代码

#include <cuda_runtime.h>
#include <stdio.h>__global__ void atomicAddAndGet(int *result, int *valueToAdd) {// 原子加法int addedValue = atomicAdd(result, *valueToAdd);// 通过原子操作后读取值，确保是加法后的值addedValue += *valueToAdd;printf("Thread %d: Added value: %d\n", threadIdx.x, addedValue);
}int main() {int result = 0;int valueToAdd = 5;int *d_result, *d_valueToAdd;// 在GPU上分配内存cudaMalloc((void**)&d_result, sizeof(int));cudaMalloc((void**)&d_valueToAdd, sizeof(int));// 将数据从主机内存复制到GPU内存cudaMemcpy(d_result, &result, sizeof(int), cudaMemcpyHostToDevice);cudaMemcpy(d_valueToAdd, &valueToAdd, sizeof(int), cudaMemcpyHostToDevice);// 启动CUDA核函数atomicAddAndGet << <1, 32 >> > (d_result, d_valueToAdd);// 将结果从GPU内存复制回主机内存cudaMemcpy(&result, d_result, sizeof(int), cudaMemcpyDeviceToHost);// 输出结果printf("Result after atomic addition: %d\n", result);// 释放GPU上的内存cudaFree(d_result);cudaFree(d_valueToAdd);return 0;
}

结果