当向量元素超过线程个数时的情况

向量元素个数为（33 * 1024）/（128 * 128）=2.x倍

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 govern12 * 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 22 #define N   (33 * 1024)23 24 __global__ void add(int *a, int *b, int *c) {25     int tid = threadIdx.x + blockIdx.x * blockDim.x;26     while (tid < N) {27         c[tid] = a[tid] + b[tid];28         tid += blockDim.x * gridDim.x;29     }30 }31 32 int main(void) {33     int *a, *b, *c;34     int *dev_a, *dev_b, *dev_c;35 36     // allocate the memory on the CPU37     a = (int*)malloc(N * sizeof(int));38     b = (int*)malloc(N * sizeof(int));39     c = (int*)malloc(N * sizeof(int));40 41     // allocate the memory on the GPU42     HANDLE_ERROR(cudaMalloc((void**)&dev_a, N * sizeof(int)));43     HANDLE_ERROR(cudaMalloc((void**)&dev_b, N * sizeof(int)));44     HANDLE_ERROR(cudaMalloc((void**)&dev_c, N * sizeof(int)));45 46     // fill the arrays 'a' and 'b' on the CPU47     for (int i = 0; i
     
      > >(dev_a, dev_b, dev_c);63 64     // copy the array 'c' back from the GPU to the CPU65     HANDLE_ERROR(cudaMemcpy(c, dev_c, N * sizeof(int),66         cudaMemcpyDeviceToHost));67 68     // verify that the GPU did the work we requested69     bool success = true;70     for (int i = 0; i