Iterative Sparse Triangular Solves for Incomplete Factorization Preconditioning
Hartwig Anzt
24 November 2015, 10:30 Salle/Bat : 465/PCRI-N
Contact :
Activités de recherche : High-performance computing
Résumé :
Sparse triangular solvers are typically parallelized using level-scheduling techniques, but parallel efficiency is poor on high-throughput architectures like GPUs. This talk proposes an iterative approach for solving sparse triangular systems when an approximation is sufficient. Although not suitable for all problems, the approach can be successful for sparse triangular matrices arising from incomplete factorizations, where an approximate solution is acceptable. Significant performance gains can be achieved when using this approach for a preconditioned Krylov subspace method.
