Sparse Linear Algebra: Applying New Features to Traditional Paradigms

View Sample PDF

Copyright: 2023
Pages: 44
Source title: Developing Linear Algebra Codes on Modern Processors: Emerging Research and Opportunities
Source Author(s)/Editor(s): Sandra Catalán Pallarés (Universidad Complutense de Madrid, Spain), Pedro Valero-Lara (Oak Ridge National Laboratory, USA), Leonel Antonio Toledo Díaz (Barcelona Supercomputing Center, Spain)and Rocío Carratalá Sáez (Universidad de Valladolid, Spain)
DOI: 10.4018/978-1-7998-7082-1.ch004

Keywords: Computer Science & IT / High Performance Computing / Information Science Reference / Programming Languages

Purchase

View Sparse Linear Algebra: Applying New Features to Traditional Paradigms on the publisher's website for pricing and purchasing information.

Abstract

This chapter shows several new programming strategies based on tasking to parallelize sparse linear algebra kernels. The reader will explore different approaches to improve the performance of these kernels thanks to a better workload distribution and comprehension of the data layout. This will be accomplished through the study of some of the most popular and widely used sparse operations, such as SpMV (sparse matrix vector multiplication), GTSV (triangular solve), or CG (conjugate gradient). Those strategies have been tested on multicore systems. Some of them equipped GPU devices, showcasing how to overcome the peculiarities of task-based parallelized kernels in the context of sparse linear algebra computations.

The IRMA Community

Research IRM

Sparse Linear Algebra: Applying New Features to Traditional Paradigms

Purchase

Abstract

Related Content

IRMA Sponsors