[Users] meeting minutes for 2023-10-12
Roland Haas
rhaas at illinois.edu
Thu Oct 12 10:00:46 CDT 2023
Present: Roland, Steve, Peter, Zach
Testsuites
==========
* Steve compiled code with nvcc and finds that some code tries to use
GPU outside of CarpetX and fails
* need to verify on different machines
* this can be a blocker for the release, needs to be sorted out
ET release
==========
* not updates from GRHayL
* SeedMagneticField is in master, no further updates
* Lucas is working on CarpetX documentation in branch and
documentation.tex
** CarpetX still contains what looks like dead code, hard to tell what
is what though.
** would also need tests on AMD GPU system. Only Forntier system uses
AMD GPUs right now. Maybe solicit help from RIT who is using Frontier
CarpetX release
===============
* worry of not yet fully at quality for ET due to incomplete review
* would favor inclusion as "preview" in release with caveat that API
may change
** would want to include to keep up momentum of using CarpetX
* current only "science" doable with stock CarpetX included in toolkit
is the WaveToy (SpaceTimeX and AsterX are not part of the release)
ET size
=======
Steve brought up large size and time used to compile the full Einstein
Toolkit. Suggests using MakeThornList script to compile only what is
needed for parameter file instead of compiling the full toolkit all the
time.
* Zach agrees that default workflow should migrate that way
* this would not compile many of the more experimental code (eg
CarpetX) for most users and reduce compilation issues
* Steve suggests to, by default, run MakeThornList on some typical
parfiles to produce a "typical" thornlist to compile
* discussed what the exemplar parfiles should be: vacuum, hydro, Canuda
* want "desktop sized" and "cluster sized" parfiles
* also update tutorial notebooks
Unanswered email
================
*
https://lists.einsteintoolkit.org/pipermail/users/2023-October/009083.html
Zach will respond. Looks like out of memory on a personal laptop.
*
https://lists.einsteintoolkit.org/pipermail/users/2023-October/009086.html
Roland will respond. The vtk files are not included the small data
available on Zenodo (says so in the website as well) nor produced by
default by the GW150914.rpar file. To produce them one has to update
the rpar file
Open tickets
============
*
https://bitbucket.org/einsteintoolkit/tickets/issues/2759
twopunctures-swap_xz-results-in-left Zach provided some more details
on the issue described in the ticket. Should have actually rotated
coordinates instead of swapping. Steve requests a pull request with
documentation warning and call to CCTK_VINFO in the code. Roland
argues against using CCTK_VWARN (one output per MPI rank).
Galley runners
==============
claimed: BBH, TOV: Steve
unclaimed: BNS, Scalar Wave, Poisson Equation
Yours,
Roland
--
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.einsteintoolkit.org/pipermail/users/attachments/20231012/e4085ddf/attachment.sig>
More information about the Users
mailing list