Routines for LU decomposition (not incomplete LU decomposition)

Hi CUDA users,

I am looking for routines that perform exact LU decomposition (not incomplete LU decomposition as in cusparse***ilu0) on a large sparse matrix, which allow access to the factors L and U matrix.

More specifically, I need U to calculate the determinant of the original sparse matrix.

Any clue to the desired routine? Many thanks!