[2/N] Cleanup header inclusions in torch_cpu by iwyu (pytorch#109964)

Further cleaning up of torch_cpu header inclusions. Pull Request resolved: pytorch#109964 Approved by: https://github.com/ezyang, https://github.com/Skylion007
Xxx14124 · Nov 19, 2023 · 226384b · 226384b
1 parent 0bd4d1f
commit 226384b
Show file tree

Hide file tree

Showing 49 changed files with 27 additions and 105 deletions.
diff --git a/aten/src/ATen/detail/CPUGuardImpl.cpp b/aten/src/ATen/detail/CPUGuardImpl.cpp
@@ -1,5 +1,4 @@
 #include <c10/core/impl/DeviceGuardImplInterface.h>
-#include <c10/macros/Macros.h>
 
 namespace at {
 namespace detail {

diff --git a/aten/src/ATen/detail/CUDAHooksInterface.cpp b/aten/src/ATen/detail/CUDAHooksInterface.cpp
@@ -1,11 +1,8 @@
 #include <ATen/detail/CUDAHooksInterface.h>
 
-#include <c10/util/Exception.h>
 #include <c10/util/CallOnce.h>
 
-#include <cstddef>
 #include <memory>
-#include <mutex>
 
 namespace at {
 namespace detail {

diff --git a/aten/src/ATen/detail/CUDAHooksInterface.h b/aten/src/ATen/detail/CUDAHooksInterface.h
@@ -2,16 +2,10 @@
 
 #include <c10/core/Allocator.h>
 #include <c10/util/Exception.h>
-#include <c10/util/Optional.h>
 #include <c10/util/Registry.h>
 
-#include <cstddef>
-#include <functional>
-#include <memory>
-
-// Forward-declares at::Context, at::Generator and at::cuda::NVRTC
+// Forward-declares at::Generator and at::cuda::NVRTC
 namespace at {
-class Context;
 struct Generator;
 namespace cuda {
 struct NVRTC;

diff --git a/aten/src/ATen/detail/HIPHooksInterface.cpp b/aten/src/ATen/detail/HIPHooksInterface.cpp
@@ -1,11 +1,10 @@
 #include <ATen/detail/HIPHooksInterface.h>
 
-#include <c10/util/Exception.h>
 #include <c10/util/CallOnce.h>
+#include <c10/util/Registry.h>
 
 #include <cstddef>
 #include <memory>
-#include <mutex>
 
 namespace at {
 namespace detail {

diff --git a/aten/src/ATen/detail/HIPHooksInterface.h b/aten/src/ATen/detail/HIPHooksInterface.h
@@ -1,13 +1,12 @@
 #pragma once
 
 #include <c10/core/Allocator.h>
-#include <ATen/core/Generator.h>
+#include <c10/core/GeneratorImpl.h>
 #include <c10/util/Exception.h>
 
 #include <c10/util/Registry.h>
 
 #include <cstddef>
-#include <functional>
 #include <memory>
 
 namespace at {

diff --git a/aten/src/ATen/detail/MPSHooksInterface.cpp b/aten/src/ATen/detail/MPSHooksInterface.cpp
@@ -1,7 +1,6 @@
 //  Copyright © 2022 Apple Inc.
 
 #include <ATen/detail/MPSHooksInterface.h>
-#include <c10/util/Exception.h>
 #include <c10/util/CallOnce.h>
 
 namespace at {

diff --git a/aten/src/ATen/detail/MPSHooksInterface.h b/aten/src/ATen/detail/MPSHooksInterface.h
@@ -8,11 +8,6 @@
 #include <c10/util/Registry.h>
 
 #include <cstddef>
-#include <functional>
-
-namespace at {
-class Context;
-}
 
 namespace at {
 

diff --git a/aten/src/ATen/detail/MTIAHooksInterface.cpp b/aten/src/ATen/detail/MTIAHooksInterface.cpp
@@ -1,18 +1,16 @@
 #include <ATen/detail/MTIAHooksInterface.h>
 
 #include <c10/util/CallOnce.h>
-#include <c10/util/Exception.h>
 
 #include <cstddef>
 #include <memory>
-#include <mutex>
 
 namespace at {
 namespace detail {
 
-static MTIAHooksInterface* MTIA_hooks = nullptr;
 
 const MTIAHooksInterface &getMTIAHooks() {
+  static MTIAHooksInterface* MTIA_hooks = nullptr;
   static c10::once_flag once;
   c10::call_once(once, [] {
     MTIA_hooks =

diff --git a/aten/src/ATen/detail/MTIAHooksInterface.h b/aten/src/ATen/detail/MTIAHooksInterface.h
@@ -1,23 +1,15 @@
 #pragma once
 
