AllScale toolchain pilot applications: PDE based solvers using a parallel development environment
release_6a52uykfr5ertpqwvb5bmts5uy
by
Fearghal O'Donncha, Roman Iakymchuk, Albert Akhriev, Philipp Gschwandtner, Peter Thoman, Thomas heller, Xavi Aguilar, Kiril Dichev, Charles Gillan, Stefano Markidis, Erwin Laure, Emanuele Ragnoli (+4 others)
2018
Abstract
AllScale is a programming environment targeting highly scalable parallel<br>
applications, simplifying application development in the ExaScale era by<br>
siloing development responsibilities. The front-end AllScale API provides a simple C++<br>
development environment and a suite of parallel constructs denoting tasks to be<br>
operated concurrently. Lower level tasks related to the machine and system level are<br>
managed by the AllScale toolchain at the core level. We present the development of two<br>
large-scale parallel applications within the AllScale API, namely, an advection-<br>
diffusion model with data assimilation and a Lagrangian space-weather simulation model<br>
based on a particle-in-cell method. Mathematical formulations and implementations are<br>
presented and we evaluate parallel constructs developed using the AllScale API. The<br>
performance of the applications from the perspective of both parallel scalability and<br>
more importantly productivity are assessed. We demonstrate how the AllScale API can<br>
greatly improve developer productivity while maintaining parallel performance in two<br>
distinct applications. Code complexity metrics demonstrate reduction in application<br>
specific implementations of up to 30% while performance tests on three different<br>
compute systems demonstrate comparable parallel scalability to an MPI version of the<br>
code.
In text/plain
format
Archived Files and Locations
application/pdf 488.3 kB
file_adunyqdeyrbbjo4i5pmpvey6uu
|
zenodo.org (repository) web.archive.org (webarchive) |
article-journal
Stage
published
Date 2018-11-23
access all versions, variants, and formats of this works (eg, pre-prints)
Datacite Metadata (via API)
Worldcat
wikidata.org
CORE.ac.uk
Semantic Scholar
Google Scholar