https://developer.nvidia.com/zh-cn/blog/how-overlap-data-transfers-cuda-cc/

分批拷贝：

for (int i = 0; i < nStreams; ++i) {
  int offset = i * streamSize;
  cudaMemcpyAsync(&d_a[offset], &a[offset],
                  streamBytes, cudaMemcpyHostToDevice, cudaMemcpyHostToDevice, stream[i]);
}

for (int i = 0; i < nStreams; ++i) {
  int offset = i * streamSize;
  kernel<<<streamSize/blockSize, blockSize, 0, stream[i]>>>(d_a, offset);
}

for (int i = 0; i < nStreams; ++i) {
  int offset = i * streamSize;
  cudaMemcpyAsync(&a[offset], &d_a[offset],
                  streamBytes, cudaMemcpyDeviceToHost, cudaMemcpyDeviceToHost, stream[i]);
}