Fix clang-tidy warnings in jit code (pytorch#138974)

Fixes #ISSUE_NUMBER

Pull Request resolved: pytorch#138974
Approved by: https://github.com/ezyang

Author: cyyever

torch/csrc/jit/backends/backend_exception.h

@@ -1,6 +1,8 @@
 #pragma once
 #include <c10/util/Exception.h>
 
+#include <utility>
+
 namespace c10 {
 class TORCH_API BackendRuntimeException : public c10::Error {
  public:
@@ -9,7 +11,7 @@ class TORCH_API BackendRuntimeException : public c10::Error {
       SourceLocation loc,
       std::string msg,
       int64_t debug_handle)
-      : c10::Error(loc, msg) {
+      : c10::Error(loc, std::move(msg)) {
     debug_handles.push_back(debug_handle);
   }
   // If rethrowing, can push another debug_handle

torch/csrc/jit/codegen/onednn/LlgaTensorImpl.h

@@ -5,6 +5,7 @@
 
 #include <oneapi/dnnl/dnnl_graph.hpp>
 #include <torch/csrc/jit/ir/ir.h>
+#include <utility>
 
 namespace torch::jit::fuser::onednn {
 
@@ -42,8 +43,8 @@ struct LlgaTensorDesc {
       desc::data_type dtype,
       desc::property_type property_type)
       : tid_(tid),
-        sizes_(sizes),
-        strides_(strides),
+        sizes_(std::move(sizes)),
+        strides_(std::move(strides)),
         dtype_(dtype),
         property_type_(property_type),
         layout_type_(desc::layout_type::strided),
@@ -221,7 +222,7 @@ struct LlgaTensorDesc {
 
  private:
   bool is_dimensionality_unknown() const {
-    return sizes_.size() == 0;
+    return sizes_.empty();
   }
 
   size_t tid_;
@@ -236,7 +237,7 @@ struct LlgaTensorDesc {
   // compute_inplace would be true, and input_tensor_index would be the index of
   // the corresponding input tensor in inputSpecs_ of the LlgaKernel object.
   bool compute_inplace_ = false;
-  size_t input_tensor_index_;
+  size_t input_tensor_index_{};
 };
 
 // Initially, oneDNN Graph also used to have blocked layout for tensors between

torch/csrc/jit/codegen/onednn/kernel.cpp

@@ -126,7 +126,7 @@ std::tuple<RunArgs, RunArgs> LlgaKernel::prepareRunArgs(
   auto numInputs = runArgsIdx_.size();
   for (const auto i : c10::irange(numInputs)) {
     auto spec = inputSpecs_[i];
-    auto input = inputs[runArgsIdx_[i]];
+    const auto& input = inputs[runArgsIdx_[i]];
     runInputs.push_back(
         {spec.logical_tensor(), Engine::getEngine(), input.data_ptr()});
   }

torch/csrc/jit/mobile/import.cpp

@@ -339,8 +339,7 @@ void BytecodeDeserializer::parseMethods(
     auto element = std::move(vals[i]);
     auto m_tuple = std::move(element.toTupleRef()).elements();
     const std::string& function_name = m_tuple[0].toStringRef();
-    auto codeTableElements =
-        std::move(std::move(m_tuple[1]).toTupleRef()).elements();
+    auto codeTableElements = std::move(m_tuple[1].toTupleRef()).elements();
     IValue* schemaTable = // older files do not store function schema
         (bytecode_version_ > 0x4L ||
          (bytecode_version_ == 0x4L && m_tuple.size() >= 3))

torch/csrc/jit/mobile/nnc/context.cpp

@@ -196,7 +196,7 @@ c10::IValue Function::serialize() const {
 }
 
 void Function::init_execution_state() const {
-  if (execution_state_.get() != nullptr) {
+  if (execution_state_ != nullptr) {
     return;
   }
 

torch/csrc/jit/passes/frozen_ops_to_mkldnn.cpp

@@ -85,7 +85,7 @@ void merge_sets(
 }
 
 // no uses of tensors in container types
-void assertNonTensorTypeDoesNotContainTensors(TypePtr type) {
+void assertNonTensorTypeDoesNotContainTensors(const TypePtr& type) {
   if (type->cast<TensorType>()) {
     return;
   }
@@ -94,7 +94,7 @@ void assertNonTensorTypeDoesNotContainTensors(TypePtr type) {
   }
 }
 
