Mark unused parameters in C++ code (pytorch#164912)

This PR adds unused parameter name comments in C++ declarations to improve code readability.

Pull Request resolved: pytorch#164912
Approved by: https://github.com/Skylion007

Author: cyyever

aten/src/ATen/Context.h

@@ -226,15 +226,15 @@ class TORCH_API Context {
   bool userEnabledMkldnn() const;
   void setUserEnabledMkldnn(bool e);
   bool benchmarkCuDNN() const;
-  void setBenchmarkCuDNN(bool);
+  void setBenchmarkCuDNN(bool /*b*/);
   int benchmarkLimitCuDNN() const;
-  void setBenchmarkLimitCuDNN(int);
+  void setBenchmarkLimitCuDNN(int /*b*/);
   bool immediateMiopen() const;
-  void setImmediateMiopen(bool);
+  void setImmediateMiopen(bool /*b*/);
   bool deterministicCuDNN() const;
-  void setDeterministicCuDNN(bool);
+  void setDeterministicCuDNN(bool /*b*/);
   bool deterministicMkldnn() const;
-  void setDeterministicMkldnn(bool);
+  void setDeterministicMkldnn(bool /*b*/);
   bool userEnabledNNPACK() const;
   void setUserEnabledNNPACK(bool e);
 
@@ -252,32 +252,32 @@ class TORCH_API Context {
   void setSDPPriorityOrder(const std::vector<int64_t>& order);
   std::array<at::SDPBackend, at::num_sdp_backends> sDPPriorityOrder();
 
-  void setSDPUseFlash(bool);
+  void setSDPUseFlash(bool /*e*/);
   bool userEnabledFlashSDP() const;
 
-  void setSDPUseMemEfficient(bool);
+  void setSDPUseMemEfficient(bool /*e*/);
   bool userEnabledMemEfficientSDP() const;
 
-  void setSDPUseMath(bool);
+  void setSDPUseMath(bool /*e*/);
   bool userEnabledMathSDP() const;
 
-  void setSDPUseCuDNN(bool);
+  void setSDPUseCuDNN(bool /*e*/);
   bool userEnabledCuDNNSDP() const;
 
-  void setAllowFP16BF16ReductionMathSDP(bool);
+  void setAllowFP16BF16ReductionMathSDP(bool /*e*/);
   bool allowFP16BF16ReductionMathSDP() const;
 
-  void setSDPUseOverrideable(bool);
+  void setSDPUseOverrideable(bool /*e*/);
   bool userEnabledOverrideableSDP() const;
 
   at::LinalgBackend linalgPreferredBackend() const;
-  void setLinalgPreferredBackend(at::LinalgBackend);
+  void setLinalgPreferredBackend(at::LinalgBackend /*b*/);
 
   at::BlasBackend blasPreferredBackend();
-  void setBlasPreferredBackend(at::BlasBackend);
+  void setBlasPreferredBackend(at::BlasBackend /*b*/);
 
   at::ROCmFABackend getROCmFAPreferredBackend();
-  void setROCmFAPreferredBackend(at::ROCmFABackend);
+  void setROCmFAPreferredBackend(at::ROCmFABackend /*b*/);
 
   // Note [Enabling Deterministic Operations]
   // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -310,9 +310,9 @@ class TORCH_API Context {
 
   bool deterministicAlgorithms() const;
   bool deterministicAlgorithmsWarnOnly() const;
-  void setDeterministicAlgorithms(bool, bool);
+  void setDeterministicAlgorithms(bool /*b*/, bool /*warn_only*/);
   bool deterministicFillUninitializedMemory() const;
-  void setDeterministicFillUninitializedMemory(bool);
+  void setDeterministicFillUninitializedMemory(bool /*b*/);
 
