Providing Reproducibility for Uncovering Non-deterministic Errors in
Runs on Supercomputers
Description
Reproducibility is highly desirable for parallel applications, but as they are
run on increasingly large and heterogeneous platforms, reproducibility of
numerical results or code behaviors is becoming less and less obtainable. The
same code can produce different results or occasional failures such as a crash
on different systems or even across different runs on the same hardware.
PRUNERS is a research and development project that aims at innovating scalable
tools and techniques to aid applications in obtaining reproducibility.
Specifically, our strategy is to accomplish this by developing a multi-level
analysis and control toolset called the PRUNERS toolset, which combines static
and dynamic analysis techniques to detect, control and eliminate targeted
sources of non-determinism, as introduced through parallel programming
libraries and APIs.
August 2018:
New publications (SWORD and TLC) on OpenMP race dectection techniques have
been added.
April 2017:
Archer, ReMPI, and NINJA have been made available and easily depolyable via
the
Spack package management tool.
June 2015:
“Pruning” Sources of Nondeterminism in Large-Scale Applications, News
Highlights in LLNL Computation and Bits and Bytes, June 2015.
Read more…
July 2014:
Supercomputing Tools Speed Simulations, Research Highlights in LLNL Science
and Technology Review (S&TR), July 2014.
Read more…
