Apply some std::move and param value fixups to aten (pytorch#92901)

I noticed a few perf issues in the latest ATen and decided to fixup a few other miscellaneous ones I noticed recently.
Pull Request resolved: pytorch#92901
Approved by: https://github.com/ezyang

Author: Skylion007

aten/src/ATen/LegacyBatchedTensorImpl.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <bitset>
+#include <utility>
 
 #include <ATen/ArrayRef.h>
 #include <ATen/SmallVector.h>
@@ -120,7 +121,7 @@ inline BatchedTensorImpl* maybeGetBatchedImpl(Tensor tensor) {
   if (!isBatchedTensor(tensor)) {
     return nullptr;
   }
-  return unsafeGetBatchedImpl(tensor);
+  return unsafeGetBatchedImpl(std::move(tensor));
 }
 
 // Returns a bitset. If bit i is set, then that means dim i is a batchdim.

aten/src/ATen/TensorIndexing.h

@@ -20,6 +20,8 @@
 
 #include <ATen/core/List.h>
 
+#include <utility>
+
 namespace at {
 namespace indexing {
 
@@ -230,7 +232,7 @@ static inline Tensor applySlice(
       return self;
     }
   }
-  return self.slice_symint(dim, start, stop, step);
+  return self.slice_symint(dim, start, stop, std::move(step));
 }
 
 static inline Tensor applySelect(

aten/src/ATen/ThreadLocalPythonObjects.cpp

@@ -2,27 +2,29 @@
 #include <ATen/ThreadLocalPythonObjects.h>
 #include <c10/util/Exception.h>
 
+#include <utility>
+
 namespace at {
 namespace impl {
 
 static thread_local ThreadLocalPythonObjects py_objects;
 
 
-void ThreadLocalPythonObjects::set(std::string key, std::shared_ptr<SafePyObject> value) {
-  py_objects.obj_dict_[key] = value;
+void ThreadLocalPythonObjects::set(const std::string& key, std::shared_ptr<SafePyObject> value) {
+  py_objects.obj_dict_[key] = std::move(value);
 }
 
-const std::shared_ptr<SafePyObject>& ThreadLocalPythonObjects::get(std::string key) {
+const std::shared_ptr<SafePyObject>& ThreadLocalPythonObjects::get(const std::string& key) {
   TORCH_CHECK(py_objects.obj_dict_.count(key));
   return py_objects.obj_dict_[key];
 }
 
-bool ThreadLocalPythonObjects::contains(std::string key) {
+bool ThreadLocalPythonObjects::contains(const std::string& key) {
   return py_objects.obj_dict_.count(key);
 }
 
-void ThreadLocalPythonObjects::set_state(const ThreadLocalPythonObjects& state) {
-  py_objects = state;
+void ThreadLocalPythonObjects::set_state(ThreadLocalPythonObjects state) {
+  py_objects = std::move(state);
 }
 
 const ThreadLocalPythonObjects& ThreadLocalPythonObjects::get_state() {

aten/src/ATen/ThreadLocalPythonObjects.h

@@ -8,12 +8,12 @@ namespace at {
 namespace impl {
 
 struct TORCH_API ThreadLocalPythonObjects {
-  static void set(std::string key, std::shared_ptr<SafePyObject> value);
-  static const std::shared_ptr<SafePyObject>& get(std::string key);
-  static bool contains(std::string key);
+  static void set(const std::string& key, std::shared_ptr<SafePyObject> value);
+  static const std::shared_ptr<SafePyObject>& get(const std::string& key);
+  static bool contains(const std::string& key);
 
   static const ThreadLocalPythonObjects& get_state();
-  static void set_state(const ThreadLocalPythonObjects& state);
+  static void set_state(ThreadLocalPythonObjects state);
 
  private:
   std::unordered_map<std::string, std::shared_ptr<c10::SafePyObject>> obj_dict_;

aten/src/ATen/core/ivalue_inl.h

@@ -944,7 +944,7 @@ struct C10_EXPORT ivalue::Future final : c10::intrusive_ptr_target {
           "Skipping setting following error on the Future since "
           "it is already marked completed (this is not necessarily "
           "an error):\n",
-          tryRetrieveErrorMessageInternal(eptr));
+          tryRetrieveErrorMessageInternal(std::move(eptr)));
       if (eptr_) {
         msg += c10::str(
             ", \nOriginal exception:\n",
@@ -1199,7 +1199,7 @@ struct C10_EXPORT ivalue::Future final : c10::intrusive_ptr_target {
   // Tries to retrieve the error message from std::exception_ptr.
   std::string tryRetrieveErrorMessageInternal(std::exception_ptr eptr) const {
     try {
-      std::rethrow_exception(eptr);
+      std::rethrow_exception(std::move(eptr));
     } catch (const std::exception& e) {
       return e.what();
     } catch (...) {

