K-ballo changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar.cct.lsu.edu | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | Buildbot: http://rostam.cct.lsu.edu/ | Log: http://irclog.cct.lsu.edu/
diehlpk_work has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
diehlpk_work has quit [Remote host closed the connection]
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
<rachitt_shah[m]> Hey ms I've filled in the form for GSoD, please let me know if you would like to interview me/any other step forward for GSoD.
<rachitt_shah[m]> Thank you, and looking forward to work with you folks for creating some amazing docs!
<ms[m]> srinivasyadav227, hkaiser ci did run, it just didn't show up because there were newer commits after the retest
<hkaiser> ms[m]: I added srinivasyadav227 to the repo, shouldn't the test run without a retest now?
<ms[m]> I've added you to the whitelist now
<ms[m]> probably not if you added him after the latest commits
<ms[m]> they should run whenever he commits though
<hkaiser> ms[m]: is that a jenkins whitelist?
<ms[m]> yeah
<hkaiser> ahh, could you add jedi18[m] as well, pls?
<ms[m]> I added him manually, but you can also comment with "add to whitelist"
<ms[m]> already done
<hkaiser> thanks
<ms[m]> tests should run now automatically, but I wouldn't be surprised if there are kinks, so let me know they still don't run
<srinivasyadav227> okay, sure ;-) thanks!
<ms[m]> rachitt_shah: thanks! we'll let you know soon about next steps
<rachitt_shah[m]> Thank you ms , looking forward to them!
hkaiser has quit [Quit: bye]
<ms[m]> hkaiser: sleepless night? :/
K-ballo has joined #ste||ar
<srinivasyadav227> ms: the tests here https://github.com/STEllAR-GROUP/hpx/pull/5319 got stuck like same as previous `retest`, all jenkins/cscs checks are running but jenkins/lsu tests are not starting
<itn[m]> do this error "Valid buildid required" means the same ?
<srinivasyadav227> itn: yes, i was reffering to "Valid buildid required"
itn[m] is now known as rainmaker6[m]
<rainmaker6[m]> yeah same was the case with me. I thought there was some error in config files which are needed to be fixed so i ignored them.
hkaiser has joined #ste||ar
<rainmaker6[m]> <srinivasyadav227 "ms: the tests here https://githu"> But it's clear for me now.
<srinivasyadav227> rainmaker6: okay, did they get automatically cleared?
<rainmaker6[m]> cleared in the sense ? executed?
<srinivasyadav227> i mean, have you applied any changes or did something to solve that?
<rainmaker6[m]> no
<srinivasyadav227> okay :)
<rainmaker6[m]> basically they sometimes gets triggered and sometimes don't.
<srinivasyadav227> rainmaker6: oh ok alright then, cool, np ;)
joe88 has joined #ste||ar
joe88 has quit [Client Quit]
Girlwithandroid[ has joined #ste||ar
<Girlwithandroid[> Hi! Is the deadline to apply for GSOD over?
<Girlwithandroid[> Can someone help with linking the relevant docs/where to submit the proposal?
<srinivasyadav227> Girl with android: i think this may help you, https://stellar-group.org/2021/04/hpx-accepted-for-google-season-of-docs-2021/ (idk much abt gsod ;) but i got this from previous discussion here)
<Girlwithandroid[> Thank you! This helps :)
<srinivasyadav227> :)
<ms[m]> Girl with android: the link is in that post as well, but here's a direct link to the application form
<Girlwithandroid[> thank you so much!
<ms[m]> srinivasyadav227, rainmaker6, the lsu ci is currently "out of order"
<ms[m]> the guy responsible for the cluster there has been out for some time, but we're hoping to get them back to normal soon
<ms[m]> "invalid build id" usually means "something unrelated to the actual build and tests went wrong"
<srinivasyadav227> ms: okay :)
<hkaiser> ms[m]: Alireza has had a difficult tooth surgery last week....
<hkaiser> sorry for the problems
<ms[m]> hkaiser: no worries! was just wondering about the status, it takes as long as it takes for him to recover :) please tell him not to stress if I'm making it sound like he should stress about it ;9
<ms[m]> hkaiser: ok, just had another look, looks like jobs are running again (the last few days no lsu jobs were running successfully, so I thought it was something else going on)
<ms[m]> the gcc 8 configurations would need some boost modules installed
<rainmaker6[m]> ms noted :)
<rainmaker6[m]> <hkaiser "ms: Alireza has had a difficult "> wishes for an early recovery
<hkaiser> rainmaker6[m]: thanks
hkaiser has quit [Quit: bye]
nanmiao has joined #ste||ar
<zao> HPX is too hard if you break your teeth biting into it ;)
<gonidelis[m]> can't believe that google mismatches us with this stupid tape
<gonidelis[m]> ms[m]: is this the proper place for the iter_sent header? https://github.com/STEllAR-GROUP/hpx/tree/master/libs/core/iterator_support/include/hpx/iterator_support
<gonidelis[m]> or should i put it in the traits folder?
