[Users] Meeting Minutes 2021-10-14
Gabella, William E
b.gabella at Vanderbilt.Edu
Thu Oct 14 10:14:56 CDT 2021
Present: Bill G, Peter D, Roland H, Atul K, Yosef Z, Gabriele B, Steve
B, Leo RW, Miguel G, Maria H, Zach E
Chair: Peter D Minutes: Bill G
* Any first business, no takers.
* next [https://docs.einsteintoolkit.org/et-docs/Release_Process ET
release] [SB, YZ, RH]
We do not have anything new this week. Steve and Yosef did meet. A
month away from freeze. Gallery example runners lined up.
** status update on include candidates:
Roland, has not looked at it since last week and cannot until Tuesday
the following week. Review by 12th of November the Feature Freeze
date. We can add it, but may not be able to compile it...the build
system and missing Boost libraries. So will have it off by default.
Believe code quality is correct and good. Wolfgang said he would add
some dummy documentation so it shows up in the automated thorn docs.
Roland will add to the ticket, for example Leo's stand-alone version.
Leo, I will try to look at that code.
Helvi has this one and is not in the meeting. Roland, can poke her
about this. Part of it uses Python at runtime, links against Python and
uses CFFI package which is not in the default Python install. Hayley
would port some of these problematic portions to Fortran to help with
this. EVen with Python3, you need the build environment
Python3-config. Also need the CFFI module which is not default.
Steve, do we need a way in Cactus to have an external library, and if
pieces are there build it, and if not do not build it. Hopefully this
would be automatic.
Yosef, should we support old clusters? Only some C++ compilers
supported. Seems that Python3 should be available or easily
accessible/buildable on most clusters.
Steve, Roland has a script that generates a thornlist from a *.par file,
maybe more of this should be done. Yosef, maybe add the thornlist to
the gallery example.
Peter, this has been a topic for a long time...handling the thornlist
and the elements that are compiled into the executable.
It is Yosef's to review. Yosef, No updates. Will work on it next week.
** gallery runners
All chosen. Need to double check with Helvi about her students.
BBH] [?] Helvi's students?
BNS] [?] Helvi's students?
scalar wave] [BG]
Roland, should finish up the Simfactory description for Expanse, ticket
#2561. Merge the pull request. Gabriele, current version on Simfactory
works but does not scale well. Have older verions using older MPIs. So
waiting for Admins to install newer versions.
Steve, could use Roland's help for the change for the index for
CarpetX. Roland, we need Erik S to look at it.
* SPEC benchmark contribution [SB, RH]
Steve, Roland we need to change some par files for that. They need
instructions for running longer and wider jobs---few minutes and GB
memory and other is a nodes worth and runs longer. See Ticket #2470,
submit cactuBSSN for SPEC v8 CPU Benchmark,
* LORENE2 as default LORENE thorn in ET [RH]
Roland, nothing new from last week. Will put it in the day after the
release and see what breaks.
* student discussion group about ET
email to list] [RH]
Peter, Roland sent email back to Nick Olsen. Users self help group for
the ET. Nothing new.
question on mailing list]
See Hee Il Kim's thread, no new updates on that. Steve will try to
open tickets sorted by update time]
#2570, update MPI tarball in ExternalLibrary, Roland, had trouble
compiling MPI where IB Verbs/Infiniband exists and a newer version of
MPI exists. Happens rarely. Not urgent. Peter D, does the MPI compile
detect the communication libraries? Roland, think its configure script
does that. Our ExternalLibraries method makes it a bit difficult for
us. Our build does not always know what the MPI configure would have
found to link to.
#2554, init_3_timelevels, Roland, no updates. A few weeks ago could
reproduce it. Yosef, can I help? Roland, I made your example simpler,
smaller grid and few time steps to be able to look at raw data, and
single MPI rank. Looking for where the NANs show up and in which
location. Seem to show up in ghost and buffer zones at the refinement
edge. On time levels not aligned with any coarse time step. So only on
timelevel 1 on the fine level. Related to temporal interpolation and
not just spatial interpolation. Roland, Yosef if you could verify that
that would be helpful. See NANs in timelevel 1 and not time level 2, I
Peter, Other tickets to discuss? No.
tickets ready for review]
#2254, Cactus thorns registration, seem like small request. Roland,
yes. And Yosef suggested a bibtex file as well as formaline's YAML
file. If someone would give it a try, that would be a good project.
#963, Improve McLachlan accuracy, Peter, I have not done anything.
#332, Cactus interp environment variables as configure options. Roland,
has been ready for review...an enhancement. There is a Pull Request.
Needs a reviewer.
#2364, opening and closing HDFS5 files, does have pull request and needs
someone to review it.
* Any update on CarpetX, likely Steve and Roland for reporting. Steve,
it basically works. Sam is looking at performance testing. Steve
working on running in 2D. Seems to be basically working. Roland, more
interested group of post-docs and students meet every Wednesday with
Erik and show what they are working on. Want to add hydrodynamics code
to it. These are people from RIT, LSU, Milan, Amsterdam. They are
making progress. at a stage where people are testing 1D shock in the
Access on a separate repository, uses Cactus but does not use ET right
now. Repo at https://bitbucket.org/eschnett/cactusamrex .
* Any other business?
* Chair and Minute taker for next week?
14 Oct: Chair Steve. Minutes Maria.
28 Oct: Chair Roland. Minutes Miguel
Research Assistant Professor
Department of Physics and Astronomy
Nashville, TN USA
More information about the Users