We describe some activities devoted to extending Parallel Sparse BLAS (PSBLAS), a library of routines providing basic Linear Algebra operations needed to build iterative sparse linear system solvers on distributed-memory parallel computers. We focus on the extension of PSBLAS functionalities for implementing parallel Additive Schwarz preconditioners, that are widely used in the solution of linear systems arising from a variety of applications. An analysis of the performance of PSBLAS-based Additive Schwarz preconditioners with Krylov solvers is reported, that has been carried out on test matrices arising from automotive engine simulations. Some comparisons with Additive Schwarz preconditioners implemented in the well-known PETSc library are also shown.
Extending PSBLAS to Build Parallel Schwarz Preconditioners
D'Ambra Pasqua
2006
Abstract
We describe some activities devoted to extending Parallel Sparse BLAS (PSBLAS), a library of routines providing basic Linear Algebra operations needed to build iterative sparse linear system solvers on distributed-memory parallel computers. We focus on the extension of PSBLAS functionalities for implementing parallel Additive Schwarz preconditioners, that are widely used in the solution of linear systems arising from a variety of applications. An analysis of the performance of PSBLAS-based Additive Schwarz preconditioners with Krylov solvers is reported, that has been carried out on test matrices arising from automotive engine simulations. Some comparisons with Additive Schwarz preconditioners implemented in the well-known PETSc library are also shown.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.