code refacotr to backend/cuda and backend/aoti

Author: Gasoonjia

CMakeLists.txt

@@ -595,7 +595,7 @@ if(EXECUTORCH_BUILD_CUDA)
   # Build common AOTI functionality (required for CUDA)
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backends/aoti)
   # Build CUDA-specific AOTI functionality
-  add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backends/aoti/cuda)
+  add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/backends/cuda)
   # Add aoti_cuda to backends - it already depends on aoti_common
   list(APPEND _executorch_backends aoti_cuda)
 endif()

backends/aoti/aoti_model_container.h

@@ -8,9 +8,9 @@
 
 #pragma once
 
+#include <executorch/backends/cuda/runtime/shims/memory.h>
 #include <executorch/extension/tensor/tensor.h>
 #include <executorch/runtime/core/error.h>
-#include "cuda/runtime/shims/memory.h"
 
 namespace executorch {
 namespace backends {

backends/aoti/common_shims.cpp

@@ -106,7 +106,6 @@ AOTITorchError aoti_torch_get_storage_size(
   throw std::runtime_error("Cannot get storage size on ETensor");
 }
 
-
 AOTITorchError aoti_torch_get_device_index(
     AOTITensorHandle tensor,
     int32_t* ret_device_index) {

backends/aoti/utils.cpp

@@ -34,7 +34,8 @@ size_t dtype_to_element_size(int32_t dtype) {
 // Map int32_t dtype to ExecutorTorch ScalarType (robust version of hardcoded
 // ScalarType::Float)
 executorch::aten::ScalarType dtype_to_scalar_type(int32_t dtype) {
-  // Convert based on known PyTorch dtype codes (without CUDA-specific dependency)
+  // Convert based on known PyTorch dtype codes (without CUDA-specific
+  // dependency)
   switch (dtype) {
     case 6: // PyTorch's float32 dtype code
       return executorch::aten::ScalarType::Float;

backends/aoti/utils.h

@@ -10,8 +10,8 @@
 
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/core/exec_aten/exec_aten.h>
-#include <cstdint>
 #include <cstddef>
+#include <cstdint>
 
 namespace executorch {
 namespace backends {

backends/aoti/cuda/CMakeLists.txt renamed to backends/cuda/CMakeLists.txt

@@ -19,7 +19,7 @@ set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 # Source root directory for executorch.
 if(NOT EXECUTORCH_ROOT)
-  set(EXECUTORCH_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/../../..)
+  set(EXECUTORCH_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/../..)
 endif()
 
 find_package(CUDAToolkit REQUIRED)