IDEAS-EXAALT-SIAM-Poster.pdf (2.52 MB)
IDEAS-EXAALT Collaboration: Adopting Continuous Integration for Long-Timescale Materials Simulation
poster
posted on 2019-02-26, 16:29 authored by Richard Zamora, Christoph Junghans, David MoultonDavid MoultonAs leadership-scale computing systems continue to grow in both size and
complexity, so do the scientific software applications that are designed
to use them. In order to leverage extreme scales efficiently, many
modern applications are composed of a collection of independent packages
and libraries. For these projects, the necessary implementation of
sustainable software practices requires developers to navigate multiple
code bases. One promising answer to this challenge is the Productivity
and Sustainability Improvement Planning (PSIP) methodology being
developed by the Interoperable Design of Extreme-scale Application
Software (IDEAS) project. The PSIP methodology promotes the clear
factoring of new software processes and capabilities into a manageable
number of critical steps with simple completion criteria. The Exascale
Atomistic Capability for Accuracy, Length and Time (EXAALT), currently
developed under the US-DOE Exascale Computing Project (ECP), is an
excellent example of a complex software framework in which PSIP has
proven valuable. Specifically, EXAALT is comprised of three
sub-projects (ParSplice, LAMMPS, and LATTE) with different physics, each
with its own development processes and dependencies. In this work, we
highlight a recent effort to implement an end-to-end
continuous-integration pipeline within the EXAALT project repository to
demonstrate the advantages of PSIP-based software development.