#include <cuda_runtime.h>
#include <iostream>
<p><strong>global</strong> void add(int<em> a, int</em> b, int<em> c, int n) {
int idx = blockIdx.x </em> blockDim.x + threadIdx.x;
if (idx < n) {
c[idx] = a[idx] + b[idx];
}
}</p><p>int main() {
const int n = 1024;
size_t bytes = n * sizeof(int);</p><pre class='brush:php;toolbar:false;'>int *h_a = new int[n], *h_b = new int[n], *h_c = new int[n];
int *d_a, *d_b, *d_c;

// 分配GPU内存
cudaMalloc(&d_a, bytes);
cudaMalloc(&d_b, bytes);
cudaMalloc(&d_c, bytes);

// 初始化输入数据
for (int i = 0; i < n; ++i) {
    h_a[i] = i;
    h_b[i] = i * 2;
}

// 主机到设备数据传输
cudaMemcpy(d_a, h_a, bytes, cudaMemcpyHostToDevice);
cudaMemcpy(d_b, h_b, bytes, cudaMemcpyHostToDevice);

// 配置kernel执行参数
int blockSize = 256;
int gridSize = (n + blockSize - 1) / blockSize;
add<<<gridSize, blockSize>>>(d_a, d_b, d_c, n);

// 设备到主机拷贝结果
cudaMemcpy(h_c, d_c, bytes, cudaMemcpyDeviceToHost);

// 清理资源
delete[] h_a; delete[] h_b; delete[] h_c;
cudaFree(d_a); cudaFree(d_b); cudaFree(d_c);

return 0;

}

#include <amp.h>
#include <vector>
#include <iostream>
<p>using namespace concurrency;</p><p>int main() {
const int n = 1024;
std::vector<int> a(n), b(n), c(n);</p><pre class='brush:php;toolbar:false;'>// 初始化数据
for (int i = 0; i < n; ++i) {
    a[i] = i;
    b[i] = i * 2;
}

// 创建array_view（延迟传输）
array_view<const int, 1> *(n, a);
array_view<const int, 1> bv(n, b);
array_view<int, 1> cv(n, c);

// 并行计算
parallel_for_each(cv.extent, [=](index<1> idx) restrict(amp) {
    cv[idx] = *[idx] + bv[idx];
});

// 强制同步获取结果
cv.synchronize();

return 0;

}