-#include <c10/core/Device.h>
 #include <c10/util/Exception.h>
 
 #include <c10/util/Registry.h>
 
-#include <cstddef>
-#include <functional>
-#include <memory>
+#include <string>
 
 namespace at {
 class Context;
 }
 
-// We use forward declaration here instead of #include <ATen/dlpack.h> to avoid
-// leaking DLPack implementation detail to every project that includes `ATen/Context.h`, which in turn
-// would lead to a conflict when linked with another project using DLPack (for example TVM)
-struct DLDevice_;
-
 namespace at {
 
 constexpr const char* MTIA_HELP =

diff --git a/aten/src/ATen/detail/ORTHooksInterface.cpp b/aten/src/ATen/detail/ORTHooksInterface.cpp
@@ -1,11 +1,10 @@
 #include <ATen/detail/ORTHooksInterface.h>
 
-#include <c10/util/Exception.h>
 #include <c10/util/CallOnce.h>
+#include <c10/util/Registry.h>
 
 #include <cstddef>
 #include <memory>
-#include <mutex>
 
 namespace at {
 namespace detail {

diff --git a/aten/src/ATen/detail/XPUHooksInterface.cpp b/aten/src/ATen/detail/XPUHooksInterface.cpp
@@ -1,18 +1,15 @@
 #include <ATen/detail/XPUHooksInterface.h>
 
 #include <c10/util/CallOnce.h>
-#include <c10/util/Exception.h>
 
-#include <cstddef>
 #include <memory>
-#include <mutex>
 
 namespace at {
 namespace detail {
 
-static XPUHooksInterface *xpu_hooks = nullptr;
 
-const XPUHooksInterface &getXPUHooks() {
+const XPUHooksInterface& getXPUHooks() {
+  static XPUHooksInterface* xpu_hooks = nullptr;
   static c10::once_flag once;
   c10::call_once(once, [] {
     xpu_hooks =

diff --git a/aten/src/ATen/detail/XPUHooksInterface.h b/aten/src/ATen/detail/XPUHooksInterface.h
@@ -9,10 +9,6 @@
 #include <functional>
 #include <memory>
 
-namespace at {
-class Context;
-}
-
 // We use forward declaration here instead of #include <ATen/dlpack.h> to avoid
 // leaking DLPack implementation detail to every project that includes `ATen/Context.h`, which in turn
 // would lead to a conflict when linked with another project using DLPack (for example TVM)

diff --git a/aten/src/ATen/native/DilatedMaxPool3d.cpp b/aten/src/ATen/native/DilatedMaxPool3d.cpp
@@ -3,7 +3,6 @@
 #include <ATen/Dispatch.h>
 #include <ATen/NamedTensorUtils.h>
 #include <ATen/native/Pool.h>
-#include <c10/util/irange.h>
 
 #ifndef AT_PER_OPERATOR_HEADERS
 #include <ATen/Functions.h>

diff --git a/aten/src/ATen/native/PointwiseOps.cpp b/aten/src/ATen/native/PointwiseOps.cpp
@@ -3,7 +3,6 @@
 #include <ATen/native/PointwiseOps.h>
 
 #include <ATen/core/Tensor.h>
-#include <ATen/TensorIterator.h>
 #include <ATen/TensorMeta.h>
 
 #ifndef AT_PER_OPERATOR_HEADERS

diff --git a/aten/src/ATen/native/xnnpack/RegisterOpContextClass.cpp b/aten/src/ATen/native/xnnpack/RegisterOpContextClass.cpp
@@ -4,7 +4,6 @@
 #include <ATen/native/xnnpack/Convolution.h>
 #include <ATen/native/xnnpack/Linear.h>
 #include <ATen/native/xnnpack/OpContext.h>
-#include <ATen/Tensor.h>
 #include <torch/custom_class.h>
 
 namespace at::native::xnnpack {

diff --git a/c10/util/Type_no_demangle.cpp b/c10/util/Type_no_demangle.cpp
@@ -1,7 +1,7 @@
-#include <c10/util/Type.h>
+#include <c10/macros/Macros.h>
 
