culaDeviceZgesv alternative

Hi,
I’m using CULA’s culaDeviceZgesv to compute the solution to a real system of linear equations.
I’m looking for a better implementation (mainly because CULA’s implementation does not support
streams).
Anyone can suggest/recommend such an alternative?

thanks
Eyal

Bump… :(

I would also need a linear system solver (possibly for equations with complex coefficients).

What I’m trying to do is to develop an LU decomposition algorithm, see

http://www.noctua-blog.com/index.php/2011/04/21/lu-matrix-decomposition-in-parallel-with-cuda/

and then using the cuBLAS cublastrsm functions to solve an upper/lower triangular linear system. Of course, criticisms to this approach or suggestions are welcome :-)