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/
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
diehlpk has quit [Ping timeout: 260 seconds]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
diehlpk has joined #ste||ar
EverYoung has joined #ste||ar
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
EverYoun_ has quit [Remote host closed the connection]
K-ballo has quit [Quit: K-ballo]
nanashi55 has quit [Ping timeout: 246 seconds]
nanashi55 has joined #ste||ar
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
diehlpk has quit [Remote host closed the connection]
hkaiser has quit [Quit: bye]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
nanashi55 has quit [Ping timeout: 248 seconds]
nanashi55 has joined #ste||ar
parsa has joined #ste||ar
parsa has quit [Client Quit]
jaafar has joined #ste||ar
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
jaafar has quit [Quit: Konversation terminated!]
jaafar has joined #ste||ar
<heller> jbjnr: woohoo, inspect
<jbjnr> someone needs to merge something, so I can test if everything is working again :)
<heller> ;)
<heller> i still think it is overkill to run inspect multiple times
<jbjnr> it is run by each builder before the config/build/test cycle
<jbjnr> doing it like that is easy - doing it in a separate set of scripts is more complicated
<jbjnr> for now I'm happy
<heller> figured
<jbjnr> if you want to improve it, then I'll push my stuff and you can modify it
<heller> what's missing now is the one status per config ;)
<jbjnr> I will add one status (on github) per build now
<jbjnr> easy to do, just change the context string to match the build site/name
<heller> *nod*
<heller> I am working on a better UI now ;)
<jbjnr> waste of time
<heller> I don't think think so
<heller> cdash can be reused
<heller> jbjnr: we should fix the test building first
<heller> Having failed tests which actually work is a bummer
<jbjnr> which tests are you referring to
<heller> The coalescing one
<jbjnr> if it was passing, it would pass
<jbjnr> it fails because it is wrong
<heller> Right
<heller> So we need to fix it
<heller> ctest is just doing a "make all" essentially?
<jbjnr> fixed - I'll push in a mo
<jbjnr> not fixed
<jbjnr> parcel_coalescing doesn't get built at all on my system
<jbjnr> there's no make rule for it
<jbjnr> HPX_WITH_PARCEL_COALESCING is ON
<zao> My build is `ninja components tests.regressions tests.unit tests.performance`
<zao> Especially `components` is omitted if you just build `tests`.
<zao> (with inspiration from the buildbot/circle)
EverYoung has joined #ste||ar
<jbjnr> zao: yup. this is not right. if you want to test (for example) coalescing, then the coalescing lib should be built as part of the test build.
<jbjnr> (I made a PR, but irc didn't show anything)
<zao> Myup.
<jbjnr> zao: I suspect other components/plugins might have the same trouble - I guess that's what our test dashboards are for...
<zao> My project will probably be a bit resourceless for a while, having hardware trouble.
<zao> There's around 1400 EUR of AMD junk under my desk that's not fit for purpose :D
EverYoung has quit [Ping timeout: 255 seconds]
jaafar has quit [Ping timeout: 240 seconds]
<simbergm> heller: I'm aware that my tests don't build, sorry about that... on it now
<heller> simbergm: already fixed ;)
<simbergm> heller: thank you! I have some catching up to do, seems like you all had a very productive weekend!
<jbjnr> error: no member named 'random_shuffle' in namespace 'std'
<jbjnr> grrrr
<heller> jbjnr: err
<heller> jbjnr: I thought I fixed this as well :(
<heller> jbjnr: look for std::shuffle
<jbjnr> "random_shuffle was deprecated in C++14 and completely removed in C++17."
<jbjnr> you might have fixed it - but has it been merged to master?
<jbjnr> if it's in a PR, then we should merge it
<heller> it should have been merged, yes
<heller> which file gives you the error?
<zao> Is there any point in having a proper git checkout when building HPX, compared to say a 'git archive' export?
<heller> zao: history
<zao> Apart from possibly commit ID in diagnostics?
<zao> I currently pipe archive into untar to the build dir, so I get the tree but not the repo.
<zao> Heck, my build image doesn't have the git program.
<zao> If there's value, I can do a shallow reference clone.
