aserio 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/
thundergroudon[m has joined #ste||ar
hkaiser has quit [Read error: Connection reset by peer]
autrilla has joined #ste||ar
taeguk[m] has joined #ste||ar
fane_faiz2 has joined #ste||ar
EverYoung has joined #ste||ar
<zao> At least the tests with two localities uses 2x4 threads.
diehlpk has joined #ste||ar
eschnett has quit [Quit: eschnett]
K-ballo has quit [Quit: K-ballo]
hkaiser has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
eschnett has joined #ste||ar
tjtn has joined #ste||ar
diehlpk has quit [Ping timeout: 248 seconds]
tjtn has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
diehlpk has joined #ste||ar
hkaiser has quit [Quit: bye]
diehlpk has quit [Ping timeout: 268 seconds]
tjtn has joined #ste||ar
fane_faiz2 has quit [Ping timeout: 246 seconds]
tjtn has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
fane_faiz2 has joined #ste||ar
diehlpk has joined #ste||ar
fane_faiz2 has quit [Ping timeout: 248 seconds]
parsa has joined #ste||ar
parsa has quit [Client Quit]
nanashi64 has joined #ste||ar
nanashi55 has quit [Ping timeout: 276 seconds]
nanashi64 is now known as nanashi55
parsa has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
jaafar has quit [Ping timeout: 255 seconds]
simbergm has joined #ste||ar
simbergm has quit [Ping timeout: 240 seconds]
simbergm has joined #ste||ar
simbergm has quit [Ping timeout: 248 seconds]
<github> [hpx] biddisco closed pull request #3055: Update documentation for running tests (master...make-tests) https://git.io/vbC91
simbergm has joined #ste||ar
mcopik has quit [Ping timeout: 255 seconds]
<jbjnr> dashboard is improving
david_pfander has joined #ste||ar
<simbergm> jbjnr: do you know why the "Experimental" builds only show test results on cdash?
<jbjnr> the experimental builds are the ones running on the LSU buildbot nihtly testing. I only added the submission of test results to them (a couple of years ago), because doing the update and config steps would have meant rewriting all the buildbot scripts
<jbjnr> those experimental builds have been coming in daily since .....
<heller> jbjnr: doing my best ;)
<jbjnr> you da man!
<simbergm> ah, ok, I hadn't looked much at cdash before
<simbergm> heller: keep 'em coming :)
<jbjnr> Until this week, I was the only person who did
<jbjnr> and I'm not sure if any else is looking now!
<heller> simbergm: regarding the stack size failure, I don't see them
<simbergm> heller: ok, I'm pretty sure now it's a gcc 4.9-only thing
<heller> jbjnr: I find the cdash dashboard still a little unclear. Too much information ;)
<heller> simbergm: can you point me to the failure?
<simbergm> hold on
<jbjnr> TMI? Config / Build/ Test/ too much? really? Are you ill?
<heller> jbjnr: no, the listings of the PRs
<jbjnr> you mean, you only want a single summary - wait till the dashboard goes completely green - then you will like it because a single PR on a single machine will go red and stand out and you will immediately know where to look.\
<heller> jbjnr: 1) Which one is still open? 2) Various builders for the same PR on the dashboard
<jbjnr> currently, it's TMI becuase everything is red
<heller> I mean: I have to hit ctrl-f to find my PR, for example
<jbjnr> can't you read?
<simbergm> now the link on github is pretty good, so you shouldn't need to go through the main dashboard
<jbjnr> yeah - what he said^
<heller> I can, I prefer to have everything in digested form on a dashboard ;)
<heller> would be great if master would be a seperated category as well
<heller> going before the Pull Requests
<jbjnr> I can put master in a separate track
<jbjnr> heller - look now
<heller> jbjnr: splendid!
<jbjnr> from now on, all master branch builds will go to master track.
<heller> what I find a bit confusing also is that the builds that are in process show up as if they were a duplicate of the already completed
<jbjnr> and the answer to your question "which ones are still going" - as each step completes, config/bui;d/test the results are uploaded, so if the space is empty, then it is either still running or cacelled/aborted/died
<heller> so, can we clean the Pull Requests entries once they are closed?
<heller> or at least not have them show up anymore on the dashboard?
<jbjnr> they stay until the end of each day and at miodnight the dashborn is reborn, so if a PR is closed today, tomorrow it is gone - is that what you are asking?
