Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

DOE Centers of Excellence Performance Portability Meeting

Technical Report ·
DOI:https://doi.org/10.2172/1332474· OSTI ID:1332474
 [1]
  1. Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Performance portability is a phrase often used, but not well understood. The DOE is deploying systems at all of the major facilities across ASCR and ASC that are forcing application developers to confront head-on the challenges of running applications across these diverse systems. With GPU-based systems at the OLCF and LLNL, and Phi-based systems landing at NERSC, ACES (LANL/SNL), and the ALCF – the issue of performance portability is confronting the DOE mission like never before. A new best practice in the DOE is to include “Centers of Excellence” with each major procurement, with a goal of focusing efforts on preparing key applications to be ready for the systems coming to each site, and engaging the vendors directly in a “shared fate” approach to ensuring success. While each COE is necessarily focused on a particular deployment, applications almost invariably must be able to run effectively across the entire DOE HPC ecosystem. This tension between optimizing performance for a particular platform, while still being able to run with acceptable performance wherever the resources are available, is the crux of the challenge we call “performance portability”. This meeting was an opportunity to bring application developers, software providers, and vendors together to discuss this challenge and begin to chart a path forward.
Research Organization:
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Sponsoring Organization:
USDOE
DOE Contract Number:
AC52-07NA27344
OSTI ID:
1332474
Report Number(s):
LLNL-TR--700962
Country of Publication:
United States
Language:
English

Similar Records

How to diagnose problem performance
Journal Article · Thu Sep 01 04:00:00 UTC 1983 · World Oil; (United States) · OSTI ID:5430176

Application of Portable Parallelization Strategies for GPUs on track reconstruction kernels
Conference · Thu Jan 01 00:00:00 UTC 2026 · J.Phys.Conf.Ser. · OSTI ID:3365690

IRIS-BLAS: Towards a Performance Portable and Heterogeneous BLAS Library
Conference · Thu Dec 01 04:00:00 UTC 2022 · OSTI ID:1973340

Related Subjects