//一些基本的cublas示例
#include<cuda_runtime.h>
#include<device_launch_parameters.h>
#include<cublas_v2.h>
#include<stdlib.h>
#include<assert.h>
#include<math.h>
//初始化向量
void vector_init(float*a, int n) {
for (int i = 0; i < n; i++) {
a[i] = (float)(rand() % 100);
}
}
//验证结果
void verify_result(float*a, float*b, float*c, float factor, int n) {
for (int i = 0; i < n; i++) {
assert(c[i] == factor * a[i] + b[i]);
}
}
int main() {
int n = 1 << 16;//向量大小
size_t bytes = n * sizeof(float);
float *h_a, *h_b, *h_c;//声明向量指针
float *d_a, *d_b;
//分配内存
h_a = (float*)malloc(bytes);
h_b = (float*)malloc(bytes);
h_c = (float*)malloc(bytes);
cudaMalloc(&d_a, bytes);
cudaMalloc(&d_b, bytes);
//初始化向量
vector_init(h_a, n);
vector_init(h_b, n);
//创建handle并初始化
cublasHandle_t handle;
cublasCreate_v2(&handle);
//将向量复制到设备上
cublasSetVector(n, sizeof(float), h_a, 1, d_a, 1);
cublasSetVector(n, sizeof(float), h_b, 1, d_b, 1);
//启动简单的saxpy内核
const float scale = 2.0f;
cublasSaxpy(handle, n, &scale, d_a, 1, d_b, 1);
//复制结果向量
cublasGetVector(n, sizeof(float), d_b, 1, h_c, 1);
//打印结果
verify_result(h_a, h_b, h_c, scale, n);
//清理创建的handle
cublasDestroy(handle);
cudaFree(d_a);
cudaFree(d_b);
free(h_a);
free(h_b);
return 0;
}