[Users] Meeting Minutes 2022-04-07
rhaas at illinois.edu
Thu Apr 7 10:02:36 CDT 2022
I re-ran the code generation for McLachlan, WeylScal4, CTThorns, and
EinsteinExact checked that there is no difference (CTThorns has an
inconsequential change in source code where (-a)*(b) changes to
Baikal when regenerated following the README file does show changes
compared to the version in the ET (wvuthorns), namely the manual commits
* 6b982d7 - WVUThorns Baikal/BaikalVacuum: Correct error message (1 year, 5 months ago) <zetienne>
* 314ad00 - WVUThorns Baikal/BaikalVacuum: GetBoundarySpecification() does not seem to work on certain processor counts. Instead use workaround adopted by GRHydro & Lean to set bndsize when Boundary_SelectVarForBC() for the BSSN constraints (1 year, 5 months ago) <zetienne>
are not present in the nrpytutorial commit used to regenerate Baikal.
> Present: Peter D, Bill G, Zach E, Yosef Z, Roland H, Atul K, Steve B,
> Chair: Peter D, Minutes: Bill G
> * next [https://urldefense.com/v3/__https://docs.einsteintoolkit.org/et-docs/Release_Process__;!!DZ3fjg!tY0RS6oUnjQ4dsLprTv_UO5S5XtSP-A1VRMtn_UViOrahhqTAb7JNucLsZ4alEqc$ ET
> [https://urldefense.com/v3/__https://docs.einsteintoolkit.org/et-docs/Release_Details__;!!DZ3fjg!tY0RS6oUnjQ4dsLprTv_UO5S5XtSP-A1VRMtn_UViOrahhqTAb7JNucLsY4-nhfB$ Details].
> ** The Kranc and NRPy+ codes, regeneration. Roland, will check and run
> make (now). Seems to be working, but with the Readme hash for NRPy+, so
> the older code.
> Zach, looking at WUThorns recently and Baikal has not been updated in a
> long time. Core C code is the same, but different in the Python and
> Codegen side. Also the automation of the CCL files has been improved.
> Because it take a long time for 8th order finite diff, it does a code
> gen in parallel, and Python's GIL does not allow parallelization easily,
> so NRPy clones itself, so when it is finished has to combine the
> results, uses Pickling. Zach found race conditions in the old version of
> Baikal that caused problems in the Codegen of Baikal and these have been
> remedied. Zach, advocates for reviewing the new NRPy+ stuff.
> Roland, we check regeneration just to see it works. You should be able
> to regen it, if one wants. So that if someone types to regen the code
> they get the latest versions. Zach suggests to move on to the newer
> version. A diff will show a lot of differences, Zach moved code between
> .h files to .c files, to speed up the compilation. Does pass the
> unittests that were set up. Zach is using the new version for BBHs, for
> his work. The "new way" is the code to use; the "old way" is still there.
> Zach can write instructions for testing. Zach, new code is in NRPy+
> master and in Jupyter notebooks. Roland, but that is not what the
> Readme files say to do. Zach, could convert the notebooks to Python
> files. Roland, prefers if could cd in Baikal_ETK and it would build
> with a Makefile. There is a one liner that runs the Jupyter by
> converting to Python and running.
> Feature freeze is the 12th of April.
> Decide on a name? Those present voted for Bernhard Riemann as the name
> of the next release. https://urldefense.com/v3/__https://en.wikipedia.org/wiki/Bernhard_Riemann__;!!DZ3fjg!tY0RS6oUnjQ4dsLprTv_UO5S5XtSP-A1VRMtn_UViOrahhqTAb7JNucLsdetOPYV$
> ** Gallery, BBH? Roland, no response from students, APS is in the way.
> Steve could do it, but better if students do it. Steve, the BBH gallery
> is fairly will automated with Python scripts.
> A little early to run the gallery examples. Usual advice, if you have
> never run an example, do it soon. And the run it again closer to and
> then after the Feature Freeze.
> ** FLRW Solver
> Roland, has not tried to compile it, will do that. Needs the Python C
> development files, which may or may not be present on certain clusters.
> Embeds Python interpreter in the compiled C code and runs preprocessed
> code. It might also need some non-standard modules. Python external
> library, maybe. If so difficult, we do not want to include it this round.
> [https://urldefense.com/v3/__https://www.einsteintoolkit.org/tools/unanswered.php__;!!DZ3fjg!tY0RS6oUnjQ4dsLprTv_UO5S5XtSP-A1VRMtn_UViOrahhqTAb7JNucLsZG3NC-b$ unanswered
> question on mailing list]
> Garrison's post about Seg Fault in CarpetReduce, when ghost zones set
> to 1 for reduction. Roland will respond and ask again for a par file.
> open tickets sorted by update time]
> #2601 Add four parameters for carpet, Liwei has added six parameters
> with all x,y,z being treated ecumenically.
> #2392 Roland, default values of keyword parameters not checked.
> Defaults for keywords were not set correctly, just taking the first one
> in a list. Been around for 10+ years. Steve does have fix for this at
> compiles time, is a pull request. That does NOT close the ticket
> because of other issues. Roland, should we keep the runtime checks? If
> we have fixed the compiletime checks for these keyword parameters.
> Steve agrees do not need a runtime check for these...so maybe can close
> this ticket. Steve, wants this PR in. AHfinder is used without seeing
> the AHFinder bug. Roland, it is a level 3 or 4 warning. AHFinder has
> the incorrect defaults; AHFinderDirect is fine, from Jonathan.
> #2518 BlueGeneQ support? Nothing done.
> # 963 Improve McLachlan accuracy
> #2589 Adding Kuibit to the tutorial server. Also a Pull Request to
> update the server used for the tutorial server. Steve and Roland have a
> discussion about the details. Steve will try to review and get it in.
> tickets ready for review]
> # 2364 is an enhancement only, so not a big deal.
> * Any other business
> * Chairs / Minutes
> April 14: Chair Roland, MInutes Atul
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://keys.gnupg.net.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 833 bytes
Desc: OpenPGP digital signature
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20220407/4f524bc7/attachment-0001.bin
More information about the Users