slim tensor supported

Author: Gasoonjia

backends/cuda/CMakeLists.txt

@@ -70,12 +70,7 @@ install(
 # CUDA-specific AOTI functionality
 set(_aoti_cuda_sources
     runtime/cuda_backend.cpp
-    runtime/common_shims.cpp
-    runtime/shims/memory.cpp
-    runtime/shims/tensor_attribute.cpp
     runtime/guard.cpp
-    runtime/shims/cuda_guard.cpp
-    runtime/shims/int4mm.cu
     runtime/platform/platform.cpp
 )
 add_library(aoti_cuda STATIC ${_aoti_cuda_sources})

backends/cuda/runtime/TARGETS

@@ -50,18 +50,9 @@ runtime.cxx_library(
     name = "runtime_shims",
     srcs = [
         "guard.cpp",
-        "shims/cuda_guard.cpp",
-        "shims/int4mm.cu",
-        "shims/memory.cpp",
-        "shims/tensor_attribute.cpp",
     ],
     headers = [
         "guard.h",
-        "shims/cuda_guard.h",
-        "shims/int4mm.cuh",
-        "shims/int4mm.h",
-        "shims/memory.h",
-        "shims/tensor_attribute.h",
         "utils.h",
     ],
     # @lint-ignore BUCKLINT: Avoid `link_whole=True` (https://fburl.com/avoid-link-whole)
@@ -72,7 +63,6 @@ runtime.cxx_library(
     visibility = ["@EXECUTORCH_CLIENTS"],
     deps = [
         ":tensor_maker",
-        "//executorch/backends/aoti:common_shims",
         "//executorch/runtime/core:core",
         "//executorch/runtime/core/exec_aten:lib",
         "//executorch/runtime/platform:platform",
@@ -100,7 +90,6 @@ runtime.cxx_library(
     visibility = ["@EXECUTORCH_CLIENTS"],
     deps = [
         ":runtime_shims",
-        "//executorch/backends/aoti:aoti_common",
         "//executorch/runtime/backend:interface",
         "//executorch/runtime/core/exec_aten/util:tensor_util",
     ],

backends/cuda/runtime/aoti_delegate_handle.h

@@ -10,6 +10,7 @@
 
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/core/evalue.h>
+#include <executorch/backends/cuda/runtime/slim/core/SlimTensor.h>
 
 namespace executorch {
 namespace backends {
@@ -21,7 +22,7 @@ using executorch::runtime::etensor::Tensor;
 extern "C" {
 
 // Type definitions
-using AOTITensorHandle = Tensor*;
+using AOTITensorHandle = standalone::slim::SlimTensor*;
 using AOTIRuntimeError = Error;
 
 // Forward declarations for AOT Inductor model container
@@ -50,12 +51,12 @@ using AOTInductorModelContainerGetNumOutputsFunc = AOTIRuntimeError (*)(
 
 using AOTInductorModelContainerRunFunc = AOTIRuntimeError (*)(
     AOTInductorModelContainerHandle container_handle,
-    Tensor** input_handles, // array of input Tensor*; handles
-                            // are stolen; the array itself is borrowed
+    AOTITensorHandle* input_handles, // array of input SlimTensor*; handles
+                                      // are stolen; the array itself is borrowed
     size_t num_inputs,
-    Tensor** output_handles, // array for writing output Tensor*; handles
-                             // will be stolen by the caller; the array itself
-                             // is borrowed
+    AOTITensorHandle* output_handles, // array for writing SlimTensor*; handles
+                                       // will be stolen by the caller; the array itself
+                                       // is borrowed
     size_t n_outputs,
     AOTInductorStreamHandle stream_handle,
     AOTIProxyExecutorHandle proxy_executor_handle);