[Reland][7/N] Fix Wextra-semi warning (pytorch#140342)

Reland of pytorch#140225 to fix a change in FBCODE_CAFFE2

Pull Request resolved: pytorch#140342
Approved by: https://github.com/kit1980

Author: cyyever

Diff for: .clang-format

@@ -101,9 +101,16 @@ SpacesInParentheses: false
 SpacesInSquareBrackets: false
 Standard:        c++17
 StatementMacros:
+  - C10_DEFINE_bool
+  - C10_DEFINE_int
+  - C10_DEFINE_int32
+  - C10_DEFINE_int64
+  - C10_DEFINE_string
   - PyObject_HEAD
   - PyObject_VAR_HEAD
   - PyException_HEAD
+  - DEFINE_BINARY
+
 TabWidth:        8
 UseTab:          Never
 ---

Diff for: aten/src/ATen/cpu/vec/vec512/vec512_bfloat16.h

@@ -1594,8 +1594,8 @@ inline std::tuple<Vectorized<float>, Vectorized<float>> convert_##name##_float(c
 inline Vectorized<type> convert_float_##name(const Vectorized<float>& a, const Vectorized<float>& b) { \
  return cvt_from_fp32<type>(__m512(a), __m512(b)); \
 }
-CONVERT_VECTORIZED_INIT(BFloat16, bfloat16);
-CONVERT_VECTORIZED_INIT(Half, half);
+CONVERT_VECTORIZED_INIT(BFloat16, bfloat16)
+CONVERT_VECTORIZED_INIT(Half, half)
 
 #else //defined(CPU_CAPABILITY_AVX512)
 
@@ -1624,8 +1624,8 @@ inline Vectorized<type> convert_float_##name(const Vectorized<float>& a, const V
   } \
   return Vectorized<type>::loadu(arr2); \
 }
-CONVERT_NON_VECTORIZED_INIT(BFloat16, bfloat16);
-CONVERT_NON_VECTORIZED_INIT(Half, half);
+CONVERT_NON_VECTORIZED_INIT(BFloat16, bfloat16)
+CONVERT_NON_VECTORIZED_INIT(Half, half)
 
 #endif // defined(CPU_CAPABILITY_AVX512)
 
@@ -1663,8 +1663,8 @@ inline void load_fp32_from_##name(const type *data, Vectorized<float>& out1, Vec
   data += Vectorized<float>::size(); \
   load_fp32_from_##name(data, out2); \
 }
-LOAD_FP32_NON_VECTORIZED_INIT(BFloat16, bf16);
-LOAD_FP32_NON_VECTORIZED_INIT(Half, fp16);
+LOAD_FP32_NON_VECTORIZED_INIT(BFloat16, bf16)
+LOAD_FP32_NON_VECTORIZED_INIT(Half, fp16)
 
 #endif
 }}}

Diff for: c10/core/SymInt.cpp

@@ -61,7 +61,6 @@ bool SymInt::has_hint() const {
     }                                                                \
   }
 
-// clang-format off
 DEFINE_BINARY(operator+, std::plus<>(), add, SymInt)
 DEFINE_BINARY(operator-, std::minus<>(), sub, SymInt)
 DEFINE_BINARY(operator*, std::multiplies<>(), mul, SymInt)
@@ -75,7 +74,6 @@ DEFINE_BINARY(sym_gt, std::greater<>(), gt, SymBool)
 DEFINE_BINARY(sym_ge, std::greater_equal<>(), ge, SymBool)
 DEFINE_BINARY(min, std::min, sym_min, SymInt)
 DEFINE_BINARY(max, std::max, sym_max, SymInt)
-// clang-format on
 
 SymInt::operator SymFloat() const {
   if (auto ma = maybe_as_int()) {

Diff for: tools/autograd/templates/VariableType.h

@@ -18,7 +18,7 @@
 
 namespace at {
   struct Quantizer;
-};
+}
 
 namespace torch { namespace autograd {
 
@@ -54,6 +54,6 @@ namespace VariableType {
   const at::Tensor & unpack(const Tensor & t, const char * name, int pos);
   at::Tensor unpack_opt(const Tensor & t, const char * name, int pos);
   std::vector<at::Tensor> unpack(const at::ITensorListRef& tl, const char *name, int pos);
-};
+}
 
 }} // namespace torch::autograd

Diff for: torch/csrc/distributed/autograd/context/context.cpp

@@ -21,7 +21,7 @@ std::unordered_set<rpc::worker_id_t> DistAutogradContext::getKnownWorkerIds()
     const {
   std::lock_guard<std::mutex> guard(lock_);
   return knownWorkerIds_;
-};
+}
 
 void DistAutogradContext::addKnownWorkerId(const rpc::worker_id_t workerId) {
   std::lock_guard<std::mutex> guard(lock_);

Diff for: torch/csrc/distributed/c10d/reducer.hpp

@@ -103,7 +103,7 @@ class TORCH_API Reducer {
   // been applied.
   void set_optimizer_in_backward() {
     optim_in_backward_ = true;
-  };
+  }
 
