Output of cuda program is not what was expected

Question

#include
#include
#include
#include


__global__ void setVal(char **c){

c[(blockIdx.y * gridDim.x         


        

Answer 1

The main problem with your code is that you're not allocating any device memory for the setValues call. You can't pass it a pointer to host memory (char *x[6]) and expect that to work; the CUDA kernels have to operate on CUDA memory. You create that memory, then operate on it, then copy it back:

#include 
#include 
#include 
#include 

__global__ void setValues(char *arr){
    arr[blockIdx.y * gridDim.x + blockIdx.x] = '4';
}

int main() {
    const int NCHARS=6;
    char *xd;

    cudaMalloc(&xd, NCHARS);
    dim3 grid(3,2);
    setValues<<>>(xd);

    char *p;
    p = (char*) malloc(20*sizeof(char));
    strcpy(p,"");

    cudaMemcpy(p, xd, NCHARS, cudaMemcpyDeviceToHost);
    p[NCHARS]='\0';

    printf("<%s>\n", p);
    getchar();

    cudaFree(xd);

    return 0;
}