Propagating errors out of GetFunction in MTL backend

BUG=dawn:303

Change-Id: Iff1903aecae4c043b222208b3eab5efdf9774b52
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/14501
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>

src/dawn_native/metal/ComputePipelineMTL.h

@@ -25,7 +25,8 @@ namespace dawn_native { namespace metal {
 
     class ComputePipeline : public ComputePipelineBase {
       public:
-        ComputePipeline(Device* device, const ComputePipelineDescriptor* descriptor);
+        static ResultOrError<ComputePipeline*> Create(Device* device,
+                                                      const ComputePipelineDescriptor* descriptor);
         ~ComputePipeline();
 
         void Encode(id<MTLComputeCommandEncoder> encoder);
@@ -33,6 +34,9 @@ namespace dawn_native { namespace metal {
         bool RequiresStorageBufferLength() const;
 
       private:
+        using ComputePipelineBase::ComputePipelineBase;
+        MaybeError Initialize(const ComputePipelineDescriptor* descriptor);
+
         id<MTLComputePipelineState> mMtlComputePipelineState = nil;
         MTLSize mLocalWorkgroupSize;
         bool mRequiresStorageBufferLength;

src/dawn_native/metal/ComputePipelineMTL.mm

@@ -19,27 +19,37 @@
 
 namespace dawn_native { namespace metal {
 
-    ComputePipeline::ComputePipeline(Device* device, const ComputePipelineDescriptor* descriptor)
+    // static
-        : ComputePipelineBase(device, descriptor) {
+    ResultOrError<ComputePipeline*> ComputePipeline::Create(
+        Device* device,
+        const ComputePipelineDescriptor* descriptor) {
+        std::unique_ptr<ComputePipeline> pipeline =
+            std::make_unique<ComputePipeline>(device, descriptor);
+        DAWN_TRY(pipeline->Initialize(descriptor));
+        return pipeline.release();
+    }
+
+    MaybeError ComputePipeline::Initialize(const ComputePipelineDescriptor* descriptor) {
         auto mtlDevice = ToBackend(GetDevice())->GetMTLDevice();
 
         const ShaderModule* computeModule = ToBackend(descriptor->computeStage.module);
         const char* computeEntryPoint = descriptor->computeStage.entryPoint;
-        ShaderModule::MetalFunctionData computeData = computeModule->GetFunction(
+        ShaderModule::MetalFunctionData computeData;
-            computeEntryPoint, SingleShaderStage::Compute, ToBackend(GetLayout()));
+        DAWN_TRY(computeModule->GetFunction(computeEntryPoint, SingleShaderStage::Compute,
+                                            ToBackend(GetLayout()), &computeData));
 
         NSError* error = nil;
         mMtlComputePipelineState =
             [mtlDevice newComputePipelineStateWithFunction:computeData.function error:&error];
         if (error != nil) {
             NSLog(@" error => %@", error);
-            GetDevice()->HandleError(wgpu::ErrorType::DeviceLost, "Error creating pipeline state");
+            return DAWN_DEVICE_LOST_ERROR("Error creating pipeline state");
-            return;
         }
 
         // Copy over the local workgroup size as it is passed to dispatch explicitly in Metal
         mLocalWorkgroupSize = computeData.localWorkgroupSize;
         mRequiresStorageBufferLength = computeData.needsStorageBufferLength;
+        return {};
     }
 