<heller> jbjnr: ahh, this hasn't been touched :/
<heller> brb
<jbjnr> ok I'll fix it then
david_pfander has joined #ste||ar
<heller> jbjnr: I need cdash 2.5 ;)
<jbjnr> we will be installing it this week
<github> [hpx] biddisco opened pull request #3067: #ifdef out random_shuffle deprecated in later c++ (master...fix_random_shuffle_error) https://git.io/vbRpj
mcopik_ has joined #ste||ar
mcopik_ has quit [Client Quit]
<jbjnr> hmmm master branch triggers an inspect error, but I suspect it is bogus
<heller> jbjnr: excellent. my UI awesomeness has to wait until then, I guess
<heller> rerurn the inspect and check again
<heller> all coming out of apex though
<jbjnr> might only get 2.
<heller> that might be related
<jbjnr> 2.4
<heller> why?
<jbjnr> 2.5 isn't released yet
<heller> oh
mcopik has quit [Ping timeout: 255 seconds]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
parsa has joined #ste||ar
K-ballo has joined #ste||ar
heller has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
heller has joined #ste||ar
hkaiser has joined #ste||ar
<heller> hmm, those timeouts on inclusive and exclusive scan showing up now don't look good
hkaiser has quit [Quit: bye]
jaafar has joined #ste||ar
<jbjnr> #3066 is safe to merge by the looks of things. the parcel coaelscing test passes with it
<jbjnr> #3067 fixes the random shuffle build error and is safe to merge too
<jbjnr> simbergm: ^
aserio has joined #ste||ar
EverYoung has joined #ste||ar
<github> [hpx] biddisco deleted fix_random_shuffle_error at 04ea962: https://git.io/vb0S7
hkaiser has joined #ste||ar
EverYoung has quit [Ping timeout: 240 seconds]
jaafar has quit [Ping timeout: 255 seconds]
<zao> Ick, Skylake builds tests and core in 32m15.477s
<github> [hpx] msimberg opened pull request #3069: Fix scheduling loop exit (master...fix-scheduling-loop-exit) https://git.io/vb096
<K-ballo> is that bad? :/
<zao> Ryzen was faster.
<zao> Seems to have been around 24m5.850s
<zao> The sorting tests are a bit annoying, takes 25-30s here.
<simbergm> jbjnr: sorry I was a bit slow, seems you already merged them, thank you :)
<simbergm> next up are probably 3051 and 3060
<zao> inclusive/exclusive scans run for a measly 7-8s on my box.
<zao> (on master)
<simbergm> partition test takes 3-4 minutes for me
<zao> 163 - tests.unit.components.distributed.tcp.migrate_component (Timeout)
<zao> 183 - tests.unit.lcos.distributed.tcp.async_cb_remote (Failed)
<zao> 464 - tests.unit.parcelset.distributed.tcp.put_parcels_with_coalescing (Failed)
<zao> jbjnr: That 464 the one I need to explicitly build something for, right?
<jbjnr> simbergm: is your 3056 safe to merge
<jbjnr> seems related to warning s and stuff
<jbjnr> zao: ?
<simbergm> yes, it only removes asserts
<jbjnr> zao: didn't understand your question
<simbergm> / tests
<zao> I'm guessing that tests.unit.parcelset.distributed.tcp.put_parcels_with_coalescing was the one you needed to build some weirdo target for?
<simbergm> probably 3066 fixed that?
<jbjnr> zao: yes - it depends on the coalescng lib that is not built by default, so the test now depends on it and it builds as part of make tests.
<zao> ehm, 'make tests' or the particular test?
<jbjnr> make tests bulds all tests, and that one test now triggers it to be built
<jbjnr> so the rule applies to that one test, but transitivle yot o the tests target
<zao> jbjnr: It's not built for me.
<zao> (I don't built `tests`, I build the three performance|regression|unit targets)
<zao> Not built if I explicitly build 'tests' either.
<zao> HPX_WITH_PARCEL_COALESCING:BOOL=ON
<jbjnr> you pulled from master yes?
<zao> Aye.
<zao> bed37de
<zao> Bah, did I just miss that commit?
<zao> jbjnr: Sorry, I get confused when the github notices don't mention all the events in the repo.
<zao> I'm on an old commit :(
<jbjnr> yup.
<jbjnr> irc is not displaying much about merges and stuff either ...
<zao> Yeah, that's the ones I meant.
<jbjnr> daint just kicked off a tranche of builds, I'll be watching the dashboard for improvements
<heller> simbergm: is #3069 the reason for the timeouts we see?
