测试 cuda gpu 一秒钟能launch 多少次 kernel

Eloudy

于 2023-05-24 11:11:40 发布

阅读量297

点赞数

文章标签： cuda

本文链接：https://blog.csdn.net/eloudy/article/details/130842950

版权

启动100000次带参数的空 kernel，总耗时124ms；

编译：

nvcc vectorAdd.cu

源文件：

vectorAdd.cu

#include <algorithm>
#include <stdio.h>
#include <iostream>
using namespace std;

__global__
void saxpy(int n, float a, float *x, float *y)
{
/*
  int i = blockIdx.x*blockDim.x + threadIdx.x;
  if (i < n)
    y[i] = a*x[i] + y[i];
*/
}

int main(void)
{
        int N = 1<<10;
        float *x, *y, *d_x, *d_y;
        x = (float*)malloc(N*sizeof(float));
        y = (float*)malloc(N*sizeof(float));

        cudaMalloc(&d_x, N*sizeof(float));
        cudaMalloc(&d_y, N*sizeof(float));

        for (int i = 0; i < N; i++) {
                x[i] = 1.0f;
                y[i] = 2.0f;
        }

        cudaMemcpy(d_x, x, N*sizeof(float), cudaMemcpyHostToDevice);
        cudaMemcpy(d_y, y, N*sizeof(float), cudaMemcpyHostToDevice);

        dim3 grid_;
        dim3 block_;

        block_.x = 256;
        grid_.x = (N+block_.x-1)/block_.x;

        cudaEvent_t start, end;

        cudaEventCreate(&start);
        cudaEventCreate(&end);
        cudaEventRecord(start);

  // Perform SAXPY on 1M elements
        for(unsigned i=0; i<100000; i++)
                saxpy<<<grid_, block_, 0, NULL>>>(N, 2.0f, d_x, d_y);


        cudaEventRecord(end);
        cudaEventSynchronize(end);
    //
        float time_ms = 0.0f;
        cudaEventElapsedTime(&time_ms, start, end);
        std::cout << "CUDA Kernel time: " << time_ms << " ms" << std::endl;

        cudaEventDestroy(start);
        cudaEventDestroy(end);

        cudaMemcpy(y, d_y, N*sizeof(float), cudaMemcpyDeviceToHost);

        float maxError = 0.0f;
        int result = 0;
        for (int i = 0; i < N; i++)
                maxError = max(maxError, abs(y[i]-4.0f));
        printf("Max error: %f\n", maxError);
        if(maxError < 1e-5){
                printf("SUCCESS\n");
        }
        else{
                result = 1;
                printf("FAILURE\n");
        }

        cudaFree(d_x);
        cudaFree(d_y);
        free(x);
        free(y);
        return result;
}