-void InplaceMKLDNNSubgraph(std::shared_ptr<Graph> graph) {
+void InplaceMKLDNNSubgraph(const std::shared_ptr<Graph>& graph) {
   // This function first calculates aliasing sets,
   // then calculates the last node each aliasing set is alive for.
   // Then we go through each node, if it's a node which has an equivalent
@@ -234,7 +234,7 @@ void InplaceMKLDNNSubgraph(std::shared_ptr<Graph> graph) {
 // innermost dimension is padded with 0s. The precondition, `aten_op(0) == 0`
 // allows us to avoid any special casing of padded elements.
 Operation createUnaryOp(
-    std::function<void(at::Tensor output, at::Tensor input)> aten_op,
+    const std::function<void(at::Tensor output, at::Tensor input)>& aten_op,
     bool inplace = false) {
   return [aten_op, inplace](Stack& stack) {
     auto a = pop(stack).toTensor();
@@ -395,7 +395,7 @@ static std::function<void(at::Tensor output, at::Tensor input)> hardtanh_helper(
     const Node* n) {
   auto min_val = n->f(attr::min_val);
   auto max_val = n->f(attr::max_val);
-  return [min_val, max_val](at::Tensor output, at::Tensor input) {
+  return [min_val, max_val](at::Tensor output, const at::Tensor& input) {
     at::cpu::hardtanh_out(output, input, min_val, max_val);
   };
 }
@@ -404,7 +404,7 @@ static std::function<void(at::Tensor output, at::Tensor input)> clamp_helper(
     const Node* n) {
   auto min_val = n->f(attr::min_val);
   auto max_val = n->f(attr::max_val);
-  return [min_val, max_val](at::Tensor output, at::Tensor input) {
+  return [min_val, max_val](at::Tensor output, const at::Tensor& input) {
     at::cpu::clamp_out(output, input, min_val, max_val);
   };
 }
@@ -415,15 +415,15 @@ const RegisterOperators MKLDNNHardSwishOpReg({
     torch::jit::Operator(
         "prim::MKLDNNHardSwish_(Tensor(a!) self) -> Tensor(a!)",
         createUnaryOp(
-            [](at::Tensor output, at::Tensor input) {
+            [](at::Tensor output, const at::Tensor& input) {
               at::cpu::hardswish_out(output, input);
             },
             true),
         AliasAnalysisKind::FROM_SCHEMA),
     torch::jit::Operator(
         "prim::MKLDNNHardSigmoid_(Tensor(a!) self) -> Tensor(a!)",
         createUnaryOp(
-            [](at::Tensor output, at::Tensor input) {
+            [](at::Tensor output, const at::Tensor& input) {
               at::cpu::hardsigmoid_out(output, input);
             },
             true),
@@ -443,15 +443,15 @@ const RegisterOperators MKLDNNHardSwishOpReg({
     torch::jit::Operator(
         "prim::MKLDNNHardSwish(Tensor a) -> Tensor",
         createUnaryOp(
-            [](at::Tensor output, at::Tensor input) {
+            [](at::Tensor output, const at::Tensor& input) {
               at::cpu::hardswish_out(output, input);
             },
             false),
         AliasAnalysisKind::FROM_SCHEMA),
     torch::jit::Operator(
         "prim::MKLDNNHardSigmoid(Tensor a) -> Tensor",
         createUnaryOp(
-            [](at::Tensor output, at::Tensor input) {
+            [](at::Tensor output, const at::Tensor& input) {
               at::cpu::hardsigmoid_out(output, input);
             },
             false),

torch/csrc/jit/passes/onnx/naming.cpp

@@ -7,7 +7,7 @@ namespace torch::jit::onnx {
 
 namespace ONNXScopeName {
 
-using NameFunc = std::string (*)(torch::jit::ScopePtr scope);
+using NameFunc = std::string (*)(const torch::jit::ScopePtr& scope);
 
 const std::string name_separator = "::";
 
@@ -48,7 +48,7 @@ std::string createFullScopeName(
   return std::string(class_name).append(name_separator).append(variable_name);
 }
 
