Robust Memory-Aware Mappings for Parallel Multifrontal Factorizations
Résumé
We study the memory scalability of the parallel multifrontal factorization of sparse matrices. We illustrate why commonly used mapping strategies (e.g. proportional mapping) cannot achieve a high memory efficiency. We propose a class of "memory-aware' algorithms that aim at maximizing performance under memory constraints. These algorithms provide both accurate memory predictions and a robust solver. We illustrate our approach with experiments performed on large matrices with the MUMPS solver.