<gonidelis[m]> the iter_sent header
K-ballo has quit [Ping timeout: 252 seconds]
K-ballo has joined #ste||ar
<ms[m]> gonidelis[m]: that was only for tests right? if yes, put it in e.g libs/core/iterator_support/tests/include/hpx/iterator_support/test or similar
<gonidelis[m]> oh ok
<gonidelis[m]> ms[m]: and then how do i include it in the source?
<ms[m]> with an interface target which has a target_include_directories or just a target_include_directories on the test directly (the former is preferable)
<ms[m]> not sure what to call it though, hpx_iterator_support_test is not great, but something in that direction
<gonidelis[m]> the thing is that i will go with a similar solution for the `libs/parallelism` tests
<gonidelis[m]> `libs/parallelism/algorithms`
<gonidelis[m]> ^^
<gonidelis[m]> that's a different module, so that means a different include
<ms[m]> hence the target, you can link to the target in the algorithms tests as well
<gonidelis[m]> but don't we want for the modules to be compilable autonomously ?
<gonidelis[m]> if i create a target in iterator_support the algorithms module which will be using the iter_sent.hpp will be dependent on the iterator_support one
<ms[m]> yep, that's ok
<gonidelis[m]> hm ok
<gonidelis[m]> are we sure that we wanna add an include directory here ?
<ms[m]> where would you like to add it?
<gonidelis[m]> it's not that i have a better suggestion, it's that we create an include dir 3-4 levels deep just for one header
<ms[m]> directories are cheap? I'd at the very least want the module name there, if we end up starting to include test headers from different modules
<gonidelis[m]> ok
<gonidelis[m]> i will go by it then l)
<gonidelis[m]> ;)
hkaiser has joined #ste||ar
<K-ballo> directories aren't all that cheap
<K-ballo> #include needs to stat a lot of directories to find the target most of the time
<K-ballo> even VS's open include file feature takes several seconds to react
<K-ballo> things like putting boost's include path first can have a (small but non negligible) effect on compilation time
diehlpk_work has joined #ste||ar
<diehlpk_work> ms[m], jbjnr Any idea what could be wrong with the mpi parcelport not working with > 64 nodes?
<hkaiser> diehlpk_work: what error do you see?
<diehlpk_work> From 1 to 32 everything works, however for > 32 nodes hpx hnags on startup
<diehlpk_work> I see a print before the hpx main or hpx init but after that brints nothing happens for 10 minutes
<hkaiser> hmmm, Summit?
<diehlpk_work> Yes
<hkaiser> but it works on other machines, doesn't it?
<diehlpk_work> Yes
<diehlpk_work> and no.
<diehlpk_work> We never tested on any other machine with so many localtities
<hkaiser> diehlpk_work: does hello_world work?
<diehlpk_work> hkaiser, Yes
<diehlpk_work> --hpx.lcos.collectives.cut_off=640000
<diehlpk_work> What does this option do?
<hkaiser> not sure, need to look - but it should work without that
<hkaiser> ok, I was not aware of this
<hkaiser> Thomas might remember
<hkaiser> diehlpk_work: the default for the cut_off is ~0x0 (unsigned), so setting any value doesn't make it larger
<diehlpk_work> Ok, I will try to go back to hpx 1.5
<diehlpk_work> Dominic and Sagiv used that verison on QB
<hkaiser> diehlpk_work: let me know if this fixes the issue, please
<diehlpk_work> -DHPX_WITH_MAX_CPU_COUNT=512 \
<diehlpk_work> hkaiser, That is per node or and should not affect the network or?
<hkaiser> this is per node
<hkaiser> not sure why you 512 cores, though
<hkaiser> shouldn't 256 be sufficient?
hkaiser has quit [Quit: bye]
<diehlpk_work> We had it before and Gregor was asking to increase the number
<gdaiss[m]> No, 256 should be sufficient, we would only exceed that if we use one locality per node and all 4 hyperthreads per core (which we don't)
<gdaiss[m]> I am merely confused by Patrick's description of the crashes as Octo-Tiger initialization seems to fail when exceeding 256 localities (we use 6 localities per node). No idea why though
<diehlpk_work> gdaiss[m], It works with 512
<diehlpk_work> but hangs for the IO
hkaiser has joined #ste||ar
<diehlpk_work> hkaiser, Funny using 512 lets the code run further for some attempts
<hkaiser> diehlpk_work: no idea what's wrong
gsodaspirant has joined #ste||ar
<gsodaspirant> Hello, Is the 2021 GSOD position filled?
gsodaspirant has quit [Client Quit]