<heller> jbjnr: the dynamic counter test needs the memory component
eschnett has quit [Quit: eschnett]
<jbjnr> heller: ok. I fix dynamic counter test the same way
<jbjnr> does having plugins built and sitting in the lib dir cause other tests to load them when they might not need to?
EverYoung has joined #ste||ar
<heller> jbjnr: yes
EverYoung has quit [Remote host closed the connection]
<jbjnr> heller: is there anything we can do about that?
EverYoung has joined #ste||ar
<jbjnr> I remember that getting plugins to laod was a PITA, but having them autoloaded all the time, might be worse ...
eschnett has joined #ste||ar
<heller> jbjnr: nothing we can do here
<heller> jbjnr: it's not that bad. Just the startup is a little slower, could be worse
<heller> I think HPX itself doesn't need plugins. hkaiser disagrees
<zao> (when building the _correct_ commit, the dependency indeed holds for the parcel lib thing)
<jbjnr> good
<zao> 100% tests passed, 0 tests failed out of 568
<zao> "something's fucky"
aserio has quit [Quit: aserio]
<heller> zao: or just awesome ;)
aserio has joined #ste||ar
<heller> jbjnr: see pm
<zao> 100% twice. Ship it :P
<hkaiser> heller: the parcel coalescing does not have to be a plugin, indeed
<heller> even more since it is tightly integrated into the core ;)
<hkaiser> heller: as long as we experiment with things, using plugins is nice method, though
<heller> yeah sure
<heller> the problems at hand are rather a build system problem than anything else
<heller> without plugins, we would have seen a linker error instead
<heller> but we are nearing a green dashboard!
<heller> closing in
<heller> rather
<heller> ?
aserio has quit [Quit: aserio]
aserio has joined #ste||ar
<github> [hpx] biddisco opened pull request #3070: Fix dynamic_counters_loaded_1508 test by adding dependency to memory_… (master...fix_dynamic_counters_loaded_1508) https://git.io/vbEfE
<heller> jbjnr: the status for pycicle seems to be broken
diehlpk_work has joined #ste||ar
<jbjnr> heller: will fix asap
EverYoung has quit [Ping timeout: 240 seconds]
<heller> hkaiser: is #3068 clearer now?
<heller> FWIW, I have seen hpx::lcos::condition_variable to be the cause for deadlocks (switching to hpx::lcos::detail::condition_variable fixed those)
<heller> finally got around to write a ticket...
mcopik has joined #ste||ar
<heller> hkaiser: regarding the assert. It's been fired with this test: https://github.com/STEllAR-GROUP/hpx/blob/master/tests/unit/util/pack_traversal_async.cpp#L176
<heller> i say the assert should be there. The test however, is then testing something that's a programming bug in the first place. I think it should be removed. Your take?
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
<zao> A-ha, finally a timeout test, sanity restored. 426 - tests.unit.parallel.executors.thread_pool_executors (Timeout)
akheir has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
Smasher has joined #ste||ar
<hkaiser> heller: yt?
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
aserio has quit [Ping timeout: 240 seconds]
aserio has joined #ste||ar
mcopik has quit [Ping timeout: 240 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
hkaiser has quit [Quit: bye]
EverYoung has joined #ste||ar
parsa has joined #ste||ar
<heller> zao: woohoo!
aserio has quit [Ping timeout: 248 seconds]
<diehlpk_work> Hi, one company is maybe interested in using hpx
<diehlpk_work> heller, yt?
<jbjnr> I'm worried by the fact that 3070 fix dynamic counters failed with a bunch of scan algorithm errors
<jbjnr> something very dubious is afoot.
mcopik has joined #ste||ar
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
hkaiser has joined #ste||ar
<heller> hkaiser: here
<hkaiser> heller: when you were refactoring the AGas components you removed the iterate() functionality from the symbolic_namespace client
<heller> Hmm
<hkaiser> I would like to redo this API and add it back - any objections?
<heller> No
<hkaiser> ok
<diehlpk_work> heller, Do you have time at the end of the week to look at the HPXCL paper?
