The acceleration of the sparse matrix-vector multiplication by the region traversal

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F08%3A03151548" target="_blank" >RIV/68407700:21230/08:03151548 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

The acceleration of the sparse matrix-vector multiplication by the region traversal

Popis výsledku v původním jazyce

Sparse matrix-vector multiplication (shortly spMV) is one of most common subroutines in the numerical linear algebra. The problem is that the memory access patterns during the spMV are irregular and the utilization of cache can suffer from low spatial ortemporal locality. This paper introduces new approach for the acceleration the spMV. This approach consists of 3 steps: 1) dividing matrix A into non-empty regions, 2) choosing an efficient way to traverse these regions (in another words choosing an efficient ordering of partial multiplications), 3) choosing the optimal type of storage for each region. In this paper, we describe aspects of these 3 steps in more detail (including fast and time-inexpensive algorithms for all steps). Our measurements proved that our approach gives a significant speedup for almost all matrices arising from various technical areas.

Název v anglickém jazyce

The acceleration of the sparse matrix-vector multiplication by the region traversal

Popis výsledku anglicky

Sparse matrix-vector multiplication (shortly spMV) is one of most common subroutines in the numerical linear algebra. The problem is that the memory access patterns during the spMV are irregular and the utilization of cache can suffer from low spatial ortemporal locality. This paper introduces new approach for the acceleration the spMV. This approach consists of 3 steps: 1) dividing matrix A into non-empty regions, 2) choosing an efficient way to traverse these regions (in another words choosing an efficient ordering of partial multiplications), 3) choosing the optimal type of storage for each region. In this paper, we describe aspects of these 3 steps in more detail (including fast and time-inexpensive algorithms for all steps). Our measurements proved that our approach gives a significant speedup for almost all matrices arising from various technical areas.

Druh

J_x - Nezařazeno - Článek v odborném periodiku (Jimp, Jsc a Jost)

CEP obor

IN - Informatika

OECD FORD obor

—

Projekt

—

Návaznosti

Z - Vyzkumny zamer (s odkazem do CEZ)

Rok uplatnění

2008

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Acta Polytechnica

ISSN

1210-2709

e-ISSN

—

Svazek periodika

48

Číslo periodika v rámci svazku

4/2008

Stát vydavatele periodika

CZ - Česká republika

Počet stran výsledku

8

Strana od-do

—

Kód UT WoS článku

—

EID výsledku v databázi Scopus

—