   // Runs allreduce or installed communication hook given GradBucket instance.
   c10::intrusive_ptr<c10::ivalue::Future> run_comm_hook(

Diff for: torch/csrc/jit/api/function_impl.cpp

@@ -16,7 +16,7 @@
 C10_DEFINE_bool(
     torch_jit_do_not_store_optimized_graph,
     false,
-    "Do not store the optimized graph.");
+    "Do not store the optimized graph.")
 
 namespace torch::jit {
 namespace {
@@ -133,8 +133,8 @@ GraphFunction::SpecializationKey GraphFunction::currentSpecialization() const {
 void preoptimizeGraph(std::shared_ptr<Graph>& graph, bool disable_autocast) {
   Inline(*graph);
 
-  // Peephole Optimize cleans up many "is None" checks and creates constant prop
-  // opportunities
+  // Peephole Optimize cleans up many "is None" checks and creates constant
+  // prop opportunities
   PeepholeOptimize(graph, true);
 
   // AliasDb construction can be slow, so run it just on immutable types

Diff for: torch/csrc/jit/passes/canonicalize_graph_fuser_ops.cpp

@@ -6,7 +6,7 @@
 namespace torch::jit {
 
 struct ChunkOutput {
-  ChunkOutput(Value* v, size_t o) : val(v), offset(o){};
+  ChunkOutput(Value* v, size_t o) : val(v), offset(o) {}
   Value* val;
   size_t offset;
 };

Diff for: torch/csrc/jit/passes/tensorexpr_fuser.cpp

@@ -29,12 +29,12 @@
 C10_DEFINE_bool(
     torch_jit_disable_cat,
     false,
-    "disable aten::cat in TE fusion groups");
+    "disable aten::cat in TE fusion groups")
 
 C10_DEFINE_bool(
     torch_jit_enable_dynamic_shape_fusion,
     false,
-    "enable TE fusion using dynamic shapes");
+    "enable TE fusion using dynamic shapes")
 
 namespace torch::jit {
 
@@ -82,9 +82,8 @@ static const OperatorSet& supported_non_eltwise_set() {
       "aten::_convolution(Tensor input, Tensor weight, Tensor? bias, int[] stride, int[] padding, int[] dilation, bool transposed, int[] output_padding, int groups, bool benchmark, bool deterministic, bool cudnn_enabled, bool allow_tf32) -> Tensor",
       "aten::matmul(Tensor self, Tensor other) -> Tensor",
   };
-  // clang-format on
   return supported_non_eltwise_set;
-};
+}
 
 bool isSupported(Node* node) {
   // For Block codegen we allow limited ops.
@@ -102,7 +101,6 @@ bool isSupported(Node* node) {
       "aten::cat(Tensor[] tensors, int dim=0) -> Tensor",
       "aten::unsqueeze(Tensor(a) self, int dim) -> Tensor(a)",
   };
-  // clang-format on
 
   if (get_tensorexpr_elementwise_set().contains(node) ||
       node->isMemberOf(supported_non_eltwise_set()) ||
@@ -903,7 +901,6 @@ class TensorExprFuser {
     static const OperatorSet pow{
       "aten::pow.Tensor_Scalar(Tensor self, Scalar exponent) -> Tensor",
     };
-    // clang-format on
 
     // Check types of input values.
     for (const Value* v : node->inputs()) {

Diff for: torch/csrc/jit/runtime/autodiff.cpp

@@ -167,7 +167,7 @@ static std::optional<std::vector<Value*>> build_script_grad(
   auto grad_inputs = insertGraph(*graph, *bw_graph, grad);
   grad_inputs = unpackOutputs(grad_inputs);
   return grad_inputs;
-};
+}
 
 namespace {
 class GradientHelper {

Diff for: torch/csrc/jit/runtime/graph_executor.cpp

@@ -56,10 +56,9 @@
 C10_DEFINE_bool(
     torch_jit_execution_plan_reuse_code_graph,
     false,
-    "Directly reuse the preprocessed graph in the CodeImpl to reduce the memory consumption. This is aggressive memory saving, and please be cautious!");
+    "Directly reuse the preprocessed graph in the CodeImpl to reduce the memory consumption. This is aggressive memory saving, and please be cautious!")
 
 namespace torch::jit {
-
 EnableProfilingGuard::EnableProfilingGuard() {
   auto& executor_mode = getExecutorMode();
   old_executor_mode = executor_mode;
@@ -432,8 +431,8 @@ struct DifferentiableGraphOp {
 