 #if HAS_DEMANGLE == 0
-#include <string>
+#include <c10/util/Type.h>
 
 namespace c10 {
 std::string demangle(const char* name) {

diff --git a/test/cpp/jit/test_argument_spec.cpp b/test/cpp/jit/test_argument_spec.cpp
@@ -1,5 +1,6 @@
 #include <gtest/gtest.h>
 
+#include <ATen/ATen.h>
 #include <torch/csrc/jit/api/function_impl.h>
 #include <torch/csrc/jit/runtime/argument_spec.h>
 #include <torch/jit.h>

diff --git a/test/cpp/jit/test_concat_opt.cpp b/test/cpp/jit/test_concat_opt.cpp
@@ -1,5 +1,6 @@
 #include <gtest/gtest.h>
 
+#include <ATen/Functions.h>
 #include <test/cpp/jit/test_utils.h>
 #include <torch/csrc/jit/ir/irparser.h>
 #include <torch/csrc/jit/passes/concat_opt.h>

diff --git a/test/cpp/jit/test_shape_analysis.cpp b/test/cpp/jit/test_shape_analysis.cpp
@@ -1,5 +1,6 @@
 #include <gtest/gtest.h>
 
+#include <ATen/ATen.h>
 #include <ATen/core/interned_strings.h>
 #include <c10/util/Exception.h>
 #include <c10/util/Optional.h>

diff --git a/test/cpp/jit/test_stack_opt.cpp b/test/cpp/jit/test_stack_opt.cpp
@@ -1,3 +1,4 @@
+#include <ATen/Functions.h>
 #include <gtest/gtest.h>
 
 #include <test/cpp/jit/test_utils.h>

diff --git a/test/cpp/lite_interpreter_runtime/test_lite_interpreter_runtime.cpp b/test/cpp/lite_interpreter_runtime/test_lite_interpreter_runtime.cpp
@@ -1,3 +1,5 @@
+#include <ATen/Functions.h>
+#include <aten/src/ATen/TensorOperators.h>
 #include <gtest/gtest.h>
 #include <test/cpp/jit/test_utils.h>
 #include <torch/csrc/autograd/generated/variable_factories.h>

diff --git a/test/cpp/lite_interpreter_runtime/test_mobile_profiler.cpp b/test/cpp/lite_interpreter_runtime/test_mobile_profiler.cpp
@@ -1,3 +1,4 @@
+#include <ATen/Functions.h>
 #include <gtest/gtest.h>
 #include <test/cpp/jit/test_utils.h>
 #include <torch/csrc/jit/api/module.h>

diff --git a/tools/iwyu/aten.imp b/tools/iwyu/aten.imp
@@ -3,34 +3,19 @@
     { include: ["@<ATen/cpu/vec/vec256/.*>", private, "<ATen/cpu/vec/vec.h>", public] },
     { include: ["<ATen/cpu/vec/vec_base.h>", private, "<ATen/cpu/vec/vec.h>", public] },
     { include: ["<ATen/cpu/vec/functional_base.h>", private, "<ATen/cpu/vec/functional.h>", public] },
+    { include: ["<ATen/core/IListRef_inl.h>", private, "<ATen/core/IListRef.h>", public] },
     { include: ["<ATen/cpu/vec/functional_bfloat16.h>", private, "<ATen/cpu/vec/functional.h>", public] },
 
     { include: ["<ATen/core/Dict_inl.h>", private, "<ATen/core/Dict.h>", public] },
     { include: ["<ATen/core/ivalue_inl.h>", private, "<ATen/core/ivalue.h>", public] },
     { include: ["<ATen/core/DimVector.h>", public, "<ATen/DimVector.h>", public] },
     { include: ["<ATen/core/Dimname.h>", public, "<ATen/Dimname.h>", public] },
-    { include: ["<ATen/core/symbol.h>", public, "<ATen/core/Dimname.h>", public] },
-    { include: ["<ATen/TensorIterator.h>", public, "<ATen/core/TensorIterator.h>", public] },
 
