Performance Analysis and Optimization of the Tiled Cholesky Factorization on NUMA Machines

Emmanuel Jeannot (1, 2)

1 RUNTIME - Efficient runtime systems for parallel architectures
2 LaBRI - Laboratoire Bordelais de Recherche en Informatique

We discuss some performance issues of the tiled Cholesky factorization on non-uniform memory access-time (NUMA) shared memory machines. We show how to optimize thread placement and data placement in order to achieve performance gain up to 50% compared to state-of-the-art libraries such as Plasma or MKL.

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Emmanuel Jeannot. Performance Analysis and Optimization of the Tiled Cholesky Factorization on NUMA Machines. PAAP 2012 - IEEE International Symposium on Parallel Architectures, Algorithms and Programming, Dec 2012, Taipei, Taiwan. ⟨hal-00772790⟩