YaskSite: Stencil Optimization Techniques Applied to Explicit ODE Methods on Modern Architectures

Alappat C, Seiferth J, Hager G, Korch M, Rauber T, Wellein G (2021)

Publication Status: Published

Publication Type: Conference contribution, Conference Contribution

Publication year: 2021

Publisher: Institute of Electrical and Electronics Engineers Inc.

Pages Range: 174-186

Article Number: 9370316

Conference Proceedings Title: CGO 2021 - Proceedings of the 2021 IEEE/ACM International Symposium on Code Generation and Optimization

Event location: Virtual, Korea, KOR

ISBN: 9781728186139

DOI: 10.1109/CGO51591.2021.9370316

Abstract

The landscape of multi-core architectures is growing more complex and diverse. Optimal application performance tuning parameters can vary widely across CPUs, and finding them in a possibly multidimensional parameter search space can be time consuming, expensive and potentially infeasible. In this work, we introduce YaskSite, a tool capable of tackling these challenges for stencil computations. YaskSite is built upon Intel's YASK framework. It combines YASK's flexibility to deal with different target architectures with the Execution-Cache-Memory performance model, which enables identifying optimal performance parameters analytically without the need to run the code. Further we show that YaskSite's features can be exploited by external tuning frameworks to reliably select the most efficient kernel(s) for the application at hand. To demonstrate this, we integrate YaskSite into Offsite, an offline tuner for explicit ordinary differential equation methods, and show that the generated performance predictions are reliable and accurate, leading to considerable performance gains at minimal code generation time and autotuning costs on the latest Intel Cascade Lake and AMD Rome CPUs.

Authors with CRIS profile

Christie Alappat Regionales Rechenzentrum Erlangen (RRZE) Georg Hager Regionales Rechenzentrum Erlangen (RRZE) Gerhard Wellein Professur für Höchstleistungsrechnen

Additional Organisation(s)

Professur für Höchstleistungsrechnen Erlangen National High Performance Computing Center (NHR@FAU)

Related research project(s)

Selbstadaption für zeitschrittbasierte Simulationstechniken auf heterogenen HPC-Systemen (SeASiTe) March 1, 2017 - Feb. 29, 2020

Involved external institutions

Universität Bayreuth DE Germany (DE)

How to cite

APA:

Alappat, C., Seiferth, J., Hager, G., Korch, M., Rauber, T., & Wellein, G. (2021). YaskSite: Stencil Optimization Techniques Applied to Explicit ODE Methods on Modern Architectures. In Jae W. Lee, Mary Lou Soffa, Ayal Zaks (Eds.), CGO 2021 - Proceedings of the 2021 IEEE/ACM International Symposium on Code Generation and Optimization (pp. 174-186). Virtual, Korea, KOR: Institute of Electrical and Electronics Engineers Inc..

MLA:

Alappat, Christie, et al. "YaskSite: Stencil Optimization Techniques Applied to Explicit ODE Methods on Modern Architectures." Proceedings of the 19th IEEE/ACM International Symposium on Code Generation and Optimization, CGO 2021, Virtual, Korea, KOR Ed. Jae W. Lee, Mary Lou Soffa, Ayal Zaks, Institute of Electrical and Electronics Engineers Inc., 2021. 174-186.

BibTeX: Download