-    { include: ["<c10/core/WrapDimMinimal.h>", public, "<ATen/WrapDimUtils.h>", public] },
-    { include: ["<c10/util/SmallVector.h>", public, "<ATen/core/DimVector.h>", public] },
-
-    { include: ["<c10/util/MaybeOwned.h>", public, "<ATen/core/TensorBase.h>", public] },
-    { include: ["<c10/core/TensorImpl.h>", public, "<ATen/core/TensorBase.h>", public] },
-    { include: ["<c10/core/UndefinedTensorImpl.h>", public, "<ATen/core/TensorBase.h>", public] },
-    { include: ["<c10/core/TensorOptions.h>", public, "<ATen/core/TensorBase.h>", public] },
-
-    { include: ["<ATen/core/TensorBase.h>", public, "<ATen/core/Tensor.h>", public] },
-    { include: ["<ATen/core/TensorBody.h>", private, "<ATen/core/Tensor.h>", public] },
     { symbol: ["at::Tensor", private, "<ATen/core/Tensor.h>", public] },
 
     { include: ["<ATen/Parallel-inl.h>", private, "<ATen/Parallel.h>", public] },
     { include: ["<ATen/ParallelOpenMP.h>", private, "<ATen/Parallel.h>", public] },
     { include: ["<ATen/ParallelNative.h>", private, "<ATen/Parallel.h>", public] },
     { include: ["<ATen/ParallelNativeTBB.h>", private, "<ATen/Parallel.h>", public] },
 
-    { include: ["<qnnpack/common.h>", public, "<pack_block_sparse.h>", public] },
-    { include: ["<qnnpack/math.h>", public, "<pack_block_sparse.h>", public] },
-    { include: ["<qnnpack/operator.h>", public, "<qnnpack_func.h>", public] },
-    { include: ["<qnnpack/log.h>", public, "<pytorch_qnnpack.h>", public] },
 ]
diff --git a/tools/iwyu/c10.imp b/tools/iwyu/c10.imp
@@ -22,4 +22,6 @@
     { include: ["<c10/core/StorageImpl.h>", private, "<c10/core/Storage.h>", public ] },
     { symbol: [ "C10_API", private, "<c10/macros/Export.h>", public ] },
     { symbol: [ "c10::complex", private, "<c10/util/complex.h>", public ] },
+    { symbol: [ "fmt::format", private, "<fmt/format.h>", public ] },
+    { symbol: [ "fmt::vformat", private, "<fmt/format.h>", public ] },
 ]
diff --git a/tools/iwyu/torch.imp b/tools/iwyu/torch.imp
@@ -1,2 +1,3 @@
 [
+    { include: ["<bits/chrono.h>", private, "<chrono>", public ] },
 ]
diff --git a/torch/csrc/jit/passes/freeze_module.cpp b/torch/csrc/jit/passes/freeze_module.cpp
@@ -10,7 +10,6 @@
 #include <torch/csrc/jit/passes/eliminate_no_ops.h>
 #include <torch/csrc/jit/passes/inliner.h>
 #include <torch/csrc/jit/passes/lower_tuples.h>
-#include <torch/csrc/jit/passes/remove_mutation.h>
 #include <torch/csrc/jit/runtime/graph_executor_impl.h>
 
 #include <stack>

diff --git a/torch/csrc/jit/passes/specialize_autogradzero.cpp b/torch/csrc/jit/passes/specialize_autogradzero.cpp
@@ -3,7 +3,6 @@
 #include <c10/util/Exception.h>
 #include <torch/csrc/jit/ir/ir.h>
 #include <torch/csrc/jit/jit_log.h>
-#include <torch/csrc/jit/passes/clear_undefinedness.h>
 #include <torch/csrc/jit/runtime/graph_executor.h>
 #include <torch/csrc/jit/runtime/profiling_record.h>
 

diff --git a/torch/csrc/jit/passes/tensorexpr_fuser.h b/torch/csrc/jit/passes/tensorexpr_fuser.h
@@ -7,8 +7,6 @@
 namespace torch {
 namespace jit {
 
-struct Graph;
-
 // Run TensorExpressions-based fuser.
 // If add_composed_op is true, creates a single operation that
 // performs both the runtime check that types align

diff --git a/torch/csrc/jit/passes/vulkan_rewrite.cpp b/torch/csrc/jit/passes/vulkan_rewrite.cpp
@@ -1,7 +1,6 @@
 #include <ATen/core/jit_type.h>
 #include <torch/csrc/jit/ir/ir.h>
 #include <torch/csrc/jit/ir/subgraph_matcher.h>
-#include <torch/csrc/jit/passes/constant_pooling.h>
 #include <torch/csrc/jit/passes/dead_code_elimination.h>
 #include <torch/csrc/jit/passes/fold_conv_bn.h>
 #include <torch/csrc/jit/passes/freeze_module.h>

diff --git a/torch/csrc/jit/passes/xnnpack_rewrite.cpp b/torch/csrc/jit/passes/xnnpack_rewrite.cpp
@@ -2,16 +2,13 @@
 #include <ATen/native/xnnpack/OpContext.h>
 
