小编Mar*_*rco的帖子

我有个问题!我需要在cuda c中初始化一个常量全局数组.要初始化数组,我需要使用for!我需要这样做,因为我必须在一些内核中使用这个数组,而我的教授告诉我将其定义为只在设备中可见的常量.

我怎样才能做到这一点？？

我想做这样的事情:

#include <stdio.h>
#include <math.h>
#define N 8

__constant__ double H[N*N];

__global__ void prodotto(double *v, double *w){

        int k=threadIdx.x+blockDim.x*blockIdx.x;

        w[k]=0;
        for(int i=0;i<N;i++) w[k]=w[k]+H[k*N+i]*v[i];
}

int main(){

        double v[8]={1, 1, 1, 1, 1, 1, 1, 1};
        double *dev_v, *dev_w, *w;
        double *host_H;
        host_H=(double*)malloc((N*N)*sizeof(double));
        cudaMalloc((void**)&dev_v,sizeof(double));
        cudaMalloc((void**)&dev_w,sizeof(double));

        for(int k=0;k<N;k++){
            host_H[2*N*k+2*k]=1/1.414;
            host_H[2*N*k+2*k+1]=1/1.414;
            host_H[(2*k+1)*N+2*k]=1/1.414;
            host_H[(2*k+1)+2*k+1]=-1/1.414;

        }

        cudaMemcpyToSymbol(H, host_H, (N*N)*sizeof(double));
        cudaMemcpy(dev_v, v, N*sizeof(double), cudaMemcpyHostToDevice); 
        cudaMemcpy(dev_w, w, N*sizeof(double), cudaMemcpyHostToDevice); 

        prodotto<<<1,N>>>(dev_v, dev_w);

        cudaMemcpy(v, dev_v, N*sizeof(double), cudaMemcpyDeviceToHost); 
        cudaMemcpy(w, dev_w, N*sizeof(double), cudaMemcpyDeviceToHost); 


        for(int i=0;i<N;i++) printf("\n%f   %f", v[i], …