向量在同一个块中用多个线程执行。
本实验是启动一个线程块，这个线程块中启动的线程个数为10个。
1 /*
2 * Copyright 1993-2010 NVIDIA Corporation. All rights reserved.
3 *
4 * NVIDIA Corporation and its licensors retain all intellectual property and
5 * proprietary rights in and to this software and related documentation.
6 * Any use, reproduction, disclosure, or distribution of this software
7 * and related documentation without an express license agreement from
8 * NVIDIA Corporation is strictly prohibited.
9 *
10 * Please refer to the applicable NVIDIA end user license agreement (EULA)
11 * associated with this source code for terms and conditions that govern
12 * your use of this NVIDIA software.
13 *
14 */
15
16
17 #include "../common/book.h"
18 #include "cuda.h"
19 #include "cuda_runtime.h"
20 #include "device_launch_parameters.h"
21 #define N 10
22
23 __global__ void add(int *a, int *b, int *c) {
24 int tid = threadIdx.x;
25 if (tid < N)
26 c[tid] = a[tid] + b[tid];
27 }
28
29 int main(void) {
30 int a[N], b[N], c[N];
31 int *dev_a, *dev_b, *dev_c;
32
33 // allocate the memory on the GPU
34 HANDLE_ERROR(cudaMalloc((void**)&dev_a, N * sizeof(int)));
35 HANDLE_ERROR(cudaMalloc((void**)&dev_b, N * sizeof(int)));
36 HANDLE_ERROR(cudaMalloc((void**)&dev_c, N * sizeof(int)));
37
38 // fill the arrays 'a' and 'b' on the CPU
39 for (int i = 0; i<N; i++) {
40 a[i] = i;
41 b[i] = i * i;
42 }
43
44 // copy the arrays 'a' and 'b' to the GPU
45 HANDLE_ERROR(cudaMemcpy(dev_a, a, N * sizeof(int), cudaMemcpyHostToDevice));
46 HANDLE_ERROR(cudaMemcpy(dev_b, b, N * sizeof(int), cudaMemcpyHostToDevice));
47
48 /*
49 第一个参数1表示启动一个执行副本，也就是启动一个线程块
50 第二个参数N表示每个线程块中启动的线程个数为N个
51 */
52 add << <1, N >> >(dev_a, dev_b, dev_c);
53
54 // copy the array 'c' back from the GPU to the CPU
55 HANDLE_ERROR(cudaMemcpy(c, dev_c, N * sizeof(int), cudaMemcpyDeviceToHost));
56
57 // display the results
58 for (int i = 0; i<N; i++) {
59 printf("%d + %d = %d
", a[i], b[i], c[i]);
60 }
61
62 // free the memory allocated on the GPU
63 HANDLE_ERROR(cudaFree(dev_a));
64 HANDLE_ERROR(cudaFree(dev_b));
65 HANDLE_ERROR(cudaFree(dev_c));
66
67 return 0;
68 }