CUDA 编程，运行时间的统计

CUDA 编程，运行时间的统计

https://blog.csdn.net/QLeelq/article/details/127528370

单位为毫秒，精度为0.5微妙

单位是ms啊

cudaEvent_t start, stop;
float time;
cudaEventCreate(&start); //创建计时
cudaEventCreate(&stop); //创建计时
cudaEventRecord(start, 0); // 开始计时


cudaStatus = cudaDeviceSynchronize();
if (cudaStatus != cudaSuccess) {        
    fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching Kernel!\n", cudaStatus);
    goto Error;
}

cudaEventRecord(stop, 0);
cudaEventSynchronize(stop);
cudaEventElapsedTime(&time, start, stop);
cudaEventDestroy(start); //销毁
cudaEventDestroy(stop);
printf("Lee with GPU cost %f ms\n", time);

---

CPU的方法是

https://blog.csdn.net/m0_37251750/article/details/102581935

单位也是ms

精度太差了，不建议使用这个方法！！！

#include<time.h>
clock_t start, end;
start = clock();
end = clock();   //结束时间
printf("Lee with CPU cost %f ms\n", (float)(end-start));