The fast solution of sparse linear systems is a very critical problem in many applications.
As of November 2015, the cuSPARSE library offers routines for the solution of sparse linear systems based on LU decomposition, in particular
cusparse<t>csrilu02
and
cusparse<t>csrsv2_solve
Furthermore, cuSPARSE offers the
cusparse<t>csrcolor
which implements graph coloring. The use of graph coloring for incomplete LU-factorization is described in Graph Coloring: More Paral...

More
# Linear Systems

# Solving sparse positive definite linear systems by Cholesky factorization

On our Git Hub website, we have added a fully worked example on how using LU factorization to solve sparse linear systems.

More
# Solving tridiagonal linear systems in CUDA

Tridiagonal linear systems can be easily solved in the framework of the cuSPARSE library by aid of the function:
cusparse<t>gtsv()
cuSPARSE also provides
cusparse<t>gtsv_nopivot()
which, at variance with the first mentioned routine, does not perform pivoting. Both the above functions solve the same linear system with multiple right hand sides. A batched routine
cusparse<t>gtsvStridedBatch()
also exists which solves multiple linear systems.
For all the above routi...

More
# Solving General Sparse Linear Systems in CUDA

One possibility to solve general sparse linear systems in CUDA is using cuSOLVER.
CuSOLVER has three useful routines:
1. cusolverSpDcsrlsvlu, which works for square linear systems (number of unknowns equal to the number of equations) and internally uses sparse LU factorization with partial pivoting;
2. cusolverSpDcsrlsvqr, which works for square linear systems (number of unknowns equal to the number of equations) and internally uses sparse QR factorization;
3. cusolverSpDcsrlsqvqr, w...

More