cusparseDcsric02 in cusparse


I try to use cusparseDcsric02 in cusparse to do incomplete cholesky decomposition.
But I don’t understand why the input and output matrix value use the same variable “csrValA_valM”,
since the number of non-zero “nnz” is not the same.
It may work for dense matrix, but what if the input is a sparse matrix?
“nnz” must be too small for the output matrix.

can anyone help?
Thank you.

Because it is incomplete cholesky decomposition,
the size of output matrix must be smaller than the input one?