   // Note [Writing Nondeterministic Operations]
   // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -356,11 +356,11 @@ class TORCH_API Context {
       Float32Op op,
       Float32Precision p);
   bool allowTF32CuDNN(std::optional<Float32Op> op = std::nullopt) const;
-  void setAllowTF32CuDNN(bool);
+  void setAllowTF32CuDNN(bool /*b*/);
   bool allowTF32OneDNN() const;
-  void setAllowTF32OneDNN(bool);
+  void setAllowTF32OneDNN(bool /*b*/);
   bool allowTF32CuBLAS() const;
-  void setAllowTF32CuBLAS(bool);
+  void setAllowTF32CuBLAS(bool /*b*/);
   Float32MatmulPrecision float32MatmulPrecision() const;
   Float32Precision float32Precision(Float32Backend backend, Float32Op op) const;
   CuBLASReductionOption allowFP16ReductionCuBLAS() const;
@@ -372,7 +372,7 @@ class TORCH_API Context {
       bool allow_reduced_precision,
       bool allow_splitk = true);
   bool allowFP16AccumulationCuBLAS() const;
-  void setAllowFP16AccumulationCuBLAS(bool);
+  void setAllowFP16AccumulationCuBLAS(bool /*b*/);
 
   // Matmuls can use a so-called "persistent" kernel which launches one CUDA
   // block for each SM on the GPU, and each block then iterates over multiple
@@ -384,7 +384,7 @@ class TORCH_API Context {
   // to make matmuls target only a subset of the SMs, so they can fully schedule
   // even next to a comms kernel, and only be a few percent slower.
   std::optional<int32_t> _SMCarveout_EXPERIMENTAL() const;
-  void _setSMCarveout_EXPERIMENTAL(std::optional<int32_t>);
+  void _setSMCarveout_EXPERIMENTAL(std::optional<int32_t> /*c*/);
 
   at::QEngine qEngine() const;
   void setQEngine(at::QEngine e);
@@ -405,7 +405,7 @@ class TORCH_API Context {
   void setDefaultMobileCPUAllocator();
   void unsetDefaultMobileCPUAllocator();
   bool allowFP16ReductionCPU() const;
-  void setAllowFP16ReductionCPU(bool);
+  void setAllowFP16ReductionCPU(bool /*b*/);
 
   // Preserved for BC
   void lazyInitCUDA() {

aten/src/ATen/MapAllocator.cpp

@@ -62,7 +62,7 @@ constexpr const char* unknown_eventname = "eventname not specified";
 #endif
 }  // namespace (anonymous)
 
-MapAllocator::MapAllocator(WithFd, std::string_view filename, int fd, int flags, size_t size)
+MapAllocator::MapAllocator(WithFd /*unused*/, std::string_view filename, int fd, int flags, size_t size)
   : filename_(filename.empty() ? unknown_filename : filename)
   , size_(0) // to be filled later
 #ifdef _WIN32
@@ -494,7 +494,7 @@ RefcountedMapAllocator::RefcountedMapAllocator(const char *filename, int flags,
 
     initializeAlloc();
 }
-RefcountedMapAllocator::RefcountedMapAllocator(WithFd, const char *filename, int fd, int flags, size_t size)
+RefcountedMapAllocator::RefcountedMapAllocator(WithFd /*unused*/, const char *filename, int fd, int flags, size_t size)
   : RefcountedMapAllocatorArgCheck(flags)
   , MapAllocator(WITH_FD, filename, flags, fd, size + map_alloc_alignment) {
 
@@ -614,7 +614,7 @@ at::DataPtr MapAllocator::makeDataPtr(std::string_view filename, int flags, size
   return {context->data(), context, &deleteMapAllocator, at::DeviceType::CPU};
 }
 
-at::DataPtr MapAllocator::makeDataPtr(WithFd, const char *filename, int fd, int flags, size_t size, size_t* actual_size_out) {
+at::DataPtr MapAllocator::makeDataPtr(WithFd /*unused*/, const char *filename, int fd, int flags, size_t size, size_t* actual_size_out) {
   auto* context = new MapAllocator(WITH_FD, filename, fd, flags, size);
   if (actual_size_out) *actual_size_out = context->size();
   return {context->data(), context, &deleteMapAllocator, at::DeviceType::CPU};
@@ -626,7 +626,7 @@ at::DataPtr RefcountedMapAllocator::makeDataPtr(const char *filename, int flags,
   return {context->data(), context, &deleteRefcountedMapAllocator, at::DeviceType::CPU};
 }
 