     ComputePipeline::~ComputePipeline() {

src/dawn_native/metal/DeviceMTL.mm

@@ -111,7 +111,7 @@ namespace dawn_native { namespace metal {
     }
     ResultOrError<ComputePipelineBase*> Device::CreateComputePipelineImpl(
         const ComputePipelineDescriptor* descriptor) {
-        return new ComputePipeline(this, descriptor);
+        return ComputePipeline::Create(this, descriptor);
     }
     ResultOrError<PipelineLayoutBase*> Device::CreatePipelineLayoutImpl(
         const PipelineLayoutDescriptor* descriptor) {
@@ -122,7 +122,7 @@ namespace dawn_native { namespace metal {
     }
     ResultOrError<RenderPipelineBase*> Device::CreateRenderPipelineImpl(
         const RenderPipelineDescriptor* descriptor) {
-        return new RenderPipeline(this, descriptor);
+        return RenderPipeline::Create(this, descriptor);
     }
     ResultOrError<SamplerBase*> Device::CreateSamplerImpl(const SamplerDescriptor* descriptor) {
         return new Sampler(this, descriptor);

src/dawn_native/metal/RenderPipelineMTL.h

@@ -25,7 +25,8 @@ namespace dawn_native { namespace metal {
 
     class RenderPipeline : public RenderPipelineBase {
       public:
-        RenderPipeline(Device* device, const RenderPipelineDescriptor* descriptor);
+        static ResultOrError<RenderPipeline*> Create(Device* device,
+                                                     const RenderPipelineDescriptor* descriptor);
         ~RenderPipeline();
 
         MTLIndexType GetMTLIndexType() const;
@@ -44,6 +45,9 @@ namespace dawn_native { namespace metal {
         wgpu::ShaderStage GetStagesRequiringStorageBufferLength() const;
 
       private:
+        using RenderPipelineBase::RenderPipelineBase;
+        MaybeError Initialize(const RenderPipelineDescriptor* descriptor);
+
         MTLVertexDescriptor* MakeVertexDesc();
 
         MTLIndexType mMtlIndexType;

src/dawn_native/metal/RenderPipelineMTL.mm

@@ -311,20 +311,31 @@ namespace dawn_native { namespace metal {
 
     }  // anonymous namespace
 
-    RenderPipeline::RenderPipeline(Device* device, const RenderPipelineDescriptor* descriptor)
+    // static
-        : RenderPipelineBase(device, descriptor),
+    ResultOrError<RenderPipeline*> RenderPipeline::Create(
-          mMtlIndexType(MTLIndexFormat(GetVertexStateDescriptor()->indexFormat)),
+        Device* device,
-          mMtlPrimitiveTopology(MTLPrimitiveTopology(GetPrimitiveTopology())),
+        const RenderPipelineDescriptor* descriptor) {
-          mMtlFrontFace(MTLFrontFace(GetFrontFace())),
+        std::unique_ptr<RenderPipeline> pipeline =
-          mMtlCullMode(ToMTLCullMode(GetCullMode())) {
+            std::make_unique<RenderPipeline>(device, descriptor);
-        auto mtlDevice = device->GetMTLDevice();
+        DAWN_TRY(pipeline->Initialize(descriptor));
+        return pipeline.release();
+    }
+
+    MaybeError RenderPipeline::Initialize(const RenderPipelineDescriptor* descriptor) {
+        mMtlIndexType = MTLIndexFormat(GetVertexStateDescriptor()->indexFormat);
+        mMtlPrimitiveTopology = MTLPrimitiveTopology(GetPrimitiveTopology());
+        mMtlFrontFace = MTLFrontFace(GetFrontFace());
+        mMtlCullMode = ToMTLCullMode(GetCullMode());
+        auto mtlDevice = ToBackend(GetDevice())->GetMTLDevice();
 
         MTLRenderPipelineDescriptor* descriptorMTL = [MTLRenderPipelineDescriptor new];
 
         const ShaderModule* vertexModule = ToBackend(descriptor->vertexStage.module);
         const char* vertexEntryPoint = descriptor->vertexStage.entryPoint;
-        ShaderModule::MetalFunctionData vertexData = vertexModule->GetFunction(
+        ShaderModule::MetalFunctionData vertexData;
-            vertexEntryPoint, SingleShaderStage::Vertex, ToBackend(GetLayout()));
+        DAWN_TRY(vertexModule->GetFunction(vertexEntryPoint, SingleShaderStage::Vertex,
+                                           ToBackend(GetLayout()), &vertexData));
+
         descriptorMTL.vertexFunction = vertexData.function;
         if (vertexData.needsStorageBufferLength) {
             mStagesRequiringStorageBufferLength |= wgpu::ShaderStage::Vertex;
@@ -332,8 +343,10 @@ namespace dawn_native { namespace metal {
 