-std::string variableName(torch::jit::ScopePtr scope) {
+std::string variableName(const torch::jit::ScopePtr& scope) {
   return parseNameFromScope(scope).second;
 }
 
@@ -58,7 +58,7 @@ std::string variableNameFromRoot(
   return nameFromRoot(scope, layer_separator, &variableName);
 }
 
-std::string className(torch::jit::ScopePtr scope) {
+std::string className(const torch::jit::ScopePtr& scope) {
   return parseNameFromScope(scope).first;
 }
 

torch/csrc/jit/passes/onnx/naming.h

@@ -9,11 +9,11 @@ namespace ONNXScopeName {
 std::string createFullScopeName(
     const std::string& class_name,
     const std::string& variable_name);
-std::string variableName(torch::jit::ScopePtr scope);
+std::string variableName(const torch::jit::ScopePtr& scope);
 std::string variableNameFromRoot(
     const torch::jit::ScopePtr& scope,
     const std::string& layer_separator);
-std::string className(torch::jit::ScopePtr scope);
+std::string className(const torch::jit::ScopePtr& scope);
 std::string classNameFromRoot(
     const torch::jit::ScopePtr& scope,
     const std::string& layer_separator);

torch/csrc/jit/passes/onnx/remove_inplace_ops_for_onnx.cpp

@@ -6,7 +6,6 @@
 #include <torch/csrc/jit/frontend/error_report.h>
 #include <torch/csrc/jit/jit_log.h>
 #include <torch/csrc/jit/passes/dead_code_elimination.h>
-#include <torch/csrc/jit/passes/onnx/helper.h>
 #include <torch/csrc/jit/passes/onnx/pattern_conversion/pattern_encapsulation.h>
 
 #include <c10/util/irange.h>

torch/csrc/jit/passes/quantization/quantization_patterns.h

@@ -25,7 +25,9 @@ std::string getExtraArgList(std::vector<std::string> extra_args) {
       extra_args.begin(),
       extra_args.end(),
       std::string(),
-      [](std::string acc, const std::string& arg) { return acc + ", " + arg; });
+      [](const std::string& acc, const std::string& arg) {
+        return acc + ", " + arg;
+      });
 }
 
 // Get the pattern we want to replace the match with

torch/csrc/jit/python/init.cpp

@@ -1732,7 +1732,7 @@ void initJITBindings(PyObject* module) {
           bool allow_numbers_as_tensors = opAllowsNumbersAsTensors(symbol);
           ToIValueAllowNumbersAsTensors g(allow_numbers_as_tensors);
           const auto overloads = getAllSortedOperatorsFor(symbol);
-          auto opWithStack = getOpWithStack(overloads, std::move(args), kwargs);
+          auto opWithStack = getOpWithStack(overloads, args, kwargs);
           std::shared_ptr<Operator> overload = std::get<0>(opWithStack);
           auto result = overload->schema().overload_name();
           if (result.empty()) {

torch/csrc/jit/python/script_init.cpp

@@ -48,7 +48,6 @@
 #include <torch/csrc/jit/runtime/instruction.h>
 #include <torch/csrc/jit/runtime/interpreter.h>
 #include <torch/csrc/jit/runtime/logging.h>
-#include <torch/csrc/jit/serialization/export_bytecode.h>
 #include <torch/csrc/jit/serialization/import_source.h>
 #include <torch/csrc/jit/serialization/pickle.h>
 #include <torch/csrc/jit/serialization/python_print.h>

torch/csrc/jit/runtime/decomposition_registry.cpp

@@ -10,9 +10,7 @@
 #include <c10/util/Exception.h>
 #include <torch/csrc/autograd/jit_decomp_interface.h>
 #include <torch/csrc/jit/ir/ir.h>
-#include <torch/csrc/jit/passes/constant_propagation.h>
 #include <torch/csrc/jit/passes/inliner.h>
-#include <torch/csrc/jit/passes/peephole.h>
 #include <torch/csrc/jit/runtime/graph_executor.h>
 #include <memory>
 #include <unordered_map>
@@ -79,8 +77,7 @@ static void DecomposeOp(Node* n) {
     return;
   }
   WithInsertPoint guard(n);
-  auto outputs =
-      insertGraph(*n->owningGraph(), *decomposition->get(), n->inputs());
+  auto outputs = insertGraph(*n->owningGraph(), **decomposition, n->inputs());
   TORCH_INTERNAL_ASSERT(outputs.size() == n->outputs().size());
   for (size_t i : c10::irange(outputs.size())) {
     n->outputs().at(i)->replaceAllUsesWith(outputs[i]);