<jbjnr> I cannot wipe existing results off the dashboard
<jbjnr> cdash is open source, so in principle we could tweak it and add that kind of feature if it was worth the effort (it isn't)
<jbjnr> PS. I am going to add the pycicle feature that will check the comments on a PR and if you add "pycicle rebuild" then it will force a build of that PR.
<heller> excellent
<simbergm> jbjnr: how far did you get with jenkins yesterday? did it seem like a dead end?
<heller> that would have been my net question ;)
<heller> what was the outcome?
<simbergm> heller: it wasn't very good, we will most likely never jenkins visible to the outside so we still need some other service to collect results (i.e. cdash)
<heller> that's what I thought
<heller> so a dead end
<simbergm> and then it's just a question if jenkins brings anything that pycicle doesn't
<jbjnr> well, it looks pretty easy to put my build scripts into jenkins and then PRs will be built as they are now and everything will be pretty much the same.
<jbjnr> The jenkins interface itself will no be visible outside CSCS
<jbjnr> but the results can still be submitted to cdash
<heller> so why use jenkins at all then?
<jbjnr> Personally, I would prefer now to just use pycicle - because it is easy to add features and tweak stuff - plus we can easily deploy it at LSU/FAU if we want.
<jbjnr> but a) heller does not like the dashboard
<jbjnr> TMI apparently
<jbjnr> b) hk will want to keep buildbot running and having two will be bad at LSU end
<simbergm> we should discuss this later with hkaiser
<jbjnr> I will run pycicle here and I'm happy. I will let the rest of you deciude what you do next
<jbjnr> if heller wants to run it at FAU -then great
<simbergm> options are basically 1) what we have now with circleci for prs 2) builds at LSU/FAU/CSCS + cdash 3) circleci for everything
<heller> jbjnr: the benfits outweigh the caveats, clearly
<heller> I am happy with what pycicle gives us now. let's use it and improve the stuff when we need it (one can clearly live with the cdash dashboard)
<github> [hpx] msimberg opened pull request #3056: Fix some build warnings caused by unused variables/unnecessary tests (master...fix-build-warnings) https://git.io/vbCFC
<jbjnr> marco say we can upgrade the cdash sever next week btw. we are running 2.2, but a 2.4 is available.
<simbergm> jbjnr: are there significant changes? :)
<heller> ok, any cool new features?
<heller> simbergm: I think circle should still be kept for things like inspect and documentation building. the free service is good enough for that
<heller> simbergm: personally i think, investing in circle-ci is a waste of money since we have soo much unused compute power available
<jbjnr> not much, just some slightly better handling of tracks etc I think, but this has been going since about 2007? and I want to wipe the database and start a fresh one- I have all my paraiew tests results from years back using space needlessley - plut 2 years+ of buildbot nightly results that can be purged.
<simbergm> heller: sure, fair point
<jbjnr> I will add inspect to pycicle
<jbjnr> it can be done as a separate step and then submit a status to github directly without appearing on the dashboard
<jbjnr> or I can make it appear as a test fail
<jbjnr> it can run before the build/etc steps and the dashboard will show it first
<simbergm> jbjnr: and you can show a separate thing on github as well? like build and test now?
<jbjnr> I do not think we can add a column to cdash
<simbergm> ah, and config, so probably yes
<simbergm> right...
<jbjnr> but we could make it set config=fail if inspeact fails
<simbergm> jbjnr: and do you think it's possible to add a link from cdash to the commit on github?
<jbjnr> that's already yhere in the update column
<heller> jbjnr: inspect only needs to run once
<heller> jbjnr: not for all configurations
<jbjnr> it shows you the modified files and you click them and it takes you to github
<simbergm> jbjnr: nice
<jbjnr> heller: irony: it would be easier to add an inspect check to the pycicle script and run it on every PR, than to customize it to only run once! let me think about that - we could just add a separate rule for that checking ...
<simbergm> btw, is it someone's intention to have intel and cuda builds working for the release?
<jbjnr> I intend to setup a build using clang and clang cuda - but I couldn't care less about intel.
<heller> simbergm: CUDA, yes, once pycicle works ;)
<heller> simbergm: CUDA is currently mostly a problem due to buildbot fuckup
<heller> intel, yeah
<heller> it's dead jim, it's dead
<heller> clang is far superior nowadays
<heller> wanna see something cool?
