,
SNA'2007

Recursive implementation of high performance numerical algebra library
Authors:Ivan Šimeček, P. Tvrdík
Keywords

High-performance, numerical algebra library.


Abstract

For good performance of every computer program, the good cache utilization is crucial. In numerical algebra libraries (such as BLAS or LAPACK) is good cache utilization achieved by explicit {\it loop restructuring} (mainly loop blocking), but it requires difficult cache behavior analysis. In this paper, we represent recursive implementation of some routines from numerical algebra library. This implementation leads to good cache utilization with no need to analyze cache behavior due to the "natural" partition of data.

 

Download:
final version (in .PDF format)

BibTex entry:
@inproceedings{JA_SNA07_REC,
author = {{\v S}ime{\v c}ek, I. and Tvrd{\'\i}k, P.},
title = {High Performance Recursive Linear Algebra Library},
booktitle = {Seminar on Numerical Analysis},
publisher = {{\' U}stav geonomy AV {\v C}R},
address = {Ostrava},
year = {2007},
pages = {116--119},
ISBN = {80-86407-12-8},
language = {English}
}

BACK