zgh136608的博客

CUDA global_reduce实现 #include <stdio.h> #include <time.h> #include <stdlib.h> #include <cuda_runtime.h> __global__ void global_reduce(float* d_out, float* d_in) { int myID = blockIdx.x * blockDim.x + threadIdx.x;//得到该线程在全局的线程号 int