<simbergm> of course ;)
<jbjnr> "once pycicle works" - WTF?
<heller> jbjnr: still some tweaks to do ;)
<heller> jbjnr: and I need to find the time to set it up here
<simbergm> once pycicle is merged into master? :)
<heller> jbjnr: of course, there will be problems once two picycle instances run at the same time
<simbergm> although jbjnr you should probably make an official python package out of it and create some nice config files :)
<jbjnr> I'm going tpo put pyccile into it's own repo at some point, so if Ben and Sam and others decide to use it, it'll need more tweaks to make it handle different "projects"
<simbergm> ... maybe
<jbjnr> ut we can merge iot to hpx initially
<jbjnr> then git filter-branch it out later
<heller> can you access this?
<heller> this is what i've been doing this week ...
<jbjnr> I can
<heller> the solid lines are the important ones
<heller> they started out almost straight flat at the beginning of this week ;)
<heller> observe how awesome clang is...
<jbjnr> indeed it is
<jbjnr> very good
<jbjnr> what is the particle solver?
<jbjnr> aha IPIC3D
<heller> yes
<heller> "space weather" particle in cell
<jbjnr> ok. is it open source?
<heller> earth's magnetic field interacting with solar ions
<heller> yes
<jbjnr> cool. I have a friend who works on that stuff
<jbjnr> and several projects here might be interested in seeing an hpxified (allscale) particle code
<heller> excellent
<heller> the code is developed at KTH
<heller> i'd consider this still to be an early prototype
<heller> the original code is pretty mature though
<heller> yes
<jbjnr> ta
<heller> (allscale uses the resource partitioner btw)
<jbjnr> yay \o/
<jbjnr> simbergm: just out of curiosity I checked the dashboard and it looks like I first added hermione (the old rostam machine) at LSU to the build submission on 2 sept 2014 http://cdash.cscs.ch/index.php?project=HPX&date=2014-09-04
<jbjnr> 3rd actually
<simbergm> jbjnr: it was much greener back then...
<heller> it was...
<heller> around the 1.0 release, we had almost everything green
<jbjnr> ta
<jbjnr> bitbucket - wow.
<github> [hpx] StellarBot pushed 1 new commit to gh-pages: https://git.io/vbWUm
<github> hpx/gh-pages 429a241 StellarBot: Updating docs
zao has quit [Ping timeout: 246 seconds]
Smasher has joined #ste||ar
hkaiser has joined #ste||ar
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
K-ballo has joined #ste||ar
zao has joined #ste||ar
<github> [hpx] msimberg pushed 19 new commits to master: https://git.io/vbWmx
<github> hpx/master fa76925 ct-clmsn: first attempt to isolate sigsegv coroutine overflows from general sigsegvs
<github> hpx/master caff555 ct-clmsn: Merge branch 'master' of https://github.com/STEllAR-GROUP/hpx into segfault-fix
<github> hpx/master 618c292 ct-clmsn: fixed issues with epsilon test
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
mstumpf has joined #ste||ar
<mstumpf> hi! long time no see everybody
<heller> mstumpf: is back!
<hkaiser> hey Martin
<heller> how are you?
<mstumpf> good :) and you guys? how is life, how is the project?
<mstumpf> apart of that, the real reason why i drop in is that I need help, and I hoped the combined c++ knowledge here might come in handy :(
<hkaiser> things are well
<hkaiser> what can we do you for?
<hkaiser> ;)
<mstumpf> here, this is the problem reproduced in a minimal testcase https://pastebin.com/4cMma8Nz
<mstumpf> it only crashes on macos, with homebrew gcc 6.2
<mstumpf> i just don't know if there is something wrong with the code or if it might be a compiler bug
<mstumpf> this is one possible output: https://pastebin.com/Vv2GVm2g
<heller> IIRC, there were some problems with thread_local on macos, don't remember the details though
<mstumpf> K-ballo: you are a c++ standard guru, i feel like. Do you have any first thoughts?
<hkaiser> mstumpf: the code looks ok
<mstumpf> i already posted on stackoverflow and #gcc, nobody had any idea what's going on
<mstumpf> And I miss you guys :) it was a good time. Hope to see all of you again :)
<K-ballo> mstumpf: nothing suspicious with it
<K-ballo> is the unordered_map relevant?
