Production quality Ecosystem for Programming and Executing eXtreme-scale Applications (EPEXA)
EPEXA is an NSF-supported R&D project that will create a
production-quality, general-purpose, community-supported, open-source
software ecosystem that attacks the twin challenges of programmer
productivity and portable performance for advanced scientific
applications on modern high-performance computers. Of special interest
are irregular and sparse applications that are poorly served by current
programming and execution models.
The project addresses central challenges from modern computational
science:
- Advances in predictive, high-fidelity simulation are characterized by
increasingly irregular and dynamic computation (block sparse,
low-rank, mixed representations, etc.).
- Ongoing technology trend in heterogeneous architectures with
dynamically changing performance, and the need to increase concurrency
at all scales.
EPEXA builds upon prior work performed by the TESSE (Task-based
Environment for Scientific Simulation at Extreme Scale) project, and in
particular aims to complete the design and provide robust implementation
of TESSE’s main C++ API, the Template Task Graph (TTG) that provides a
powerful data-flow programming model. Specific project aims include:
- To extend, complete, and harden the successful TESSE research
prototype providing its first production quality implementation using
a community-based, science-driven approach.
- To grow and support the user community, associated applications, and
research use cases.
- To create a community to design, maintain, support and to grow EPEXA
in the future.
- To transform the scalability of key parts of new and existing
numerical simulation codes, including enabling the development of new
DSLs by utilizing the API of the EPEXA runtime, and by migration paths
for both applications and application programmers to follow.
- Robert Harrison
- IACS, Stony Brook University
- multiresolution numerical analysis (MADNESS) with applications to chemistry, material science, nuclear physics, and others.
- dynamic programming algorithms
- Edward Valeev
- George Bosilca and Thomas Herault
- Scott Thornton
- Florian Bischoff
- Wolfgang Bangerth
- Department of Mathematics, Colorado State University
- deal.II open-source finite-element engineering and physics
- Mohammad Mahdi Javanmard
- Poornima Nookala
EPEXA Software Ecosystem
Presentations and other resources
NSF CSSI 2020 PI meeting poster
Funding
The EPEXA project is supported by the National
Science Foundation under grants OAC-1931387 at
Stony Brook University, OAC1931347 at Virginia
Tech, and ACI-1450300 at the University of
Tennesse, Knoxville.
The TESSE project was supported by the National
Science Foundation under grants ACI-1450344 at
Stony Brook University, ACI-1450262 at Virginia
Tech, and OAC-1931384 at the University of
Tennesse, Knoxville.