torch/csrc/jit/runtime/interpreter/code_impl.h

@@ -2,6 +2,7 @@
 
 #include <memory>
 #include <unordered_map>
+#include <utility>
 #include <vector>
 
 #include <c10/util/irange.h>
@@ -945,7 +946,11 @@ struct MobileCodeImpl : CodeImpl {
       bool support_default_args_before_out,
       bool emit_promoted_ops,
       size_t remaining_bailout_depth)
-      : CodeImpl(graph, function_name, remaining_bailout_depth, false),
+      : CodeImpl(
+            graph,
+            std::move(function_name),
+            remaining_bailout_depth,
+            false),
         emit_default_input_instructions_(emit_default_input_instructions),
         support_default_args_before_out_(support_default_args_before_out),
         emit_promoted_ops_(emit_promoted_ops) {

torch/csrc/jit/runtime/interpreter/preprocess_graph.cpp

@@ -209,6 +209,6 @@ PreprocessGraph::PreprocessGraph(Graph& g) : graph(g.copy()) {
   dropUnused(graph->block());
   // fill in move_flags by scanning blocks;
   insertLastUses(*graph);
-  can_emit_inline = std::move(CanEmitInline(*graph.get()).can_emit_inline_);
+  can_emit_inline = std::move(CanEmitInline(*graph).can_emit_inline_);
 }
 } // namespace torch::jit::interpreter

torch/csrc/jit/runtime/script_profile.h

@@ -46,8 +46,8 @@ class TORCH_API SourceStats : public CustomClassHolder {
  public:
   using LineMap = c10::Dict<int64_t, c10::intrusive_ptr<InstructionStats>>;
 
-  SourceStats(SourceRef source, LineMap lineMap)
-      : source_(std::move(source)), lineMap_(std::move(lineMap)) {}
+  SourceStats(SourceRef source, const LineMap& lineMap)
+      : source_(std::move(source)), lineMap_(lineMap) {}
 
   const SourceRef& getSourceRef() const {
     return source_;

torch/csrc/jit/tensorexpr/ir.h

@@ -859,12 +859,9 @@ class TORCH_API Intrinsics : public ExprNode<Intrinsics> {
     }
   }
 
-  Intrinsics(
-      IntrinsicsOp op_type,
-      Dtype dtype,
-      const std::vector<ExprPtr>& params)
+  Intrinsics(IntrinsicsOp op_type, Dtype dtype, std::vector<ExprPtr> params)
       : ExprNodeBase(IntrinsicsDtype(op_type, dtype)),
-        params_(params),
+        params_(std::move(params)),
         op_type_(op_type) {
     if (OpArgCount(op_type) != nparams()) {
       throw malformed_input("bad arg count in Intrinsics");

torch/csrc/jit/tensorexpr/loopnest.cpp

@@ -25,11 +25,6 @@
 #include <torch/csrc/jit/tensorexpr/ir_verifier.h>
 #include <torch/csrc/jit/tensorexpr/tensor.h>
 
-#include <stdexcept>
-#include <unordered_map>
-#include <unordered_set>
-#include <vector>
-
 namespace torch::jit::tensorexpr {
 
 LoopNest::LoopNest(const LoopNest& other)

torch/csrc/jit/testing/hooks_for_testing.cpp

@@ -7,14 +7,14 @@ namespace torch::jit {
 static ModuleHook emit_module_callback;
 void didFinishEmitModule(Module module) {
   if (emit_module_callback) {
-    emit_module_callback(module);
+    emit_module_callback(std::move(module));
   }
 }
 
 static FunctionHook emit_function_callback;
 void didFinishEmitFunction(StrongFunctionPtr fn) {
   if (emit_function_callback) {
-    emit_function_callback(fn);
+    emit_function_callback(std::move(fn));
   }
 }