NUMA-Aware Strategies for the Heterogeneous Execution of SPMV on Modern Supercomputers

Cita com:
hdl:2117/366919
Document typeConference report
Defense date2021
Rights accessOpen Access
Except where otherwise noted, content on this work
is licensed under a Creative Commons license
:
Attribution-NonCommercial-ShareAlike 3.0 Generic
Abstract
The sparse matrix-vector product is a widespread operation amongst the scientific computing community. It represents the dominant computational cost in many large-scale simulations relying on iterative methods, and its performance is sensitive to the sparse pattern, the storage format, and kernel implementation, and the target computing architecture. In this work, we are devoted to the efficient execution of the sparse matrix-vector product on (potentially hybrid) modern supercomputers with non-uniform memory access configurations. A hierarchical parallel implementation is proposed to minimize the number of processes participating in distributed-memory parallelization. As a result, a single process per computing node is enough to engage all its hardware and ensure efficient memory access on manycore platforms. The benefits of this approach have been demonstrated on up to 9,600 cores of MareNostrum 4 supercomputer, at Barcelona Supercomputing Center.
CitationAlvarez, X. [et al.]. NUMA-Aware Strategies for the Heterogeneous Execution of SPMV on Modern Supercomputers. A: European Congress on Computational Methods in Applied Sciences and Engineering. "14th WCCM-ECCOMAS Congress 2020: collection of papers presented at the 14th edition of the WCCM-ECCOMAS, virtual congress, January, 11-15, 2021". 2021, p. 1-10. ISBN 978-84-121101-7-3. DOI 10.23967/wccm-eccomas.2020.223.
ISBN978-84-121101-7-3
Other identifiershttps://www.scipedia.com/public/Alvarez-Farre_et_al_2021a
Files | Description | Size | Format | View |
---|---|---|---|---|
Draft_Content_721176048p5653.pdf | 256,0Kb | View/Open |