aten/src/ATen/core/jit_type.h

@@ -16,6 +16,7 @@
 #include <ostream>
 #include <sstream>
 #include <type_traits>
+#include <utility>
 
 namespace torch {
 namespace jit {
@@ -239,7 +240,7 @@ struct TORCH_API OptionalType : public UnionType {
 
   std::string annotation_str_impl(TypePrinter printer = nullptr) const override {
     std::stringstream ss;
-    ss << "Optional[" << getElementType()->annotation_str(printer) << "]";
+    ss << "Optional[" << getElementType()->annotation_str(std::move(printer)) << "]";
     return ss.str();
   }
 };
@@ -906,7 +907,7 @@ struct TORCH_API ListType
 
   std::string annotation_str_impl(TypePrinter printer = nullptr) const override {
     std::stringstream ss;
-    ss << "List[" << getElementType()->annotation_str(printer) << "]";
+    ss << "List[" << getElementType()->annotation_str(std::move(printer)) << "]";
     return ss.str();
   }
 };
@@ -1001,7 +1002,7 @@ struct TORCH_API DictType : public SharedType {
   std::string annotation_str_impl(TypePrinter printer = nullptr) const override {
     std::stringstream ss;
     ss << "Dict[" << getKeyType()->annotation_str(printer) << ", "
-       << getValueType()->annotation_str(printer) << "]";
+       << getValueType()->annotation_str(std::move(printer)) << "]";
     return ss.str();
   }
 
@@ -1046,7 +1047,7 @@ struct TORCH_API FutureType
 
   std::string annotation_str_impl(TypePrinter printer = nullptr) const override {
     std::stringstream ss;
-    ss << "Future[" << getElementType()->annotation_str(printer) << "]";
+    ss << "Future[" << getElementType()->annotation_str(std::move(printer)) << "]";
     return ss.str();
   }
 };
@@ -1078,7 +1079,7 @@ struct TORCH_API RRefType
 
   std::string annotation_str_impl(TypePrinter printer = nullptr) const override {
     std::stringstream ss;
-    ss << "RRef[" << getElementType()->annotation_str(printer) << "]";
+    ss << "RRef[" << getElementType()->annotation_str(std::move(printer)) << "]";
     return ss.str();
   }
 };

aten/src/ATen/core/jit_type_base.h

@@ -3,6 +3,7 @@
 #include <functional>
 #include <memory>
 #include <string>
+#include <utility>
 
 #include <ATen/core/qualified_name.h>
 #include <ATen/core/type_ptr.h>
@@ -451,7 +452,7 @@ struct TORCH_API Type {
         return *renamed;
       }
     }
-    return annotation_str_impl(printer);
+    return annotation_str_impl(std::move(printer));
   }
   std::string annotation_str() const {
     // Overload instead of define a default value for `printer` to help

c10/util/ThreadLocalDebugInfo.cpp

@@ -27,8 +27,8 @@ std::shared_ptr<ThreadLocalDebugInfo> ThreadLocalDebugInfo::current() {
 
 /* static */
 void ThreadLocalDebugInfo::_forceCurrentDebugInfo(
-    const std::shared_ptr<ThreadLocalDebugInfo>& info) {
-  debug_info = info;
+    std::shared_ptr<ThreadLocalDebugInfo> info) {
+  debug_info = std::move(info);
 }
 
 /* static */
@@ -39,7 +39,7 @@ void ThreadLocalDebugInfo::_push(
   debug_info = std::make_shared<ThreadLocalDebugInfo>();
   debug_info->parent_info_ = prev_info;
   debug_info->kind_ = kind;
-  debug_info->info_ = info;
+  debug_info->info_ = std::move(info);
 }
 
 /* static */
@@ -86,8 +86,8 @@ DebugInfoGuard::DebugInfoGuard(std::shared_ptr<ThreadLocalDebugInfo> info) {
   if (!info) {
     return;
   }
-  prev_info_ = debug_info;
-  debug_info = info;
+  prev_info_ = std::move(debug_info);
+  debug_info = std::move(info);
   active_ = true;
 }
 

c10/util/ThreadLocalDebugInfo.h

@@ -41,7 +41,7 @@ class C10_API ThreadLocalDebugInfo {
 
   // Internal, use DebugInfoGuard/ThreadLocalStateGuard
   static void _forceCurrentDebugInfo(
-      const std::shared_ptr<ThreadLocalDebugInfo>& info);
+      std::shared_ptr<ThreadLocalDebugInfo> info);
 
   // Push debug info struct of a given kind
   static void _push(DebugInfoKind kind, std::shared_ptr<DebugInfoBase> info);