ICL Research Profile

PLASMA

Parallel Linear Algebra Software for Multicore Architectures

Overview

The Parallel Linear Algebra Software for Multicores and Accelerators (PLASMA) implements a set of fundamental linear algebra routines using the latest updates to the OpenMP standard including advanced tasking and GPU offload pragmas. PLASMA includes, among others, LAPACK-equivalent routines for solving linear systems of equations, linear least square problems, parallel BLAS, and parallel matrix norms.

Over the last decade, PLASMA has been used on various systems using Intel CPUs and coprocessors as well as AMD, IBM POWER, and ARM processors. As a research vehicle, PLASMA is an example of a modern design for new dense linear algebra algorithms. At the same time, PLASMA benefits from the continuous evolution of the OpenMP standard that now includes offloading functionality and enables porting to hardware accelerators. The latest PLASMA release, version 23.8.2 from August 2023, added a tighter integration with vendor libraries, convenience scripts for easier configuration, and other general bug fixes and integration with external projects such as the ECP-sponsored Spack package manager.

Find out more at https://github.com/icl-utk-edu/plasma/

In Collaboration With

  1. University of California Berkeley
  2. University of Colorado Denver

With Support From

  1. Fujitsu
  2. MathWorks
  3. National Science Foundation
  4. The United States Department of Energy

Project Handout