Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Roland Haas
    @rhaas80
    with typical branch names about as content free as "foo" and "bar" are.
    Roland Haas
    @rhaas80
    having said that, I am not sure if bitbucket would let me enforce the linear history that github's fast-forward only lets me enforce. There does not seem to be a way to restrict what types of merges are allowed.
    Erik Schnetter
    @eschnett
    yes, only allowing merge commit is what i had in mind. the idea is to forbid accidental pushes, and require explicit merges from branches.
    or maybe i should simply check out things via https, that will prevent pushes as well
    but then, maybe i should just renounce write access to the repo, and submit proper pull requests from a fork (which i should do anyway)
    Erik Schnetter
    @eschnett
    what kind of presentations do you want for the august ET workshop? would a presentation on CarpetX be suitable?
    Steven R. Brandt
    @stevenrbrandt
    Yes, a presentation on CarpetX is highly desirable
    Current state of the project, whatever movies exist, how to contribute, etc.
    Steven R. Brandt
    @stevenrbrandt
    Will CarpetX support anything other than 3D?
    Erik Schnetter
    @eschnett
    i answered this in the amrex channel. Yes, in the sense that you can define lower-dimensional grids that are 1 cell wide in the other directions.
    Erik Schnetter
    @eschnett
    CarpetX now supports different number of ghost zones in different directions
    Steven R. Brandt
    @stevenrbrandt
    Great!
    Erik Schnetter
    @eschnett
    note to self: if you don't register evolved variables with MoL, then it won't evolve these variables
    Erik Schnetter
    @eschnett
    my weekend attempt to make MoL and CarpetX was foiled by...
    ... MoL using grid arrays, which are not yet supported by CarpetX.
    Erik Schnetter
    @eschnett
    roland – i corrected some errors in Cartoon2D. these corrections might only be present in the funhpc branch, not on master. (i didn't follow up because i gave up on Cartoon2D, and instead am implementing these symmetries now in McLachlan directly.)
    if you think this is relevant for federico, i can have a closer look.
    Vassili
    @mewesv
    Hi all, I was playing around with MoL over the weekend and came across the following doubt: does CCTK_DELTA_TIME change for some of the RK methods implemented in MoL depending on the intermediate_step?
    Erik Schnetter
    @eschnett
    @mewesv I didn't see your message earlier! do you still have this question?
    actually, the answer is yes, it does. each intermediate step is treated as a first-order step with the respective CCTK_DELTA_TIME.
    Vassili
    @mewesv
    yes, but in some this is done in the respective MoL_Add routine, and in others CCTK_DELTA_TIME is altered.. I was wondering if this was by design or not
    Erik Schnetter
    @eschnett
    yes, this was by design.
    in the CarpetX prototype of MoL, we don't do this any more. cctk_time changes (obviously), but cctk_delta_time remains fixed.
    Vassili
    @mewesv
    thanks!
    Erik Schnetter
    @eschnett
    Quick poll: Who would be interested in a get-started tutorial on CarpetX? It's not production ready yet, but far enough for feedback from science applications.
    Roland Haas
    @rhaas80
    +1
    Roland Haas
    @rhaas80
    I would
    Erik Schnetter
    @eschnett
    my secret ploy to re-invigorate this chat room isn't working yet.
    Roland Haas
    @rhaas80
    I'll bring it up in the ET call today.
    mbabiuc
    @mbabiuc
    Hi!
    Steven R. Brandt
    @stevenrbrandt
    +1
    Bill Gabella
    @gabella
    Yosef and I would like to join a CarpetX tutorial.
    pdiener
    @pdiener
    I would be interested in a tutorial.
    Roland Haas
    @rhaas80
    having such a tutorial on Thursday in the ET call slot (9am Central time) was suggested. Would that work for you, Erik? Thus the time would be limited to ~1hr.
    Roberto De Pietri
    @roberto.depietri_gitlab
    I would be interested inn a tutorial ad to use it.
    yosef-zlochower
    @yosef-zlochower
    I would be interested in the tutorial
    Roland Haas
    @rhaas80
    6 interested persons so far.
    Erik Schnetter
    @eschnett
    that looks promising!
    mbabiuc
    @mbabiuc
    @rhaas80 and @zachetienne I am bringing again the issue of using simfactory on thornyflat. Specifying only --cores and --num-threads is not enough. Is is possible that simfactory defaults on the system configuration? I ended up using something like this:
    ./simfactory/bin/sim create-submit simulation –configuration=<conf> --walltime=<time> --procs=<nodes*ppn> -- ppn-used=<ppn> --num-threads=1 --par=<par>
    mbabiuc
    @mbabiuc
    Here nodes and ppn are specified with the #PBS -l command. This was simfactory does not compain anymore. Not sure if it's right though because have another problem now. My simulation runs on another machine than the one I created the executable. A workaround that I found is to use: ./simfactory/bin/sim setup-silent before ./simfactory/bin/sim create-submit.
    It seems like every node in that cluster has a different name. I already have 10 *wvu.edu.ini in Cactus/simfactory/mdb/machines/ and most likely I'll create more. Is there a workaround if I use simfactory on a cluster?
    Roland Haas
    @rhaas80
    since you mention PBS, I am assuming that thornyflat is a cluster?
    mbabiuc
    @mbabiuc
    Yes
    Roland Haas
    @rhaas80
    In that case simfactory's setup-silent command is not expected to be useful since it is only designed to get peoplesomething usable on their laptops to run the tutorials
    mbabiuc
    @mbabiuc
    Hmm. Without the *ini file though the simulation fails. How to I create it without sim setup?
    Roland Haas
    @rhaas80
    as long as one sets the correct max-num-smt, num-smt and max-num-threads settings in the ini file then even hyperthreading support will work. I won't auto-detect it though (at best it could detect the processor setup on the logn nodes which may not even be the one on the compute nodes if eg the login nodes have more powerfull processors or more sockets).
    usually for clusters I pick one from a "similar" cluster then edit it.
    you could eg take the one for cori (cori.ini) which has hyperthreading turned on. stampede2-knl.ini also has smt settings.
    uhm, well ok. stampede2-knl only sets smt in the copy in my simfactory checkout yet since I did not yet get around to test and push the changes after the most recent hypertrheading discussion in the ET calls.