<heller> I don't remember why I removed it. Not even that I removed it ;)
<heller> diehlpk_work: maybe
<heller> Remind me where to find it again please
<hkaiser> it's still in the symbol namespace server components
<heller> Ok
<hkaiser> just not in the client anymore
akheir has quit [Remote host closed the connection]
<diehlpk_work> Deadline is in 10 days
<heller> Requires a broadcast now, I guess
<hkaiser> but I'd like to change the api away from the callback
akheir has joined #ste||ar
<hkaiser> yes, using broadcast
<heller> For registering names?
parsa has joined #ste||ar
<heller> Or the iteration itself?
<heller> Is there a particular use case you have in mind?
EverYoun_ has quit [Remote host closed the connection]
<heller> hkaiser: another thing I'd like to see in the symbolic namespace is explicit placing of symbols, through '/<locality>/...'
aserio has joined #ste||ar
<zao> jbjnr: Still running concurrent test suites, but https://gist.github.com/zao/8cda93b4da6c64ff897701f9c4a605ff
<zao> Didn't bother making separate Testing directories, so it's possible that things trample on each other a bit.
<zao> (50 concurrent test runs)
<zao> (this is master as of c05035a)
<heller> You should submit to the dashboard
<zao> Feels like it could get a bit noisy with dozens upon dozens of testruns for a particular commit.
<zao> But I don't know how well cdash filters/compacts.
<zao> Doing a test run now to see if I get usable Test.xml out of it.
<jbjnr> if you run multiple sets of tests in the same build dir, the ctest TAG will be overwritten and only one would be valid
<jbjnr> assuming you use ctest -D xxx etc
jaafar has joined #ste||ar
<zao> Yeah, I haven't figured out how to generate/override the tag, so all my builds for a day end up as YYMMDD-hhmm, cycling once a day.
<jbjnr> ctest picks up the TAG from the build dir, so really you should use N build dirs for N concurrent sets of tests.
<jbjnr> I think you might be able to provide a TAG on the cmd line, but I doubt its worth the effort ...
<zao> I've got a proof-of-concept where I bind a separate build/Testing directory into each run, but haven't adapted the scripts to it yet.
<zao> My approach here will probably ultimately be to shove results into a RDBMS and have an API to ask about failure/success/timeout rates of the tests for a commit.
<zao> Or whatever one might care about in a dashboard.
<jbjnr> are you sure you don't want to jusr run pycicle and submit results here?
<jbjnr> lots of hacking work still needed to do cleaning up of dirs etc - just need some basic python skills
<zao> Quite possible, if I manage to secure a solid machine and can shoehorn it into the containers.
<zao> My setup is a bit weird, as I don't have many variants when it comes to compilers and versions.
<zao> What I can do is run a lot of test suites.
<zao> A build of tests and deps takes 35min, so if I have variants the CPU time rapidly exhausts.
<jbjnr> pycicle currently has a=only the option of gcc or clang and on greina, it only builds gcc, so there are not many variants yet implemented.
<zao> Where does it live? Part of `tools` in HPX or is the bulk in some other repo?
<jbjnr> it lives in tools, but only in pycicle brnch so far. in a few more days, I reckon it'll be ready for others to use
<jbjnr> still a bit hacky at the moment and not completeley robust
<zao> Scans are quite prominent indeed.
<zao> (just fails, no timeouts)
* zao implements sleep
david_pfander1 has joined #ste||ar
EverYoung has joined #ste||ar
Smasher has quit [Remote host closed the connection]
K-ballo1 has joined #ste||ar
gentryx_ has joined #ste||ar
hkaiser has quit [Quit: bye]
mcopik has quit [*.net *.split]
K-ballo has quit [*.net *.split]
david_pfander has quit [*.net *.split]
simbergm has quit [*.net *.split]
gentryx has quit [*.net *.split]
david_pfander1 is now known as david_pfander
K-ballo1 is now known as K-ballo
mcopik has joined #ste||ar
simbergm has joined #ste||ar
eschnett has quit [Quit: eschnett]
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
aserio1 has joined #ste||ar
aserio has quit [Ping timeout: 276 seconds]
aserio1 is now known as aserio
hkaiser has joined #ste||ar
akheir has quit [Remote host closed the connection]
aserio1 has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
aserio1 has quit [Ping timeout: 240 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
ct-clmsn has joined #ste||ar
ct-clmsn is now known as Guest39294
Guest39294 has quit [Client Quit]
eschnett has joined #ste||ar
parsa has joined #ste||ar
eschnett has quit [Quit: eschnett]