-at::DataPtr RefcountedMapAllocator::makeDataPtr(WithFd, const char *filename, int fd, int flags, size_t size, size_t* actual_size_out) {
+at::DataPtr RefcountedMapAllocator::makeDataPtr(WithFd /*unused*/, const char *filename, int fd, int flags, size_t size, size_t* actual_size_out) {
   auto* context = new RefcountedMapAllocator(WITH_FD, filename, fd, flags, size);
   if (actual_size_out) *actual_size_out = context->size() - map_alloc_alignment;
   return {context->data(), context, &deleteRefcountedMapAllocator, at::DeviceType::CPU};

aten/src/ATen/MapAllocator.h

@@ -25,7 +25,7 @@ class TORCH_API MapAllocator {
  public:
   MapAllocator(std::string_view filename, int flags, size_t size);
   MapAllocator(
-      WithFd,
+      WithFd /*unused*/,
       std::string_view filename,
       int fd,
       int flags,
@@ -59,14 +59,14 @@ class TORCH_API MapAllocator {
     return flags_;
   }
 
-  static MapAllocator* fromDataPtr(const at::DataPtr&);
+  static MapAllocator* fromDataPtr(const at::DataPtr& /*dptr*/);
   static at::DataPtr makeDataPtr(
       std::string_view filename,
       int flags,
       size_t size,
       size_t* actual_size_out);
   static at::DataPtr makeDataPtr(
-      WithFd,
+      WithFd /*unused*/,
       const char* filename,
       int fd,
       int flags,
@@ -105,13 +105,13 @@ class TORCH_API RefcountedMapAllocator : private RefcountedMapAllocatorArgCheck,
  public:
   RefcountedMapAllocator(const char* filename, int flags, size_t size);
   RefcountedMapAllocator(
-      WithFd,
+      WithFd /*unused*/,
       const char* filename,
       int fd,
       int flags,
       size_t size);
 
-  static RefcountedMapAllocator* fromDataPtr(const at::DataPtr&);
+  static RefcountedMapAllocator* fromDataPtr(const at::DataPtr& /*dptr*/);
   RefcountedMapAllocator(const RefcountedMapAllocator&) = delete;
   RefcountedMapAllocator(RefcountedMapAllocator&&) = delete;
   RefcountedMapAllocator& operator=(const RefcountedMapAllocator&) = delete;
@@ -122,7 +122,7 @@ class TORCH_API RefcountedMapAllocator : private RefcountedMapAllocatorArgCheck,
       size_t size,
       size_t* actual_size_out);
   static at::DataPtr makeDataPtr(
-      WithFd,
+      WithFd /*unused*/,
       const char* filename,
       int fd,
       int flags,

aten/src/ATen/NestedTensorImpl.cpp

@@ -273,7 +273,7 @@ c10::SymInt NestedTensorImpl::sym_numel_custom() const {
   return NestedTensorImpl::numel_custom();
 }
 
-c10::SymBool NestedTensorImpl::sym_is_contiguous_custom(MemoryFormat) const {
+c10::SymBool NestedTensorImpl::sym_is_contiguous_custom(MemoryFormat /*memory_format*/) const {
   return nested_tensor_impl_is_contiguous(this);
 }
 IntArrayRef NestedTensorImpl::sizes_custom() const {

aten/src/ATen/NestedTensorImpl.h

@@ -115,7 +115,8 @@ struct TORCH_API NestedTensorImpl : public c10::TensorImpl {
   // with real implementations
   int64_t numel_custom() const override;
   c10::SymInt sym_numel_custom() const override;
-  c10::SymBool sym_is_contiguous_custom(MemoryFormat) const override;
+  c10::SymBool sym_is_contiguous_custom(
+      MemoryFormat /*memory_format*/) const override;
   int64_t size_custom(int64_t d) const override {
     return this->size(d);
   }

aten/src/ATen/Parallel.h

@@ -14,7 +14,7 @@ inline int64_t divup(int64_t x, int64_t y) {
 TORCH_API void init_num_threads();
 
 // Sets the number of threads to be used in parallel region
-TORCH_API void set_num_threads(int);
+TORCH_API void set_num_threads(int /*nthreads*/);
 
 // Returns the maximum number of threads that may be used in a parallel region
 TORCH_API int get_num_threads();
@@ -37,7 +37,7 @@ inline void lazy_init_num_threads() {
   }
 }
 