     {
       auto inputs = last(stack, num_inputs);
-      // hook up the outputs of df to the gradient functions of the inputs that
-      // require gradients
+      // hook up the outputs of df to the gradient functions of the inputs
+      // that require gradients
       for (auto idx : grad.df_output_vjps) {
         grad_fn->addOutputForIValue(inputs[idx]);
       }
@@ -455,8 +454,8 @@ struct DifferentiableGraphOp {
       // TODO - XXX - if any output is the same tensor multiple times, views
       // have to be setup here. We need to refactor autograd until it is safe
       // for tensors to be constructed without all the viewing infrastructure.
-      // this is currently intentionally not done here so we can get an idea of
-      // our perf before introducing overhead for correctness
+      // this is currently intentionally not done here so we can get an idea
+      // of our perf before introducing overhead for correctness
       for (auto idx : grad.df_input_vjps) {
         grad_fn->addInputIValue(outputs[idx]);
       }
@@ -501,7 +500,8 @@ struct DifferentiableGraphOp {
       detach(stack[i]);
     }
   }
-  // Capture (save) inputs that would be required to subsequently run backwards
+  // Capture (save) inputs that would be required to subsequently run
+  // backwards
   void captureInputs(
       DifferentiableGraphBackward& grad_fn,
       at::ArrayRef<IValue> inputs) const {
@@ -736,8 +736,10 @@ struct GraphExecutorImpl : public GraphExecutorImplBase {
     runOptimization(opt_graph);
 
     // Phase 4. If this graph will be differentiated, we need to slice out the
-    //          symbolically differentiable subgraphs for further optimizations.
-    // Phase 5. Apply non-differentiable optimizations to the graphs we've found
+    //          symbolically differentiable subgraphs for further
+    //          optimizations.
+    // Phase 5. Apply non-differentiable optimizations to the graphs we've
+    // found
     //          (or the whole graph if we know we won't need its derivative).
     if (needsGradient(opt_graph)) {
       auto diff_nodes = CreateAutodiffSubgraphs(
@@ -781,8 +783,8 @@ struct GraphExecutorImpl : public GraphExecutorImplBase {
 
   // NOLINTNEXTLINE(cppcoreguidelines-non-private-member-variables-in-classes)
   ArgumentSpecCreator arg_spec_creator_;
-  // Populated only when optimize is false (and in that case plan_cache will be
-  // unused). The compiled version of graph.
+  // Populated only when optimize is false (and in that case plan_cache will
+  // be unused). The compiled version of graph.
   // NOLINTNEXTLINE(cppcoreguidelines-non-private-member-variables-in-classes)
   ExecutionPlan fallback;
 

Diff for: torch/csrc/jit/runtime/interpreter.cpp

@@ -49,12 +49,12 @@ using torch::distributed::autograd::DistAutogradContainer;
 C10_DEFINE_bool(
     torch_jit_enable_rethrow_caught_exception,
     false,
-    "enable rethrowing caught exception");
+    "enable rethrowing caught exception")
 
 C10_DEFINE_bool(
     torch_jit_enable_expanded_stacks,
     false,
-    "When true we will attemps to pre-expand node stacks and cache expanded stacks.");
+    "When true we will attemps to pre-expand node stacks and cache expanded stacks.")
 
 namespace torch::jit {
 

Diff for: torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp

@@ -41,44 +41,44 @@
 C10_DEFINE_bool(
     torch_jit_enable_new_executor,
     true,
-    "If this flag is set to false TorchScript will be using the legacy/original executor");
+    "If this flag is set to false TorchScript will be using the legacy/original executor")
 
 C10_DEFINE_bool(
     torch_jit_disable_warning_prints,
     false,
-    "Disables warning.warn prints in TorchScript graph");
+    "Disables warning.warn prints in TorchScript graph")
 
 C10_DEFINE_bool(
     torch_jit_static_then_dynamic,
     false,
-    "fuse on two static compilations then 10 dynamic");
+    "fuse on two static compilations then 10 dynamic")
 
 C10_DEFINE_bool(
     torch_jit_always_dynamic,
     false,
-    "fuse on 12 dynamic compilations");
+    "fuse on 12 dynamic compilations")
 
 C10_DEFINE_bool(
     torch_jit_release_profiling_graph_after_optimization,
     false,
-    "After getOptimizedPlanFor release the optimization record for reduction of memory in inference. This is aggressive memory saving, and please be cautious!");
+    "After getOptimizedPlanFor release the optimization record for reduction of memory in inference. This is aggressive memory saving, and please be cautious!")
 
 C10_DEFINE_int32(
     torch_jit_release_profiling_graph_delay_in_seconds,
     60,
-    "How long to wait before releasing the profiling graph after optimizaiton is done. Only used if torch_jit_release_profiling_graph_after_optimization is set to true.");
+    "How long to wait before releasing the profiling graph after optimizaiton is done. Only used if torch_jit_release_profiling_graph_after_optimization is set to true.")
 
 constexpr size_t kDefaultNumProfiledRuns = 1;
 constexpr size_t kDefaultBailoutDepth = 20;
 
 C10_DEFINE_int64(
     torch_jit_num_profiled_runs,
     kDefaultNumProfiledRuns,
-    "Number of profiling runs");
+    "Number of profiling runs")
 C10_DEFINE_int64(
     torch_jit_bailout_depth,
     kDefaultBailoutDepth,
-    "Number of re-specializations");
+    "Number of re-specializations")
 
 namespace torch::jit {