         const ShaderModule* fragmentModule = ToBackend(descriptor->fragmentStage->module);
         const char* fragmentEntryPoint = descriptor->fragmentStage->entryPoint;
-        ShaderModule::MetalFunctionData fragmentData = fragmentModule->GetFunction(
+        ShaderModule::MetalFunctionData fragmentData;
-            fragmentEntryPoint, SingleShaderStage::Fragment, ToBackend(GetLayout()));
+        DAWN_TRY(fragmentModule->GetFunction(fragmentEntryPoint, SingleShaderStage::Fragment,
+                                             ToBackend(GetLayout()), &fragmentData));
+
         descriptorMTL.fragmentFunction = fragmentData.function;
         if (fragmentData.needsStorageBufferLength) {
             mStagesRequiringStorageBufferLength |= wgpu::ShaderStage::Fragment;
@@ -372,9 +385,7 @@ namespace dawn_native { namespace metal {
             [descriptorMTL release];
             if (error != nil) {
                 NSLog(@" error => %@", error);
-                device->HandleError(wgpu::ErrorType::DeviceLost,
+                return DAWN_DEVICE_LOST_ERROR("Error creating rendering pipeline state");
-                                    "Error creating rendering pipeline state");
-                return;
             }
         }
 
@@ -385,6 +396,8 @@ namespace dawn_native { namespace metal {
             MakeDepthStencilDesc(GetDepthStencilStateDescriptor());
         mMtlDepthStencilState = [mtlDevice newDepthStencilStateWithDescriptor:depthStencilDesc];
         [depthStencilDesc release];
+
+        return {};
     }
 
     RenderPipeline::~RenderPipeline() {

src/dawn_native/metal/ShaderModuleMTL.h

@@ -43,9 +43,10 @@ namespace dawn_native { namespace metal {
                 [function release];
             }
         };
-        MetalFunctionData GetFunction(const char* functionName,
+        MaybeError GetFunction(const char* functionName,
                                SingleShaderStage functionStage,
-                                      const PipelineLayout* layout) const;
+                               const PipelineLayout* layout,
+                               MetalFunctionData* out) const;
 
       private:
         ShaderModule(Device* device, const ShaderModuleDescriptor* descriptor);

src/dawn_native/metal/ShaderModuleMTL.mm

@@ -92,17 +92,22 @@ namespace dawn_native { namespace metal {
         return {};
     }
 
-    ShaderModule::MetalFunctionData ShaderModule::GetFunction(const char* functionName,
+    MaybeError ShaderModule::GetFunction(const char* functionName,
                                          SingleShaderStage functionStage,
-                                                              const PipelineLayout* layout) const {
+                                         const PipelineLayout* layout,
+                                         ShaderModule::MetalFunctionData* out) const {
+        ASSERT(!IsError());
+        ASSERT(out);
         std::unique_ptr<spirv_cross::CompilerMSL> compiler_impl;
         spirv_cross::CompilerMSL* compiler;
         if (GetDevice()->IsToggleEnabled(Toggle::UseSpvc)) {
             // Initializing the compiler is needed every call, because this method uses reflection
             // to mutate the compiler's IR.
-            mSpvcContext.InitializeForMsl(mSpirv.data(), mSpirv.size(), GetMSLCompileOptions());
+            if (mSpvcContext.InitializeForMsl(mSpirv.data(), mSpirv.size(),
-            // TODO(rharrison): Handle initialize failing
+                                              GetMSLCompileOptions()) !=
-
+                shaderc_spvc_status_success) {
+                return DAWN_DEVICE_LOST_ERROR("Unable to initialize instance of spvc");
+            }
             compiler = reinterpret_cast<spirv_cross::CompilerMSL*>(mSpvcContext.GetCompiler());
         } else {
             // If these options are changed, the values in DawnSPIRVCrossMSLFastFuzzer.cpp need to
@@ -147,12 +152,10 @@ namespace dawn_native { namespace metal {
             }
         }
 
