SC15 Austin, TX

Analyzing the Performance of a Sparse Matrix Vector Multiply for Extreme Scale Computers

Authors: Amanda Bienz (University of Illinois at Urbana-Champaign), Jon Calhoun (University of Illinois at Urbana-Champaign), Luke Olson (University of Illinois at Urbana-Champaign), Marc Snir (University of Illinois at Urbana-Champaign), William D. Gropp (University of Illinois at Urbana-Champaign)

Abstract: As high-performance computing systems continue to progress towards extreme scale, the scalability of applications becomes critical. The scalability of an algorithm is dependent on interconnect properties, such as latency and bandwidth, and is often limited by network contention. Sparse matrix-vector multiplication (SpMV) is fundamental to a large class of HPC applications. We investigate the performance and scalability of SpMV routines in the widely used software packages PETSc, Hypre, and Trilinos. Through the use of an asynchronous multiplication, we show an improvement in scalability and performance of the SpMV operation when applied to various matrices.

Poster: pdf
Two-page extended abstract: pdf

Poster Index