Fix some warnings that only happen in a release build

microsoft · Feb 27, 2024 · a43aaad · a43aaad
1 parent a6aa359
commit a43aaad
Show file tree

Hide file tree

Showing 3 changed files with 28 additions and 7 deletions.
diff --git a/onnxruntime/contrib_ops/cuda/bert/packed_multihead_attention_impl.cu b/onnxruntime/contrib_ops/cuda/bert/packed_multihead_attention_impl.cu
@@ -381,7 +381,7 @@ void InvokeTranspose(
     const T* query, const T* key, const T* value, const T* bias, T* output,
     const int batch_size, const int sequence_length,
     const int num_heads, const int qk_head_size, const int v_head_size,
-    AttentionQkvFormat source_format, AttentionQkvFormat target_format,
+    [[maybe_unused]] AttentionQkvFormat source_format, AttentionQkvFormat target_format,
     const int32_t* token_offset, int32_t token_count,
     cudaStream_t stream) {
   if (key != nullptr && value != nullptr) {

diff --git a/onnxruntime/contrib_ops/cuda/moe/ft_moe/moe_gemm_kernels_template.h b/onnxruntime/contrib_ops/cuda/moe/ft_moe/moe_gemm_kernels_template.h
@@ -20,6 +20,12 @@
 #pragma GCC diagnostic ignored "-Wstrict-aliasing"
 #endif
 
+// Ignore CUTLASS warning C4100: unreferenced formal parameter
+#if defined(_MSC_VER)
+#pragma warning(push)
+#pragma warning(disable : 4100)
+#endif
+
 #include "cutlass/array.h"
 #include "cutlass/numeric_conversion.h"
 #include "cutlass/layout/matrix.h"
@@ -36,6 +42,10 @@
 #include "layout_traits_helper.h"
 #include "moe_cutlass_kernel.h"
 
+#if defined(_MSC_VER)
+#pragma warning(pop)
+#endif
+
 #ifdef __GNUC__
 #pragma GCC diagnostic pop
 #endif

diff --git a/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h b/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h
@@ -13,7 +13,8 @@ using namespace onnxruntime;
 namespace onnxruntime {
 
 common::Status LoadDynamicLibrary(onnxruntime::PathString library_name);
-common::Status CreateTensorRTCustomOpDomainList(std::vector<OrtCustomOpDomain*>& domain_list, const std::string extra_plugin_lib_paths);
+common::Status CreateTensorRTCustomOpDomainList(std::vector<OrtCustomOpDomain*>& domain_list,
+                                                const std::string extra_plugin_lib_paths);
 common::Status CreateTensorRTCustomOpDomainList(TensorrtExecutionProviderInfo& info);
 void ReleaseTensorRTCustomOpDomain(OrtCustomOpDomain* domain);
 void ReleaseTensorRTCustomOpDomainList(std::vector<OrtCustomOpDomain*>& custom_op_domain_list);
@@ -23,16 +24,22 @@ struct TensorRTCustomKernel {
       : compute_stream_(compute_stream) {
   }
 
-  void Compute(OrtKernelContext* context){};  // The implementation is in TensorRT plugin. No need to implement it here.
+  void Compute(OrtKernelContext* /*context*/){
+      // The implementation is in TensorRT plugin. No need to implement it here.
+  };
 
  private:
   void* compute_stream_;
 };
 
 struct TensorRTCustomOp : Ort::CustomOpBase<TensorRTCustomOp, TensorRTCustomKernel> {
-  explicit TensorRTCustomOp(const char* provider, void* compute_stream) : provider_(provider), compute_stream_(compute_stream) {}
+  explicit TensorRTCustomOp(const char* provider, void* compute_stream) : provider_(provider),
+                                                                          compute_stream_(compute_stream) {
+  }
 
-  void* CreateKernel(const OrtApi& /* api */, const OrtKernelInfo* info) const { return new TensorRTCustomKernel(info, compute_stream_); };
+  void* CreateKernel(const OrtApi& /* api */, const OrtKernelInfo* info) const {
+    return new TensorRTCustomKernel(info, compute_stream_);
+  };
 
   const char* GetName() const { return name_; };
 
@@ -46,15 +53,19 @@ struct TensorRTCustomOp : Ort::CustomOpBase<TensorRTCustomOp, TensorRTCustomKern
 
   ONNXTensorElementDataType GetInputType(size_t /*index*/) const { return ONNX_TENSOR_ELEMENT_DATA_TYPE_UNDEFINED; };
 
-  OrtCustomOpInputOutputCharacteristic GetInputCharacteristic(size_t) const { return OrtCustomOpInputOutputCharacteristic::INPUT_OUTPUT_VARIADIC; };
+  OrtCustomOpInputOutputCharacteristic GetInputCharacteristic(size_t) const {
+    return OrtCustomOpInputOutputCharacteristic::INPUT_OUTPUT_VARIADIC;
+  };
 
   size_t GetOutputTypeCount() const { return num_outputs_; };
 
   void SetOutputTypeCount(size_t num) { num_outputs_ = num; };
 
   ONNXTensorElementDataType GetOutputType(size_t /*index*/) const { return ONNX_TENSOR_ELEMENT_DATA_TYPE_UNDEFINED; };
 
-  OrtCustomOpInputOutputCharacteristic GetOutputCharacteristic(size_t) const { return OrtCustomOpInputOutputCharacteristic::INPUT_OUTPUT_VARIADIC; };
+  OrtCustomOpInputOutputCharacteristic GetOutputCharacteristic(size_t) const {
+    return OrtCustomOpInputOutputCharacteristic::INPUT_OUTPUT_VARIADIC;
+  };
 
   bool GetVariadicInputHomogeneity() const {
     return false;  // heterogenous