 #include <torch/csrc/jit/ir/ir.h>
-#include <torch/csrc/jit/ir/subgraph_matcher.h>
-#include <torch/csrc/jit/passes/constant_pooling.h>
 #include <torch/csrc/jit/passes/constant_propagation.h>
 #include <torch/csrc/jit/passes/fold_conv_bn.h>
 #include <torch/csrc/jit/passes/freeze_module.h>
 #include <torch/csrc/jit/passes/fuse_linear.h>
 #include <torch/csrc/jit/passes/fuse_relu.h>
 #include <torch/csrc/jit/passes/graph_rewrite_helper.h>
 #include <torch/csrc/jit/passes/hoist_conv_packed_params.h>
-#include <torch/csrc/jit/passes/inliner.h>
 #include <torch/csrc/jit/passes/mobile_optimizer_type.h>
 #include <torch/csrc/jit/passes/prepack_folding.h>
 #include <torch/csrc/jit/passes/remove_dropout.h>

diff --git a/torch/csrc/jit/runtime/autodiff.cpp b/torch/csrc/jit/runtime/autodiff.cpp
@@ -3,10 +3,8 @@
 #include <ATen/core/functional.h>
 #include <c10/util/Exception.h>
 #include <c10/util/irange.h>
-#include <torch/csrc/jit/frontend/ir_emitter.h>
 #include <torch/csrc/jit/jit_log.h>
 #include <torch/csrc/jit/passes/common_subexpression_elimination.h>
-#include <torch/csrc/jit/passes/constant_pooling.h>
 #include <torch/csrc/jit/passes/dead_code_elimination.h>
 #include <torch/csrc/jit/passes/inliner.h>
 #include <torch/csrc/jit/passes/lower_tuples.h>

diff --git a/torch/csrc/jit/runtime/autodiff.h b/torch/csrc/jit/runtime/autodiff.h
@@ -3,8 +3,6 @@
 #include <torch/csrc/Export.h>
 #include <torch/csrc/jit/ir/ir.h>
 
-#include <ATen/ATen.h>
-
 #include <memory>
 #include <vector>
 

diff --git a/torch/csrc/jit/runtime/operator.cpp b/torch/csrc/jit/runtime/operator.cpp
@@ -1,7 +1,6 @@
 #include <torch/csrc/jit/runtime/operator.h>
 
 #include <ATen/ATen.h>
-#include <ATen/core/alias_info.h>
 #include <ATen/core/interned_strings.h>
 #include <c10/util/irange.h>
 #include <torch/csrc/jit/frontend/edit_distance.h>

diff --git a/torch/csrc/jit/runtime/profiling_record.cpp b/torch/csrc/jit/runtime/profiling_record.cpp
@@ -5,7 +5,6 @@
 #include <torch/csrc/jit/codegen/cuda/interface.h>
 #include <torch/csrc/jit/jit_log.h>
 #include <torch/csrc/jit/passes/clear_profiling.h>
-#include <torch/csrc/jit/passes/constant_propagation.h>
 #include <torch/csrc/jit/passes/tensorexpr_fuser.h>
 #include <torch/csrc/jit/runtime/autodiff.h>
 #include <torch/csrc/jit/runtime/graph_executor.h>

diff --git a/torch/csrc/jit/runtime/register_c10_ops.cpp b/torch/csrc/jit/runtime/register_c10_ops.cpp
@@ -4,7 +4,6 @@
 #include <torch/csrc/jit/frontend/tracer.h>
 #include <torch/csrc/jit/ir/ir.h>
 #include <torch/csrc/jit/runtime/operator.h>
-#include <unordered_set>
 
 namespace torch::jit {
 

diff --git a/torch/csrc/jit/runtime/simple_graph_executor_impl.cpp b/torch/csrc/jit/runtime/simple_graph_executor_impl.cpp
@@ -1,7 +1,6 @@
 #include <torch/csrc/jit/runtime/profiling_graph_executor_impl.h>
 
 #include <c10/util/Optional.h>
-#include <torch/csrc/jit/jit_log.h>
 #include <torch/csrc/jit/runtime/simple_graph_executor_impl.h>
 #include <mutex>