cuda in c++ project cuda kernel produces no output

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m porting cpu random generator to gpu, it is part of larger c++ project.

I have random.cu file with kernel function and simple function to invoke that kernel from c++ host code

void cudarand(float* rnd)

{

    int* seed=(int*)malloc(5000*sizeof(int));

    for (int i =0 ; i<5000; i++)

    {

	seed[i]=(int)abs(10000000*rand()/ (RAND_MAX + 1.0));

    }

    size_t size = sizeof(rnd);

    int* cuda_seed;

    float* cuda_rnd;

    cudaMalloc(&cuda_seed, sizeof(seed));

    cudaMemcpy(cuda_seed, seed, sizeof(seed), cudaMemcpyHostToDevice);

    cudaMalloc(&cuda_rnd, sizeof(rnd));

    cudarandKernel<<<1,5000>>>(cuda_seed, cuda_rnd);

    cudaMemcpy(rnd, cuda_rnd, sizeof(rnd), cudaMemcpyDeviceToHost);

    cudaFree(cuda_rnd);

}

In host code I include “random.h”, which also is included in random.cu file

random.h is

#ifndef RANDOM_H

#define RANDOM_H

void cudarand(float* rnd);

#endif

From host code I pass array which should be filled by my kernel, but it remains empty. I tried that same kernel in separate project with only one .cu file and it works just perfect. What may be wrong?

Thanks in advance.

2

Fixed. It seems cuda didn’t like sizeof(seed) :)