hkaiser 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/
<hkaiser> K-ballo: yah, I think so
K-ballo has quit [Quit: K-ballo]
hkaiser has quit [Quit: bye]
nikunj has joined #ste||ar
heller has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
heller has joined #ste||ar
<jbjnr> diehlpk_work: I've pushed scripts/data and also software versions
eschnett has quit [Quit: eschnett]
K-ballo has joined #ste||ar
K-ballo has quit [Client Quit]
K-ballo has joined #ste||ar
nikunj has quit [Quit: Leaving]
hkaiser has joined #ste||ar
eschnett has joined #ste||ar
eschnett has quit [Quit: eschnett]
aserio has joined #ste||ar
<aserio> hkaiser: ty?
<aserio> yt?**
<zao> K-ballo: Ick.
<hkaiser> aserio: here
<aserio> hkaiser: please see pm
eschnett has joined #ste||ar
<parsa> K-ballo: any easy workarounds to get the code working?
<K-ballo> operate line by line instead? read the whole line into a string stream
<parsa> i thought of using scanf, but i don't know if that has any side-effects on the stream
<parsa> so stringstream is not affected by this bug?
<zao> A stringstream constructed from a line of text ends at the end of the line, avoiding it considering anything on the next line?
<K-ballo> it is, the formatting functionality of iostreams is shared
<K-ballo> right, operating on isolated lines stops the stream from parsing "too far"
hkaiser has quit [Quit: bye]
<parsa> K-ballo: works. thank you!
hkaiser has joined #ste||ar
<aserio> hkaiser: Don't forget about the FleCSI Meeting!
<hkaiser> aserio: I won't ;-)
<hkaiser> thanks!
<hkaiser> aserio: btw, bibel got 20% overall speedup when running on that branch compared to old master
<hkaiser> Bibek*
<hkaiser> (sorru for butchering your name bibek)
<hkaiser> simbergm: see above, the cache-line changes give us 20% speedup on high contention scenarios
<hkaiser> in Phylanx
<aserio> hkaiser: wow!
<aserio> hkaiser, bibek: what about the 16 thread case?
jaafar has quit [Ping timeout: 250 seconds]
<simbergm> hkaiser: nice job! :D
<simbergm> jbjnr: I can't for the life of me figure out what's wrong with the moodycamel branch
<simbergm> if you're motivated thread_pool_executors_test fails usually very quickly (hangs)
<simbergm> I know that it's not caused by the moodycamel queue only, not by your other changes to how tasks/threads are constructed
<simbergm> at some point an important thread gets scheduled but but then it disappears somewhere
<simbergm> i.e. doesn't actually get run
<simbergm> I
<simbergm> I'm wondering if it has anything to do with the (non-)guarantees it provides: https://github.com/cameron314/concurrentqueue#reasons-not-to-use
<simbergm> shouldn't, but I don't know...
<K-ballo> we should cherry pick that fix to tuple
<bibek> aserio: it was the same, around 20%
<aserio> bibek: so still much worse but about 20% better?
<diehlpk_work> jbjnr, Can you script extract the total time?
<bibek> the height of the skyscrappers have also decreased, its not slower than the 1 thread case anymore but still we need to do more.
<bibek> aserio
<aserio> wow!
<bibek> ill show you the plots when you are back
<simbergm> K-ballo yeah, good point
aserio has quit [Ping timeout: 255 seconds]
<simbergm> jbjnr: you might like to know that that same test doesn't seem to hang with your scheduler, but I'm guessing it's just luck... ;)
jaafar has joined #ste||ar
jaafar_ has joined #ste||ar
jaafar has quit [Read error: Connection reset by peer]
eschnett has quit [Quit: eschnett]
daissgr has joined #ste||ar
aserio has joined #ste||ar
<parsa> hkaiser: ping
<parsa> jbjnr: ping
eschnett has joined #ste||ar
eschnett has quit [Quit: eschnett]
<jbjnr> parsa: ??
<hkaiser> jbjnr: ob meeting
eschnett has joined #ste||ar
<hkaiser> paper
<jbjnr> sorry was out for a walk in the rain
<jbjnr> I'm too tired to think about the paper any more
<hkaiser> k
<jbjnr> what do you need from me?
<jbjnr> Got family in room here, so can't join call
<hkaiser> you need to finsihed the LF writeup and description of graphs
<jbjnr> I don't know what to say - got any ideas?
<jbjnr> I thought the network part would only be a small fraction of the paper
<hkaiser> what for? the graphs
<hkaiser> ?
<hkaiser> at least the graphs need explanations
<jbjnr> yeah - what to say in the scaling part of the paper
<jbjnr> I thought I wrote that. I'll look now
Vir has joined #ste||ar
<jbjnr> ps hkaiser tell diehlpk_work that we can extract anything we want with the scripts
<hkaiser> jbjnr: ack, he read this just now
<jbjnr> (assuming the numbers are in the slurm.out file)
eschnett has quit [Quit: eschnett]
<hkaiser> aserio: yt?
<aserio> hkaiser: yep
<hkaiser> could you schedule another OB paper meeting for tomorrow same time (1.30pm), please?
<hkaiser> aserio ^^
<aserio> Sure :)
<hkaiser> thanks!
hkaiser has quit [Quit: bye]
aserio has quit [Quit: aserio]
jaafar_ is now known as jaafar
daissgr has quit [Ping timeout: 252 seconds]
<diehlpk_work> jbjnr, Sorry for removing your stuff
aserio has joined #ste||ar
aserio has quit [Quit: aserio]
daissgr has joined #ste||ar
<jbjnr> diehlpk_work: wasted my whole night with getting node level scripts to run, will try to recover text tomorrow, though I don't care any more and you can just leave it deleted.
hkaiser has joined #ste||ar
daissgr has quit [Ping timeout: 258 seconds]
jaafar has quit [Ping timeout: 250 seconds]