-        MetalFunctionData result;
-
         {
             spv::ExecutionModel executionModel = SpirvExecutionModelForStage(functionStage);
             auto size = compiler->get_entry_point(functionName, executionModel).workgroup_size;
-            result.localWorkgroupSize = MTLSizeMake(size.x, size.y, size.z);
+            out->localWorkgroupSize = MTLSizeMake(size.x, size.y, size.z);
         }
 
         {
@@ -167,9 +170,14 @@ namespace dawn_native { namespace metal {
                                                              options:nil
                                                                error:&error];
             if (error != nil) {
-                // TODO(cwallez@chromium.org): forward errors to caller
+                // TODO(cwallez@chromium.org): Switch that NSLog to use dawn::InfoLog or even be
+                // folded in the DAWN_VALIDATION_ERROR
                 NSLog(@"MTLDevice newLibraryWithSource => %@", error);
+                if (error.code != MTLLibraryErrorCompileWarning) {
+                    return DAWN_VALIDATION_ERROR("Unable to create library object");
                 }
+            }
+
             // TODO(kainino@chromium.org): make this somehow more robust; it needs to behave like
             // clean_func_name:
             // https://github.com/KhronosGroup/SPIRV-Cross/blob/4e915e8c483e319d0dd7a1fa22318bef28f8cca3/spirv_msl.cpp#L1213
@@ -178,13 +186,13 @@ namespace dawn_native { namespace metal {
             }
 
             NSString* name = [NSString stringWithFormat:@"%s", functionName];
-            result.function = [library newFunctionWithName:name];
+            out->function = [library newFunctionWithName:name];
             [library release];
         }
 
-        result.needsStorageBufferLength = compiler->needs_buffer_size_buffer();
+        out->needsStorageBufferLength = compiler->needs_buffer_size_buffer();
 
-        return result;
+        return {};
     }
 
 }}  // namespace dawn_native::metal

src/tests/end2end/ObjectCachingTests.cpp

@@ -139,14 +139,16 @@ TEST_P(ObjectCachingTest, ComputePipelineDeduplicationOnShaderModule) {
     wgpu::ShaderModule module =
         utils::CreateShaderModule(device, utils::SingleShaderStage::Compute, R"(
             #version 450
+            shared uint i;
             void main() {
-                int i = 0;
+                i = 0;
             })");
     wgpu::ShaderModule sameModule =
         utils::CreateShaderModule(device, utils::SingleShaderStage::Compute, R"(
             #version 450
+            shared uint i;
             void main() {
-                int i = 0;
+                i = 0;
             })");
     wgpu::ShaderModule otherModule =
         utils::CreateShaderModule(device, utils::SingleShaderStage::Compute, R"(
@@ -195,8 +197,9 @@ TEST_P(ObjectCachingTest, ComputePipelineDeduplicationOnLayout) {
     desc.computeStage.module =
         utils::CreateShaderModule(device, utils::SingleShaderStage::Compute, R"(
             #version 450
+            shared uint i;
             void main() {
-                int i = 0;
+                i = 0;
             })");
 
     desc.layout = pl;
@@ -311,8 +314,9 @@ TEST_P(ObjectCachingTest, RenderPipelineDeduplicationOnFragmentModule) {
     wgpu::ShaderModule otherModule =
         utils::CreateShaderModule(device, utils::SingleShaderStage::Fragment, R"(
             #version 450
+            layout (location = 0) out vec4 color;
             void main() {
-                int i = 0;
+                color = vec4(0.0);
             })");
 
     EXPECT_NE(module.Get(), otherModule.Get());