TFE_Op* matmul = MatMulOp(ctx, hcpu, hgpu);
    if (cpu_op) {
      string cpu_device_name;
      ASSERT_TRUE(GetDeviceName(ctx, &cpu_device_name, "CPU"));
      TFE_OpSetDevice(matmul, cpu_device_name.c_str(), status.get());
    } else {
      TFE_OpSetDevice(matmul, gpu_device_name.c_str(), status.get());
    }
    ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());

    TFE_TensorHandle* retvals[1];