Roadmap for the Development of a Linear Algebra Library for Exascale Computing: SLATE: Software for Linear Algebra Targeting Exascale
Abdelfattah, Ahmad and Anzt, Hartwig and Bouteiller, Aurelien and Danalis, Anthony and Dongarra, Jack and Gates, Mark and Haidar, Azzam and Kurzak, Jakub and Luszczek, Piotr and Tomov, Stanimire and Wood, Stephen and Wu, Panruo and Yamazaki, and YarKhan
ScaLAPACK was first released in 1995, 22 years ago. To put it in perspective, this was one year after version 1.0 of the MPI standard was released, and two years before the OpenMP Fortran 1.0 specification was released. The fastest machine on the TOP500 list was the Japanese Numerical Wind Tunnel, with peak performance of 235.8 GFLOPS. This was the year when Microsoft acquired NCSA Mosaic to build the Internet Explorer. The past two decades witnessed tectonic shifts in the hardware technology, followed by paradigm shifts in the software technology, and a plethora of algorithmic innovation in scientific computing. At the same time, no viable replacement for ScaLAPACK emerged, that would channel this technological progress into a robust software package. SLATE is meant to be this replacement, and the objective of this document is to provide an overview of the cutting-edge solutions required to accomplish that mission.
Published 2017-09-12 04:00:00 as ut-eecs-17-752 (ID:612)