High-Level Strategies for Parallel Shared-Memory Sparse Matrix-Vector Multiplication

The sparse matrix-vector multiplication is an important computational kernel, but is hard to efficiently execute even in the sequential case. The problems--namely low arithmetic intensity, inefficient cache use, and limited memory bandwidth--are magnified as the core count on shared-memory parallel...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	IEEE transactions on parallel and distributed systems 2014-01, Vol.25 (1), p.116-125
Hauptverfasser:	Yzelman, Albert-Jan Nicholas, Roose, Dirk
Format:	Artikel
Sprache:	eng
Schlagworte:	Bandwidth cache-oblivious Computer architecture high-performance computing Hilbert space-filling curve Indexes Kernel matrix reordering NUMA architectures Particle separators shared-memory parallelism Sparse matrices sparse matrix partitioning Sparse matrix-vector multiplication Vectors
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!