使用CUBLAS查找最大值和最小值

#include <stdio.h>
#include <cuda_runtime.h>

int main() {
    float a[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f};
    float b[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f};
    float c[] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f};
    float *d;
    cudaMalloc(&d, 5 * sizeof(float));
    cudaMemcpy(d, a, 5 * sizeof(float), cudaMemcpyHostToDevice);

    int block_size = 256;
    int grid_size = (2 * block_size + 2) / block_size;

    float alpha = 1.0f;
    float beta = 0.0f;

    cublasHandle_t h;
    cublasCreate(&h);

    cublasSgemm(h, CUBLAS_OP_T, CUBLAS_OP_N, 5, 5, 5, 5, &alpha, d, 5, d, 5, d, 5, &beta, c, 5);

    float max_value = 0.0f;
    float min_value = 0.0f;

    for (int i = 0; i < 5; i++) {
        max_value = fmax(max_value, c[i]);
        min_value = fmin(min_value, c[i]);
    }

    printf("Max value: %f, Min value: %f\n", max_value, min_value);

    cublasDestroy(h);
    cudaFree(d);

    return 0;
}