Coarse-Grained Column Agglomeration Parallel Algorithm for LU Factorization Using Multi-Threaded MATLAB

MATLAB programing language is one of the most popular scientific computing tools, especially for solving linear algebra problems. LU factorization is an essential component for the direct solution of linear equations systems. This paper studied a coarse-grained column agglomeration parallel algorith...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Mathematics (Basel) Ročník 13; číslo 2; s. 298
Hlavní autoři: Sabir, Osama, Alebrahim, Reza
Médium: Journal Article
Jazyk:angličtina
Vydáno: Basel MDPI AG 01.01.2025
Témata:
Agglomeration
Algebra
Algorithms
Benchmarks
C plus plus
Central processing units
coarse-grained column agglomeration
Communication
CPUs
Decomposition
Efficiency
Factorization
Gaussian elimination
high-level languages
Linear algebra
Linear equations
Matlab
parallel MATLAB
Parallel processing
Random access memory
Software
ISSN:2227-7390, 2227-7390
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:MATLAB programing language is one of the most popular scientific computing tools, especially for solving linear algebra problems. LU factorization is an essential component for the direct solution of linear equations systems. This paper studied a coarse-grained column agglomeration parallel algorithm in MATLAB to analyze the implementation performance among all the available computation resources. In this paper, we focus on parallelizing the LU decomposition without pivoting algorithm using Gaussian elimination under MATLAB R2020b platform. Numerical experiments were provided to demonstrate the efficiency of CPU parallelization. Performances of the present methods were assessed by comparing the speed and accuracy of different coarse-grained column agglomeration algorithms using different sizes of matrices. Different algorithms were implemented in a four-core Xeon E3-1220 v3 @ 3.10 GHz CPU with 16 GB RAM memory.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2227-7390
2227-7390
DOI:10.3390/math13020298