Dissertation/ Thesis

A Computational study of sparse or structured matrix operations

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: A Computational study of sparse or structured matrix operations
Συγγραφείς: Aimaiti, Nuerrennisahan (Nurgul), University of Lethbridge. Faculty of Arts and Science
Συνεισφορές: Hossain, Shahadat
Στοιχεία εκδότη: Universtiy of Lethbridge, Department of Mathematics and Computer Science
Department of Mathematics and Computer Science
Arts and Science
Έτος έκδοσης: 2018
Συλλογή: University of Lethbridge Institutional Repository
Θεματικοί όροι: Sparse matrices -- Data processing, Java (Computer program language), Algebras, linear, High performance computing, Mathematical optimization -- Data processing, Numerical calculations -- Data processing, sparse data structure, CRS, Compressed Row Storage, JSA, Java Sparse Array, diagonal, BLAS, Basic Linear Algebra Subroutines
Περιγραφή: Matrix computation is an important area in high-performance scientific computing. Major computer manufacturers and vendors typically provide architecture- aware implementation libraries such as Basic Linear Algebra Subroutines (BLAS). In this thesis, we perform an experimental study of a subset of matrix operations, where the matrices are dense, sparse, or structured in Java. We implement a subset of BLAS operations in Java and compare their performance with standard data structures Compressed Row Storage (CRS) and Java Sparse Array (JSA) for dense and sparse structured matrices. The diagonal storage format is shown to be a viable alternative for dense and structured matrices.
Τύπος εγγράφου: thesis
Περιγραφή αρχείου: application/pdf
Γλώσσα: English
Relation: Thesis (University of Lethbridge. Faculty of Arts and Science); https://hdl.handle.net/10133/5268
Διαθεσιμότητα: https://hdl.handle.net/10133/5268
Αριθμός Καταχώρησης: edsbas.FAD7E847
Βάση Δεδομένων: BASE
Περιγραφή
Η περιγραφή δεν είναι διαθέσιμη