33 lines
721 B
Plaintext
33 lines
721 B
Plaintext
#include <cstdio>
|
|
using namespace std;
|
|
|
|
|
|
__global__ void hell_from__gpu()
|
|
{
|
|
// 核函数不支持 c++ 的 iostream。
|
|
|
|
// 输出流的缓存顺序。
|
|
// printf("gpu: hello world! ");
|
|
|
|
const int bx = blockIdx.x;
|
|
const int by = blockIdx.y;
|
|
const int bz = blockIdx.z;
|
|
|
|
const int tx = threadIdx.x;
|
|
const int ty = threadIdx.y;
|
|
const int tz = threadIdx.z;
|
|
|
|
printf("gpu: hello world! block(%d, %d, %d) -- thread(%d, %d, %d)\n", bx, by, bz, tx, ty, tz);
|
|
}
|
|
|
|
|
|
int main()
|
|
{
|
|
printf("nvcc: hello world!\n");
|
|
|
|
const dim3 block_size(2, 4);
|
|
hell_from__gpu<<<1, block_size>>>();
|
|
cudaDeviceSynchronize(); // 同步主机和设备，否则无法输出字符串。
|
|
|
|
return 0;
|
|
} |