[inductor] Move AOTInductor runtime headers (#108564)

Summary: Move AOTInductor runtime header files into its own subdirectory, to separate them from to-be-added libtorch C interface.

Reviewed By: frank-wei

Differential Revision: D48905038

Pull Request resolved: https://github.com/pytorch/pytorch/pull/108564
Approved by: https://github.com/frank-wei

benchmarks/dynamo/common.py

@@ -1149,7 +1149,7 @@ class AOTInductorModelCache:
 
             # Use a utility function for easier benchmarking
             source = """
-            #include <torch/csrc/inductor/aot_inductor_model.h>
+            #include <torch/csrc/inductor/aot_runtime/model.h>
 
             torch::aot_inductor::AOTInductorModel model;
 

setup.py

@@ -1291,6 +1291,7 @@ def main():
         "include/torch/csrc/distributed/autograd/rpc_messages/*.h",
         "include/torch/csrc/dynamo/*.h",
         "include/torch/csrc/inductor/*.h",
+        "include/torch/csrc/inductor/aot_runtime/*.h",
         "include/torch/csrc/jit/*.h",
         "include/torch/csrc/jit/backends/*.h",
         "include/torch/csrc/jit/generated/*.h",
@@ -1333,8 +1334,8 @@ def main():
         "include/THH/*.h*",
         "include/THH/generic/*.h",
         "include/sleef.h",
-        "_inductor/codegen/*.cpp",
         "_inductor/codegen/*.h",
+        "_inductor/codegen/aot_runtime/*.cpp",
         "share/cmake/ATen/*.cmake",
         "share/cmake/Caffe2/*.cmake",
         "share/cmake/Caffe2/public/*.cmake",

test/cpp/aot_inductor/test.cpp

@@ -3,7 +3,7 @@
 #include <vector>
 
 #include <c10/cuda/CUDAStream.h>
-#include <torch/csrc/inductor/aot_inductor_interface.h>
+#include <torch/csrc/inductor/aot_runtime/interface.h>
 #include <torch/torch.h>
 
 namespace torch {

test/inductor/test_aot_inductor.py

@@ -37,7 +37,7 @@ class AOTInductorModelRunner:
 
         # Use a utility function for easier testing
         source = """
-        #include <torch/csrc/inductor/aot_inductor_model.h>
+        #include <torch/csrc/inductor/aot_runtime/model.h>
 
         torch::aot_inductor::AOTInductorModel model;
 

torch/_inductor/codegen/aot_runtime/interface.cpp

@@ -1,8 +1,8 @@
-#include <torch/csrc/inductor/aot_inductor_interface.h>
-#include <torch/csrc/inductor/aot_inductor_model_container.h>
+#include <torch/csrc/inductor/aot_runtime/interface.h>
+#include <torch/csrc/inductor/aot_runtime/model_container.h>
 #include <ATen/core/dispatch/Dispatcher.h>
 #include <iostream>
-#include <torch/csrc/inductor/proxy_executor.h>
+#include <torch/csrc/inductor/aot_runtime/proxy_executor.h>
 #include <stdexcept>
 #include <vector>
 

torch/_inductor/codegen/wrapper.py

@@ -930,7 +930,7 @@ class CppWrapperCodeGen(WrapperCodeGen):
     def write_header(self):
         if V.graph.aot_mode:
             with open(
-                os.path.join(os.path.dirname(__file__), "aot_inductor_interface.cpp")
+                os.path.join(os.path.dirname(__file__), "aot_runtime", "interface.cpp")
             ) as f:
                 self.header.splice(f.read())
         else:

torch/csrc/inductor/aot_runtime/model.h

@@ -7,7 +7,7 @@
 #include <ATen/ATen.h>
 
 #include <c10/cuda/CUDAGuard.h>
-#include <torch/csrc/inductor/proxy_executor.h>
+#include <torch/csrc/inductor/aot_runtime/proxy_executor.h>
 
 #define AOT_VECTOR_SIZE_CHECK(vec, expected_size) \
   {                                               \

torch/csrc/inductor/aot_runtime/model_container.h

@@ -4,8 +4,8 @@
 #include <mutex>
 #include <shared_mutex>
 
-#include <torch/csrc/inductor/aot_inductor_model.h>
-#include <torch/csrc/inductor/proxy_executor.h>
+#include <torch/csrc/inductor/aot_runtime/model.h>
+#include <torch/csrc/inductor/aot_runtime/proxy_executor.h>
 
 namespace torch {
 namespace aot_inductor {