About: Architecture Dependent Linear Code Optimizations

Facets (new session)
Description
Metadata
Settings
- owl:sameAs
- Inference Rule:

About: Architecture Dependent Linear Code Optimizations Goto Sponge NotDistinct Permalink

An Entity of Type : http://linked.opendata.cz/ontology/domain/vavai/Vysledek, within Data Space : linked.opendata.cz associated with source document(s)

Attributes	Values
rdf:type	skos:Concept http://linked.opendata.cz/ontology/domain/vavai/Vysledek
Description	Každý řádně definovaný matematický algoritmus může být relativně snadno implementován v nějakém vyšším prg. jazyce za pomoci knihoven pro LA. Ale takto vzniklé kódy nerespektují vnitřní architekturu CPU a mají nízkou lokalitu dat. Tento nedostatek se dá zmenšit pomocí SW transfromací kódu. V tomto článku demonstrujeme tuto myšlenku na kódech dvou nepřímých řešičů. Také ilustrujeme možnosti nově vyvinuté SW transformace zvané dynamická změna smyslu cyklu (dynamic loop reversal) pro zlepšení využití skryté paměti. (cs) Every mathematical algorithm for the numerical algebra can be relatively easily rewrited into pseudocode: every mathematic operation in the algorithm is simply transformed into one subroutine from the linear algebra package. This simple algorithm leads to very transparent, error-free codes. But these codes do not respect the inner architecture of the CPU and the memory hierarchy and suffer from low temporal and spatial locality. These drawbacks can be reduced by the applications of SW transformation techniques. These transformations result in better temporal and spatial locality or in more efficient utilization of inner pipelines. In this paper, we will demonstrate idea of SW transformation on 2 codes of iterative solvers. We will also represent the new code restructuring transformation called dynamic loop reversal. This transformation offers a new way to improve the cache utilization. Every mathematical algorithm for the numerical algebra can be relatively easily rewrited into pseudocode: every mathematic operation in the algorithm is simply transformed into one subroutine from the linear algebra package. This simple algorithm leads to very transparent, error-free codes. But these codes do not respect the inner architecture of the CPU and the memory hierarchy and suffer from low temporal and spatial locality. These drawbacks can be reduced by the applications of SW transformation techniques. These transformations result in better temporal and spatial locality or in more efficient utilization of inner pipelines. In this paper, we will demonstrate idea of SW transformation on 2 codes of iterative solvers. We will also represent the new code restructuring transformation called dynamic loop reversal. This transformation offers a new way to improve the cache utilization. (en)
Title	Architecture Dependent Linear Code Optimizations Architecture Dependent Linear Code Optimizations (en) Optimalizace lineárních kódů v závislosti na architektuře (cs)
skos:prefLabel	Architecture Dependent Linear Code Optimizations Architecture Dependent Linear Code Optimizations (en) Optimalizace lineárních kódů v závislosti na architektuře (cs)
skos:notation	RIV/68407700:21230/05:03107566!RIV06-MSM-21230___
http://linked.open...avai/riv/aktivita	P Z
http://linked.open...avai/riv/aktivity	P(IBS3086102), Z(MSM6840770014)
http://linked.open...vai/riv/dodaniDat	2006
http://linked.open...aciTvurceVysledku	Šimeček, Ivan Tvrdík, Pavel
http://linked.open.../riv/druhVysledku	A(V) - Audiovizuální tvorba se vzdáleným přístupem
http://linked.open...iv/duvernostUdaju	S - Úplné a pravdivé údaje nepodléhající ochraně podle zvláštních právních předpisů
http://linked.open...titaPredkladatele	České vysoké učení technické v Praze / Fakulta elektrotechnická
http://linked.open...dnocenehoVysledku	513013
http://linked.open...ai/riv/idVysledku	RIV/68407700:21230/05:03107566
http://linked.open...riv/jazykVysledku	eng - angličtina
http://linked.open.../riv/klicovaSlova	Cholesky factorization; Conjugate Gradient algorithm; High-performance; cache hierarchy; dynamic loop reversal; processor architecture (en)
http://linked.open.../riv/klicoveSlovo	dynamic loop reversal High-performance Conjugate Gradient algorithm cache hierarchy processor architecture Cholesky factorization
http://linked.open...i/riv/kodPristupu	L
http://linked.open...ontrolniKodProRIV	[691290ADC249]
http://linked.open...i/riv/mistoVydani	Praha
http://linked.open...n/vavai/riv/nosic	neuvedeno
http://linked.open...in/vavai/riv/obor	IN
http://linked.open...ichTvurcuVysledku	2 (xsd:int)
http://linked.open...cetTvurcuVysledku	2 (xsd:int)
http://linked.open...vavai/riv/projekt	Parallel Algorithms for Large Scale Simulation on PC Clusters
http://linked.open...UplatneniVysledku	2005
http://linked.open...iv/tvurceVysledku	Šimeček, Ivan Tvrdík, Pavel
http://linked.open...n/vavai/riv/zamer	Výzkum perspektivních informačních a komunikačních technologií
https://schema.org/isbn	80-01-03201-9
http://localhost/t...ganizacniJednotka	21230

Faceted Search & Find service v1.16.118 as of Jun 21 2024

Alternative Linked Data Documents: ODE Content Formats:

RDF

ODATA

Microdata

About

OpenLink Virtuoso version 07.20.3240 as of Jun 21 2024, on Linux (x86_64-pc-linux-gnu), Single-Server Edition (126 GB total memory, 112 GB memory in use)
Data on this page belongs to its respective rights holders.
Virtuoso Faceted Browser Copyright © 2009-2024 OpenLink Software