main.cu

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include "core.cu"

// CUDA通用函数声明
void (*func)(int, int, int, float *, float *, int **);

// 生成随机数
float getRand()
{
    return rand() / double(RAND_MAX);
}

/**
 * @brief Get the Sample object
 * 
 * @param k int 样例维度
 * @param m int 测试点数量
 * @param n int 参考点数量
 * @param s_points float** 测试点数据指针
 * @param r_points float** 参考点数据指针
 */
void getSample(int k, int m, int n, float **s_points, float **r_points)
{
    float *tmp;
    tmp = (float *)malloc(sizeof(float) * k * m);
    for (int i = 0; i < k * m; i++)
        tmp[i] = getRand();
    *s_points = tmp;
    tmp = (float *)malloc(sizeof(float) * k * n);
    for (int i = 0; i < k * n; i++)
        tmp[i] = getRand();
    *r_points = tmp;
}

// 样例设置
int samples[] = {
    3,  1,      1024,
    16, 1,      1024,
    3,  1,      65536,
    16, 1,      65536,

    3,  1024,   1024,
    16, 1024,   1024,
    3,  1024,   65536,
    16, 1024,   65536,

    3,  1024,   1048576,
    16, 1024,   1048576
};

int total = 0;   // 样例个数
int seed = 1000; // 随机种子
long st, et;     // 开始和结束的时间

/**
 * @brief 调用对应的核函数
 * 
 * @param v 指定 version 版本
 */
void test(int v)
{
    srand(seed);
    for (int i = 0; i < total; ++i)
    {
        int k = samples[3 * i];
        int m = samples[3 * i + 1];
        int n = samples[3 * i + 2];
        float *s_points, *r_points;
        getSample(k, m, n, &s_points, &r_points);
        int *results;
        st = getTime();
        (*func)(k, m, n, s_points, r_points, &results);
        et = getTime();
        printf("CudaCall %d, %2d, %4d, %10d, %10.3fms\n", v, k, m, n, (et - st) / 1e6);
        free(s_points);
        free(r_points);
    }
}

// 主函数
int main()
{
    total = sizeof(samples) / (3 * sizeof(*samples)); // 样例个数
    // 运行全部的优化版本
    for (int v = 0; v < 14; ++v)
    {
        switch (v)
        {
        case 0:
            func = &v0::cudaCall;
            break;
        case 1:
            func = &v1::cudaCall;
            break;
        case 2:
            func = &v2::cudaCall;
            break;
        case 3:
            func = &v3::cudaCall;
            break;
        case 4:
            func = &v4::cudaCall;
            break;
        case 5:
            func = &v5::cudaCall;
            break;
        case 6:
            func = &v6::cudaCall;
            break;
        case 7:
            func = &v7::cudaCall;
            break;
        case 8:
            func = &v8::cudaCall;
            break;
        case 9:
            func = &v9::cudaCall;
            break;
        case 10:
            func = &v10::cudaCall;
            break;
        case 11:
            func = &v11::cudaCall;
            break;
        case 12:
            func = &v12::cudaCall;
            break;
        case 13:
            func = &v13::cudaCall;
            break;
        default:
            break;
        }
        printf("\nRunning CUDACALL %d...\n", v);
        test(v);
    }
}