[Users] Meeting Minutes 2021-09-30
Gabella, William E
b.gabella at Vanderbilt.Edu
Thu Sep 30 09:50:13 CDT 2021
Present: Yosef Z, Bill G, Peter D, Miguel G, Leo RW, Roland H, Atul K,
Helvi W, Gabriele B
Chair: Yosef Z, Minutes: Bill G
* next ET Release [SB, YZ, RH]
[Steve is not present.] No specific updates. Roland, we do need to
"poke" the various items. And look for gallery testers. Yosef, Kuibit
reviewer has no news. Roland, RePriMand reviewer also no news.
* "hack-a-thon" to clean up tickets, documentation, web pages
Yosef, Put on the mailing list for scheduling. Roland, Go through the
tickets and look for easier or harder tickets and go from there. Go
through and fix warnings. Roland, would make a list; Ian hinder tried
to classify as urgent and not so urgent. Code authors could go through
the tickets. Some older components in the toolkit might not have good
question on mailing list]
Aug 13, THC thorn. Roland, we have kept hands off because either a THC
or Kadath question, or both, not in the toolkit. So not for us.
Sep 1, ET and recent intel compilers. Steve seems to be handling this
(last exchange 25 Aug), but we should still document the problem with
the compiler and if one or other is the correct one. Intel compilers on
Bayou system at LSU. Peter has access to intel 21 machines and can try
something, would test suite work? Likely need a BNS parameter file.
Yosef, have older intel at RIT cluster. Atul, there is the BNS gallery
example to use.
open tickets sorted by update time]
No discussion...lots of them.
tickets ready for review]
#963, improve McLachlan accuracy. Peter will take a look at this.
#2568, Roland can make the messages more clear, or document.
#2570, update MPI tarball. Roland two things changed here, update the
tarball, 4.0.1 current version fails to compile on USX (infiniband, and
low level support UCT), 4.0.6 works with both versions. This happened
on Deep Bayou to Roland when installing own MPI. Other change in the
Pull Request, make MPI link against all libraries that HWLOC uses, udev
(?) and NUMA. Usually works on systems with dynamic linking but on
static it can fail.
#2563, isfinite() supersedes finite(). Helvi, a problem on the new Macs
with the M1 chip. Roland will apply it today unless someone seriously
#332, Cactus and environment vars. No discussion.
#2364, avoid HDF5 files opening many times. It is an enhancement.
Roland, put this in for Blue Waters where filesystem was slow. Ticket
is lingering in case anyone wants it/needs it.
* Chair and minute taker
Chair, Roland Minutes, Atul
* any other business
Gabriele B, hack-a-thon suggestion, would be useful to have more param
files that cover more cases; there are so many features in the ET. Did
not find them earlier because no param files showing them. Helvi, keep
in mind they should be consistent on the web site or wherever we keep them.
Roland, more examples of param files, we might want to turn some into
gallery examples to keep them up-to-date, keep them from "aging out" and
no longer working with the toolkit.
Bill, we might also considered a place that is not so well curated, a
"community contributed" site, where users can upload parameters files,
results, and some comments on when and how it ran. A user at your own
risk kind of site. Seems we would get many more param files, but there
would be overlaps, lack of quality control, etc, etc.
Helvi W, been asked to co-organize a workshop for numerical relativity
in Aug 2022, with SpEC, with David Radice, Leo Stein, probably at Brown
university. Mix of few days of hack-a-thon/tutorials, discussions, and
working together. It is an ICERM program, another three weeks for
computational gravity during the semester. Recall, ICERM fall 2020 was
about Computational Relativity, https://icerm.brown.edu/programs/sp-f20/ .
Research Assistant Professor
Department of Physics and Astronomy
Nashville, TN USA
More information about the Users