-TORCH_API void set_thread_num(int);
+TORCH_API void set_thread_num(int /*id*/);
 
 class TORCH_API ThreadIdGuard {
  public:
@@ -130,7 +130,7 @@ inline scalar_t parallel_reduce(
 TORCH_API std::string get_parallel_info();
 
 // Sets number of threads used for inter-op parallelism
-TORCH_API void set_num_interop_threads(int);
+TORCH_API void set_num_interop_threads(int /*nthreads*/);
 
 // Returns the number of threads used for inter-op parallelism
 TORCH_API size_t get_num_interop_threads();

aten/src/ATen/SparseCsrTensorImpl.cpp

@@ -252,7 +252,7 @@ void SparseCsrTensorImpl::set_stride(int64_t dim, int64_t new_stride) {
 void SparseCsrTensorImpl::set_storage_offset(int64_t storage_offset) {
   TORCH_CHECK(false, "Sparse ", at::sparse_csr::layoutToString(layout_, /*upper=*/true), " tensors do not have set_storage_offset.");
 }
-c10::SymBool SparseCsrTensorImpl::sym_is_contiguous_custom(MemoryFormat) const {
+c10::SymBool SparseCsrTensorImpl::sym_is_contiguous_custom(MemoryFormat /*memory_format*/) const {
   TORCH_CHECK(false, "Sparse ", at::sparse_csr::layoutToString(layout_, /*upper=*/true), " tensors do not have is_contiguous");
 }
 } // namespace at

aten/src/ATen/SparseCsrTensorImpl.h

@@ -32,10 +32,10 @@ struct TORCH_API SparseCsrTensorImpl : public TensorImpl {
 
  public:
   explicit SparseCsrTensorImpl(
-      at::DispatchKeySet,
+      at::DispatchKeySet /*key_set*/,
       at::Device device,
       Layout layout,
-      const caffe2::TypeMeta);
+      const caffe2::TypeMeta /*data_type*/);
 
   void resize_(int64_t nnz, IntArrayRef size);
   void resize_and_clear_(
@@ -86,7 +86,8 @@ struct TORCH_API SparseCsrTensorImpl : public TensorImpl {
  protected:
   IntArrayRef strides_custom() const override;
   SymIntArrayRef sym_strides_custom() const override;
-  SymBool sym_is_contiguous_custom(MemoryFormat) const override;
+  SymBool sym_is_contiguous_custom(
+      MemoryFormat /*memory_format*/) const override;
 
  public:
   void set_size(int64_t dim, int64_t new_size) override;

aten/src/ATen/SparseTensorImpl.h

@@ -46,7 +46,9 @@ struct TORCH_API SparseTensorImpl : public TensorImpl {
 
  public:
   // Public for now...
-  explicit SparseTensorImpl(at::DispatchKeySet, const caffe2::TypeMeta);
+  explicit SparseTensorImpl(
+      at::DispatchKeySet /*key_set*/,
+      const caffe2::TypeMeta /*data_type*/);
 
   void release_resources() override;
 
@@ -384,8 +386,8 @@ struct TORCH_API SparseTensorImpl : public TensorImpl {
 
  private:
   explicit SparseTensorImpl(
-      at::DispatchKeySet,
-      const caffe2::TypeMeta,
+      at::DispatchKeySet /*key_set*/,
+      const caffe2::TypeMeta /*data_type*/,
       at::Tensor indices,
       at::Tensor values);
 

aten/src/ATen/TensorIndexing.h

@@ -112,10 +112,10 @@ TORCH_API std::ostream& operator<<(std::ostream& stream, const Slice& slice);
 // `torch.tensor([1, 2])`) | `torch::tensor({1, 2})`
 struct TORCH_API TensorIndex final {
   // Case 1: `at::indexing::None`
-  TensorIndex(std::nullopt_t) : type_(TensorIndexType::None) {}
+  TensorIndex(std::nullopt_t /*unused*/) : type_(TensorIndexType::None) {}
 
   // Case 2: "..." / `at::indexing::Ellipsis`
-  TensorIndex(at::indexing::EllipsisIndexType)
+  TensorIndex(at::indexing::EllipsisIndexType /*unused*/)
       : type_(TensorIndexType::Ellipsis) {}
   TensorIndex(const char* str) : TensorIndex(at::indexing::Ellipsis) {
     TORCH_CHECK_VALUE(