Present: Roland, Deborah, Steve, Lucas, Peter, Beyhan, Maxwell, Jordan, Noora, Yosef, Zach
Chair: Peter
Minute taker: Deborah
Minutes:
New ETK release done! Kuibit release wrong on download page will be fixed Bitbucket issues Zach been working on figuring out what exactly needs to be done to convert over to github with the help of clankers has simplified the process greatly into just a few steps some trickiness about mapping bitbucket users to github users maybe just list the bitbucket user maybe just email address if that’s possible clankers didn’t have a plan for handling users yet was just going to do names in plain text Zach has changed this to make a user map between bitbucket and github users Accumulated updates after the release CarpetX / AMReX there will be a breaking change to CarpetX which we wanted to do early in the release cycle McLachlan updates being brought back into focus Mailing list questions Jordan about using GetComponents can it be used to update everything to the newest release? Roland: the script doesn’t change branches so GetComponents can’t switch you to a new release typically just create a new Cactus directory with the new release Steve: it’s a feature we could add, in principle, if there is interest Jordan about bug in CarpetRegrid2 has opened a ticket Question about IllinoisGRMHD C2P failure and Psi4 extraction Psi4 aspect has been answered still waiting for help about IllinoisGRMHD part it’s crashing immediately getting nans Zach: in IllinoisGRMHD itself, where that error message appears, give some helpful tips about how to go about debugging turn on poisoning likely an issue of feeding in some data but not all the data that illinois GRMHD needs will reply to the email Open tickets Test Derivs fails on stampede with NaNs Steve reproduced the bug but with the help of clankers may have found a solution Needs more eyes Split the ET right now we have a giant thornlist with basically two sets of thorns (Carpet and CarpetX) could split up the Carpet and the CarpetX thornlists could help with code duplication comes with some tradeoffs Noora encountering an issue with compiling when compiling, it was pulling the Carpet versions instead of the CarpetX versions some of the thorns aren’t part of the current ETK release Maxwell what about splitting them but keeping the name differences? Steve if we get rid of the X in some names, then things like TwoPunctures don’t have to be duplicated Lucas which thorns can be made compatible across drivers? twopunctures seems like an exception at some point, we’re probably going to have a breaking change Roland we can make some level of compatibility with just a little bit of performance loss Steve went through with the help of AI to see what thorns would be easy to make compatible with both Yosef development and code maintenance is mostly a volunteer effort deprecate Carpet once CarpetX can replicate all science proposes that long term compatibility doesn’t need to be an overarching goal Steve agrees that we will have to move on and abandon code eventually needs further discussion Zach solving many of these issues could be done by formalizing Roland’s tool that creates a thornlist from the parameter files more informative error messages would be useful as well Jordan can someone explain the difference between Carpet and CarpetX? Lucas and Zach: it’s very nontrivial. loops are different for example. ccl files differ a bit Zach: much easier with clankers Yosef which thorns are actually compatible with both Steve: shared a markdown file with that Roland: only a few things, like Time his impression is that there’s just a handful of thorns that are the same and most things are different treats them as entirely different codes Peter what’s the timescale for actually getting CarpetX to the point of replacing Carpet for production runs let’s discuss this further in future meetings AOB Steve made a few updates to the ETK tutorial server should run more smoothly now ----------------------------------------- Deborah Ferguson Assistant Professor The University of Rhode Island