<mstumpf> K-ballo: i didn't test with other datatypes, but the segfault itself is definitely in the destructor of unordered_map
<mstumpf> K-ballo: at one point it just lifelocked because upon deleting the internal datastructure it had a circular linked list that it tried to delete
<mstumpf> i don't know how any list at all can happen in the first place because no data was ever assigned to the unordered_map
<mstumpf> so first guess compiler bug?
<simbergm> jbjnr: do you know what's up with these tests?
<simbergm> ah, do you have a 30 second timeout?
<mstumpf> @everyone was nice talking to you guys again, gotta leave though. Thanks for being my knowledge resource :) wish you all the best
<jbjnr> simbergm: no timeout should be 45s in my cmake scripts, but some tests override it with 100 I seem to recall. Should not be 30
<jbjnr> does look suspicious though
<simbergm> some time out after 30s :(
mstumpf has quit [Quit: Page closed]
<jbjnr> ok, then something is being picked up wrong.
<jbjnr> I'll check
<simbergm> thanks
<jbjnr> aha. found it - my bad
<hkaiser> jbjnr: can you teach pycicle about expected compilation failures?
<hkaiser> those are still flagged as build errors even if they shouldn't
<jbjnr> hkaiser: can you tell me which ones you mean and I'll have a try
<zao> hkaiser: Do you know off-chance if it's possible to disable all form of core pinning in HPX when running outside of SLURM?
<hkaiser> the test's names usually start with fail_
<hkaiser> zao: --hpx:bind=none
<zao> And can I sneak that into the test runner script somehow with CTest?
<zao> https://i.imgur.com/d7dfh8W.png <- only reason cores 9,11,13,15 were used at all was because a second locality was involved in some tests.
<hkaiser> zao: need to look into that, there probably is a way - perhaps a env variable
<hkaiser> gtg now, sorry
<zao> np
hkaiser has quit [Quit: bye]
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
<zao> Bah, master doesn't compile.
<zao> Well, tests.
<K-ballo> that looks like the bind front/back stuff, was it merged already?
<zao> Not sure.
<zao> Haven't kept track on work, might already be fixed in some PR.
<simbergm> I think that broke when heller was fixing some other tests, I think he's aware (or hkaiser)
<K-ballo> uh, seems it was.. sorry about that
<simbergm> heller: ping?
<K-ballo> the PR broke it, not fixed it
<heller> it was me?
<heller> hmmm
<simbergm> maybe not...
<zao> Was this the thing you couldn't repro unless you did C++14, K-ballo?
<K-ballo> zao: yep
<K-ballo> I was looking into it, but it got merged before I got to it
<K-ballo> a couple days ago even
tjtn has joined #ste||ar
aserio has joined #ste||ar
<simbergm> K-ballo: okay, thought it was from some other stuff
<K-ballo> with C++14 the algorithm dispatch machinery is not sfinae-friendly (for diagnostics reasons), and something is inadvertently instantiating those with a const deferred_call/one-shot
<K-ballo> I need to figure out what and fix it, but it's not particularly easy to trace...
tjtn has quit [Client Quit]
<simbergm> sorry, heller, your PR was merged around the same time, thought it broke because of that :(
<simbergm> K-ballo: hkaiser claimed that 3039 would fix those tests
diehlpk has quit [*.net *.split]
Smasher has quit [*.net *.split]
jbjnr has quit [*.net *.split]
parsa[w] has quit [*.net *.split]
auviga has quit [*.net *.split]
zao has quit [*.net *.split]
david_pfander has quit [*.net *.split]
twwright has quit [*.net *.split]
Vir has quit [*.net *.split]
wash[m] has quit [*.net *.split]
K-ballo has quit [*.net *.split]
nanashi55 has quit [*.net *.split]
diehlpk_work has quit [*.net *.split]
daissgr has quit [*.net *.split]
heller has quit [*.net *.split]
<simbergm> but we'll see, and I'm not 100% sure if we were talking about the same thing
taeguk[m] has quit [Ping timeout: 248 seconds]
thundergroudon[m has quit [Ping timeout: 248 seconds]
diehlpk has joined #ste||ar
autrilla has quit [Ping timeout: 255 seconds]
heller has joined #ste||ar
daissgr has joined #ste||ar
wash[m] has joined #ste||ar
Smasher has joined #ste||ar
nanashi55 has joined #ste||ar
K-ballo has joined #ste||ar
zao has joined #ste||ar
david_pfander has joined #ste||ar
auviga has joined #ste||ar
Vir has joined #ste||ar
jbjnr has joined #ste||ar
parsa[w] has joined #ste||ar
twwright has joined #ste||ar
diehlpk_work has joined #ste||ar
tjtn has joined #ste||ar
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
thundergroudon[m has joined #ste||ar
autrilla has joined #ste||ar
taeguk[m] has joined #ste||ar
jaafar has joined #ste||ar
hkaiser has joined #ste||ar
aserio has quit [Quit: aserio]
aserio has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
aserio has joined #ste||ar
autrilla has quit [Ping timeout: 240 seconds]
thundergroudon[m has quit [Ping timeout: 250 seconds]
taeguk[m] has quit [Ping timeout: 252 seconds]
patg[[w]] has joined #ste||ar
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
<aserio> wash[m]: Will you be joining the meeting today?
