TFE_Context* ctx = TFE_NewContext(opts, status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_DeleteContextOptions(opts);

  TFE_TensorHandle* input1 = TestMatrixTensorHandle(ctx);
  TFE_TensorHandle* input2 = TestMatrixTensorHandle(ctx);
  TFE_TensorHandle* dim = TestScalarTensorHandle(ctx, 0);
  TFE_Op* concatOp = TFE_NewOp(ctx, "Concat", status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);

      /*is_async=*/async, /*enable_streaming_enqueue=*/true,
      /*in_flight_nodes_limit=*/0);
  TFE_ContextSetExecutorForThread(ctx, executor);

  TFE_TensorHandle* m = TestMatrixTensorHandle(ctx);
  TFE_Op* matmul = MatMulOp(ctx, m, m);
  TFE_TensorHandle* retvals[2] = {nullptr, nullptr};
  int num_retvals = 2;
  TFE_Execute(matmul, &retvals[0], &num_retvals, status);
  EXPECT_EQ(1, num_retvals);

// Return a tensor handle containing a 2x2 matrix of doubles
TFE_TensorHandle* DoubleTestMatrixTensorHandle(TFE_Context* ctx);

// Return a tensor handle containing a 2x2 matrix of floats
TFE_TensorHandle* TestMatrixTensorHandle(TFE_Context* ctx);

// Return a tensor handle containing 2D matrix containing given data and
// dimensions
TFE_TensorHandle* TestMatrixTensorHandleWithInput(TFE_Context* ctx,

  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TF_DeleteTensor(t);
  TF_DeleteStatus(status);
  return th;
}

TFE_TensorHandle* TestMatrixTensorHandle(TFE_Context* ctx) {
  int64_t dims[] = {2, 2};
  float data[] = {1.0f, 2.0f, 3.0f, 4.0f};
  TF_Status* status = TF_NewStatus();
  TF_Tensor* t = TFE_AllocateHostTensor(ctx, TF_FLOAT, &dims[0],