// `struct_size` is used for struct member compatibility check between core TF
// and plug-ins with the same C API minor version. More info here:
// https://github.com/tensorflow/community/blob/master/rfcs/20200612-stream-executor-c-api/C_API_versioning_strategy.md
#define TF_VALIDATE_STRUCT_SIZE(STRUCT_NAME, STRUCT_OBJ, SIZE_VALUE_NAME) \
  do {                                                                    \

        "//tensorflow:android": [
            "//tensorflow/c/experimental/stream_executor:stream_executor_hdrs",
            "//tensorflow/core:portable_tensorflow_lib_lite",
        ],
        "//conditions:default": [
            ":tf_tensor",
            "//tensorflow/c/experimental/stream_executor",
            "//tensorflow/c/experimental/stream_executor:stream_executor_internal",
            "//tensorflow/core:framework",

        otherwise = [
            "@local_xla//xla/stream_executor/cuda:all_runtime",
            "@local_xla//xla/stream_executor/cuda:cuda_driver",
            "@local_xla//xla/stream_executor/cuda:cuda_platform",
            "@local_xla//xla/stream_executor/cuda:cudnn_plugin",
            "@local_xla//xla/stream_executor/cuda:cufft_plugin",
            "@local_xla//xla/stream_executor:cuda_platform",
        ],
    ),

  if (TF_GetCode(status) != TF_OK) return -1;
  auto stream = gcs_file->gcs_client.ReadObject(
      bucket, object, gcs::ReadRange(offset, offset + buffer_size));
  TF_SetStatusFromGCSStatus(stream.status(), status);
  if ((TF_GetCode(status) != TF_OK) &&
      (TF_GetCode(status) != TF_OUT_OF_RANGE)) {
    return -1;
  }
  int64_t read;
  auto content_length = stream.headers().find("content-length");

  // Load the library.
  TF_Status* status = TF_NewStatus();
  string lib_path =
      tensorflow::GetDataDependencyFilepath(tensorflow::io::JoinPath(
          "tensorflow", "c", "experimental", "stream_executor", "test",
          "test_pluggable_device.so"));
  TF_Library* lib = TF_LoadPluggableDeviceLibrary(lib_path.c_str(), status);
  TF_Code code = TF_GetCode(status);
  string status_msg(TF_Message(status));

@@ -3,6 +3,11 @@
    backported.  */
 #define _ISOMAC 1
 
+#define __GLIBC_USE(F)	__GLIBC_USE_ ## F
+
+# define __attribute_alloc_size__(params) \
+  __attribute__ ((__alloc_size__ params))
+
 #include <iostream>
 
 using namespace std;
diff --git a/sysdeps/aarch64/dl-machine.h b/sysdeps/aarch64/dl-machine.h
index 185402f..bbdeae0 100644
--- a/sysdeps/aarch64/dl-machine.h
+++ b/sysdeps/aarch64/dl-machine.h

  * Added function `experimental.extension_type.as_dict()`, which converts an instance of `tf.experimental.ExtensionType` to a `dict` representation.
* `stream_executor`
  * Top level `stream_executor` directory has been deleted, users should use equivalent headers and targets under `compiler/xla/stream_executor`.
* `tf.nn`

// have been executed. Note that "execution" here refers to kernel execution /
// scheduling of copies, etc. Similar to sync execution, it doesn't guarantee
// that lower level device queues (like GPU streams) have been flushed.
//
// This call may not block for execution of ops enqueued concurrently with this
// call.
TF_CAPI_EXPORT extern void TFE_ExecutorWaitForAllPendingNodes(
    TFE_Executor*, TF_Status* status);

// read current values from streamz counters defined in other modules.
//
// The code under test will have created streamz counters like this:
// auto* streamz = tensorflow::monitoring::Counter<1>::New("name",
// "description", "label");
// and then incremented that counter for various values of label:
// streamz->GetCell("label-value")->IncrementBy(1);
//

    // Special-cased operation for packing per-device tensors into one parallel
    // tensor.
    if (inputs.size() != parallel_device.num_underlying_devices()) {
      std::string message(absl::StrCat(
          "The parallel device ", parallel_device_name, " expected ",
          parallel_device.num_underlying_devices(),
          " inputs to TPUReplicatedInput, but got ", inputs.size()));