Matrix Inversion

I am currently working on the development of several CUDA applications. I am in dire need of a matrix inversion algorithm for non-square complex matrices. Normally this would be done with SVD or equivalent, but as it stands, the BLAS package contains no routines to handle such things.

I do know of a rather lackluster way to invert complex matrices using real matrix inversion algorithms, but I don’t even have those.

So, to get to my question, does anyone have any code that they have written that might be useful, or are there any plans for extending BLAS to include a LAPACK type API?

Many thanks in advance for any help on the matter.

Cheers,
Sean