使用CUDA Runtime API测耗时

#include "cuda_runtime.h"
#include "device_launch_parameters.h"

#include <iostream>

//省略错误判断
int main()
{
    cudaEvent_t start, end;
    float elapsedTime;

    cudaEventCreate(&start);
    cudaEventCreate(&end);
    cudaEventRecord(start, 0);
    for (int i = 0; i < 99999; i++)
        ;
    cudaEventRecord(end, 0);
    cudaEventSynchronize(end);
    cudaEventElapsedTime(&elapsedTime, start, end);
    cudaEventDestroy(start);
    cudaEventDestroy(end);
    std::cout << elapsedTime << "ms" << std::endl;
    getchar();

    return 0;
}