How to accumulate vector in a CUDA kernel

Asked Apr 25 '17 at 13:56

Active Apr 25 '17 at 15:22

Viewed 232 times

How to add a value to a thrust::vector on the CUDA device ?

I want to do something like (push_back in kernel):

__global__ void genVect( thrust::device_vector<int> & D ) 
{
        int i = threadIdx.x + blockIdx.x * blockDim.x;
        int j = i*i*i;
        D.push_back( j ); // i need this method in kernel
}

int main(void)
{
    thrust::device_vector<int> D;
    genVect<<< 1, 10 >>>( D );
    for(int i = 0; i < D.size(); i++)  std::cout << D[i] << std::endl;
    return 0;
}

ok, i try differently

__global__ void genVect( int * p )
{
    int i = threadIdx.x + blockIdx.x * blockDim.x;
    int j = i*i*i;
    p->push_back( j ); // where is push_back() 
}

int main(void)
{
    thrust::device_vector<int> & D ; 
    int * pD = thrust::raw_pointer_cast(D.data());
    genVect<<< 1, 100 >>>( pD );
    for(int i = 0; i < 100; i++)  std::cout << D[i] << std::endl;
    return 0;
}

edited Apr 25 '17 at 15:22

asked Apr 25 '17 at 13:56

Faik Stack

1

The two word answer is you can't – talonmies Apr 25 '17 at 14:13
not duplicate: i need push_back in CUDA kernel function – Faik Stack Apr 25 '17 at 14:41
1

It is a duplicate. You can't use thrust::device_vector **at all**. Whether you can do `push_back` is irrelevent – talonmies Apr 25 '17 at 14:45
I need to accumulate values for vector in the Cuda kernel. I don't understend how to do it. – Faik Stack Apr 25 '17 at 15:08
http://stackoverflow.com/a/21788662/681865 -- this isn't the question you asked, but probably the answer you want – talonmies Apr 26 '17 at 05:50
talonmies thank you – Faik Stack Apr 26 '17 at 07:05

How to accumulate vector in a CUDA kernel

0 Answers0