cusparse Ax=B?

Accelerated Computing GPU-Accelerated Libraries
#1

Hello,I want to use cusparse in order to solve Ax=B but I can’t find what function to use from the docs!http://docs.nvidia.com/cuda/cusparse/

Also,because I used cula functions ,for example the function culaSparseCudaDcooCgJacobi does it have an equal in cusparse?

What about preconditions? Like culaSparseJacobiOptionsInit?

Thanks!

#2

cuSparse has that for triangular sparse systems, and it is a two step process.

cusparseScsrsm_analysis

cusparseScsrsm_solve

How large are the matrices? cuBLAS is much easier to use so if you can represent as dense I would go that route.

#3

I don’t want triangular system unfortunately.
The matrices are 128x128 or 512x512.

Thanks

#4

Those are small enough where you really do not need to use cusparse.

Since CULA is a semi-commercial product it has more functionality that the free cuBLAS library.

Check out MAGMA, as it is free and now evidently supports both Windows and linux. That will certainly have the functionality you need.

http://icl.cs.utk.edu/magma/software/view.html?id=192

#5

Ok, thank you!