cuSPARSE for solving Ax=b on matrix ~ 230400x230400

Hi sorry for the question, probably it was already discussed.
But I want speed up my application which is solve Ax=b on integer sparse matrices about 230400x230400
Is it real for for CUDA cuSPARSE library?

Currently I use the CPU-based, self-created solver. Which is take A matrix in triplet form, convert it in column compressed form and after apply LU factorization, solve the linear equation.

Thank you.

You might want to take a look at the cusolver sample code that performs a linear system solve on a sparse system:

http://docs.nvidia.com/cuda/cuda-samples/index.html#cusolversp-linear-solver-

You could certainly use cusparse as well. However you would have to build your own solver out of the cusparse pieces

Thanks for the answer.
But I want to completely replace my solver with a csSPARSE.

I.e. convert triplet matrix A to CSC form, and then give to cuSPARSE A, b for solving, and get back the x.

Is it possible for csSPARSE?
And what is really confuse me it’s a size of input data (~ 230400x230400)

Probably exist a more hi-level library-wrappers for cuSPARSE ?

I quick look over the forum topic’s and found out that very often people suggest to use MAGMA, but I can’t found on MAGMA documentation that it suitable for big sparse linear system’s.

I have never used MAGMA, but it does seem to have support for solving sparse systems:

http://icl.cs.utk.edu/projectsfiles/magma/doxygen/sparse-iter.html
The MAGMA Sparse-Iter Package
The sparse-iter package recently added to the MAGMA software stack contains sparse BLAS routines as well as functions to handle the complete iterative solution process of a sparse linear system of equations.

Whether the existing sparse solver support suffices for your use case, I am unable to determine. Here is a short blurb about sparse matrix support in MAGMA, that includes some performance numbers:

http://developer.download.nvidia.com/compute/academia/whitepapers/Achievement2014_UTK.pdf