david_pfander has quit [Ping timeout: 240 seconds]
tjtn has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<zao> I'm not a clever person... made changes to HPX_AddTest.cmake, in a version that doesn't actually compile.
patg[[w]] has quit [Quit: Leaving]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
daissgr1 has joined #ste||ar
thundergroudon[m has joined #ste||ar
mcopik has joined #ste||ar
autrilla has joined #ste||ar
taeguk[m] has joined #ste||ar
taeguk[m] has quit [Remote host closed the connection]
thundergroudon[m has quit [Remote host closed the connection]
autrilla has quit [Remote host closed the connection]
thundergroudon[m has joined #ste||ar
tjtn has joined #ste||ar
<diehlpk_work> zbyerly, Do we know when the OpenSuCO paper will be published?
aserio has quit [Ping timeout: 248 seconds]
autrilla has joined #ste||ar
taeguk[m] has joined #ste||ar
aserio has joined #ste||ar
hkaiser has quit [Quit: bye]
jaafar has quit [Ping timeout: 240 seconds]
hkaiser has joined #ste||ar
hkaiser_ has joined #ste||ar
hkaiser has quit [Ping timeout: 255 seconds]
thundergroudon[m has quit [Ping timeout: 240 seconds]
taeguk[m] has quit [Ping timeout: 252 seconds]
autrilla has quit [Ping timeout: 250 seconds]
jakemp has joined #ste||ar
aserio has quit [Ping timeout: 248 seconds]
hkaiser_ has quit [Quit: bye]
aserio has joined #ste||ar
ct-clmsn has joined #ste||ar
ct-clmsn is now known as ct_clmsn_
ct_clmsn_ is now known as ct_clmsn__
ct_clmsn__ is now known as ct-clmsn
hkaiser has joined #ste||ar
parsa has joined #ste||ar
jbjnr has quit [Remote host closed the connection]
jbjnr has joined #ste||ar
hkaiser has quit [Quit: bye]
aserio has quit [Remote host closed the connection]
aserio has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
ct-clmsn has quit [Ping timeout: 276 seconds]
aserio has quit [Ping timeout: 240 seconds]
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
daissgr1 has quit [*.net *.split]
daissgr has quit [*.net *.split]
heller has quit [*.net *.split]
diehlpk_work has quit [Ping timeout: 240 seconds]
diehlpk_work has joined #ste||ar
daissgr1 has joined #ste||ar
daissgr has joined #ste||ar
heller has joined #ste||ar
mcopik has quit [Remote host closed the connection]
aserio has joined #ste||ar
tjtn has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mcopik has joined #ste||ar
daissgr1 has quit [*.net *.split]
daissgr has quit [*.net *.split]
heller has quit [*.net *.split]
daissgr1 has joined #ste||ar
daissgr has joined #ste||ar
heller has joined #ste||ar
aserio has quit [Quit: aserio]
<diehlpk_work> Hi, with current master I get this error when starting hpx with -t
<diehlpk_work> {what}: unrecognised option '-t': HPX(unhandled_exception)
hkaiser has joined #ste||ar
<zao> The tests that compile for me on master run for me with -t
<zao> Don't have any HPX software of my own to test.
<diehlpk_work> Ok, I will investigate more
<zao> (I may of course be doing things wrong :P)
parsa has joined #ste||ar
tjtn has joined #ste||ar