I have been learning how to use cuda to accelerate sparse matrix operations. And I meet a problem that how to transpose a very large sparse matriix. I use "csr" format to store non-zero values of sparse matrix, but I dont know how to transpose it. Please help me, thank you.
__global__ void transpose_csr(int csrRow[], int csrCol[])//change corresponding value's index for example a[0][1]=>a[1][0]
{
int i = threadIdx.x;
int temp;
temp = csrRow[i];
csrRow[i] = csrCol[i];
csrCol[i] = temp;
}
OddEvenSortGPU(TCSrRowdA, TCSrColndA, TCSrValA, NNzA);
//sort the value by row index
Comments
Post a Comment