CUDA event 记录GPU端运行时间

摘录自CUDA sample：asyncAPI

...
  cudaEvent_t start, stop;
  checkCudaErrors(cudaEventCreate(&start));
  checkCudaErrors(cudaEventCreate(&stop));
  
  StopWatchInterface *timer = NULL;
  sdkCreateTimer(&timer);
  sdkResetTimer(&timer);

  checkCudaErrors(cudaDeviceSynchronize());
  float gpu_time = 0.0f;

  sdkStartTimer(&timer);
  cudaEventRecord(start, 0);
  cudaMemcpyAsync(d_a, a, nbytes, cudaMemcpyHostToDevice, 0);
  increment_kernel<<<blocks, threads, 0, 0>>>(d_a, value);
  cudaMemcpyAsync(a, d_a, nbytes, cudaMemcpyDeviceToHost, 0);
  cudaEventRecord(stop, 0);
  sdkStopTimer(&timer);

  unsigned long int counter = 0;
  while (cudaEventQuery(stop) == cudaErrorNotReady){
    counter ++;
  }
  checkCudaErrors(cudaEventElapsedTime(&gpu_time, start, stop));

  printf("time spent executing by the GPU: %.2f\n", gpu_time);
  printf("time spent by CPU in CUDA calls: %.2f\n", sdkGetTimerValue(&timer));
  printf("CPU executed %lu iterations while waiting for GPU to finish\n", counter);

  bool bFinalResults = correct_output(a, n ,value);

  checkCudaErrors(cudaEventDestroy(start));
  checkCudaErrors(cudaEventDestroy(stop));
...