Function CudaLaunchKernel

tensorflow/core/util/gpu_kernel_helper.h:127–130

// builds
template <typename... Args>
auto CudaLaunchKernel(Args&&... args)
    -> decltype(GpuLaunchKernel(std::forward<Args>(args)...)) {
  return GpuLaunchKernel(std::forward<Args>(args)...);
}

__host__ __device__ inline tensorflow::bfloat16 GpuLdg(
    const tensorflow::bfloat16* address) {