONOS_validation";
-      if (TmpDev->isLayerSupported(ValidationLayer))
-        Layers.push_back(ValidationLayer.data());
+      if (Config.EnableValidationLayer) {
+        const llvm::StringRef ValidationLayer = "VK_LAYER_KHRONOS_validation";
+        if (TmpDev->isLayerSupported(ValidationLayer))
+          Layers.push_back(ValidationLayer.data());
+      }
+
+      if (Config.EnableDebugLayer) {
+        const llvm::StringRef DebugUtilsExtensionName = "VK_EXT_debug_utils";
+        if (TmpDev->isExtensionSupported(DebugUtilsExtensionName))
+          Extensions.push_back(DebugUtilsExtensionName.data());
+      }
 
-      const llvm::StringRef DebugUtilsExtensionName = "VK_EXT_debug_utils";
-      if (TmpDev->isExtensionSupported(DebugUtilsExtensionName))
-        Extensions.push_back(DebugUtilsExtensionName.data());
-#endif
       CreateInfo.ppEnabledLayerNames = Layers.data();
       CreateInfo.enabledLayerCount = Layers.size();
       CreateInfo.ppEnabledExtensionNames = Extensions.data();
@@ -1338,8 +1341,8 @@ class VKContext {
 };
 } // namespace
 
-llvm::Error Device::initializeVKDevices() {
-  return VKContext::instance().initialize();
+llvm::Error Device::initializeVKDevices(const DeviceConfig Config) {
+  return VKContext::instance().initialize(Config);
 }
 
 void Device::cleanupVKDevices() { VKContext::instance().cleanup(); }
diff --git a/test/Bugs/UAV-Sequental-Consistency.yaml b/test/Bugs/UAV-Sequental-Consistency.yaml
@@ -60,9 +60,11 @@ DescriptorSets:
 # UHD drivers. This issue does not reproduce when compiled with Clang because
 # LLVM's load-store optimization performs correct optimizations that hide the
 # problem.
+# This test fails randomly but inconsistently without validation enabled. With
+# validation enabled it fails consistently every time.
 # Tracking issue: https://github.com/llvm/offload-test-suite/issues/226
 # XFAIL: Intel-Memory-Coherence-Issue-226 && !Clang
 
 # RUN: split-file %s %t
 # RUN: %dxc_target -T cs_6_5 -Fo %t.o %t/source.hlsl
-# RUN: %offloader %t/pipeline.yaml %t.o
+# RUN: %offloader -validation-layer %t/pipeline.yaml %t.o
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
diff --git a/test/lit.cfg.py b/test/lit.cfg.py
diff --git a/test/lit.site.cfg.py.in b/test/lit.site.cfg.py.in
diff --git a/tools/api-query/api-query.cpp b/tools/api-query/api-query.cpp
diff --git a/tools/offloader/offloader.cpp b/tools/offloader/offloader.cpp
