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 has quit [Quit: bye]
eschnett has quit [Quit: eschnett]
K-ballo has quit [Quit: K-ballo]
nikunj has quit [Ping timeout: 250 seconds]
david_pfander has joined #ste||ar
nikunj has joined #ste||ar
hkaiser has joined #ste||ar
K-ballo has joined #ste||ar
<hkaiser> simbergm: yt?
<hkaiser> simbergm: I'd like to start merging things
<hkaiser> possibly starting with: #3702 (cuda), #3736 (dataflow fixes)
<hkaiser> this should take care of the compilation problems on various platforms we see
<hkaiser> after that in any order: #3146, #3734, #3761, #3768, #3770, #3774, #3775, #3776
<hkaiser> simbergm: any oppinions?
<jbjnr> hkaiser: any sign of diehlpk_work today?
<hkaiser> sure, give him another hour or so
<jbjnr> ok
<jbjnr> I have new data and new plots, but need the number of grids on level 17
<hkaiser> nod, cool
<hkaiser> jbjnr: did your boss like the news about the LF perf?
<jbjnr> no responses yet. nobody here really cares about my network stuff.
<jbjnr> that's why they don't want me working on it
<hkaiser> they do care, their just to shy to express it
<hkaiser> they're*
<jbjnr> nah.
<jbjnr> anyway. I'm happy and that's what matters to me.
<hkaiser> this important for the LA work you guys are doing
<jbjnr> might be, but we need collectives for that
<hkaiser> jbjnr: that's the next step
<jbjnr> so there's still a lot to do. This is only good for simple stuff send/recv etc
<hkaiser> jbjnr: sure, however octotiger is not really something simple
<hkaiser> anyways, thanks for doing all this
<hkaiser> I'm happy too!
<jbjnr> the good news is that I've really overhauled a lot of the LF work and now I know exactly how to fit the fflib collectives into the existing framework without major breakages
<hkaiser> nice
<hkaiser> let's get the LF into a realease first
<jbjnr> getting the bootup code working with sockets made me need some extra things, now that they're in, I can see how to add the fflib stuff easily.
<jbjnr> Bad news: ....
<hkaiser> and make it easy enough to use
<jbjnr> kevinfs cori runs faile with LF
<jbjnr> kevin's cori runs faile with LF
<jbjnr> ^failed
<hkaiser> ok
<jbjnr> and I can't log in to try to fix it.
<hkaiser> why can't you log in?
<hkaiser> don't havr an account?
<jbjnr> so we will get no data or that. unless we can dig out the last paper's LF runs and plot it
<jbjnr> we had some good data when we did the other paper, but not for so many nodes
<hkaiser> Kevin can definitely add you to the Cori allocation we have
<jbjnr> it's my login that has expired
<hkaiser> get a new one, should not take long
<jbjnr> getting security clearnace agai and all that will take time
<jbjnr> not by wednesday
<jbjnr> I'm too ired
<hkaiser> right
<hkaiser> jbjnr: the data we have is good enough for this paer
<hkaiser> so don't worry too much
<jbjnr> the paper is terrible though
<hkaiser> it isn't terrible, come on
<hkaiser> the results are worth publishing
<jbjnr> it has to be submitted in 2 days and it is only half written. it's a shambles
<hkaiser> stop it, everybody is cooking with water only, so we'll be fine
eschnett has joined #ste||ar
eschnett has quit [Quit: eschnett]
<K-ballo> hkaiser: the card should be on your way now
<hkaiser> K-ballo: ok
<hkaiser> I'll let you know once its here
<K-ballo> thanks
hkaiser has quit [Quit: bye]
<diehlpk_work> jbjnr, I am available
<jbjnr> diehlpk_work: hkyt?
<jbjnr> do either of you know the PI and project details for the CORI account. I want to reset my NIM access and see if I can login and ix libfabric on cori
aserio has joined #ste||ar
hkaiser has joined #ste||ar
<hkaiser> jbjnr: Use Alice Koeniges as PI
hkaiser has quit [Client Quit]
hkaiser has joined #ste||ar
<simbergm> hkaiser: sorry, been away the weekend, back at work tomorrow again
<diehlpk_work> jbjnr, mention XPRESS as the project, and Alice Koniges as the PI.
<simbergm> Go ahead and merge if test are looking decent
<simbergm> tests
<diehlpk_work> jbjnr, Also poke Kevin, since he can approve your request
<jbjnr> diehlpk_work: hkaiser I think I'm ok now. reset my password and configured @FA, seem to be able to access NIM now
<diehlpk_work> Sounds good
<simbergm> Don't remember the pr numbers but if you could wait with moodycamel and the cache line one that would be good
<jbjnr> ok. gotta go now, but managed to log into cori. will try to fix LF tonight
<jbjnr> last thing - diehlpk_work could you send me the path to your 4096 mpi level 16 run slurm.out file, so I can add it to my collection for plotting.
<jbjnr> bye
<diehlpk_work> jbjnr, pdiehl@daint103:/scratch/snx3000/pdiehl/PowerTiger/slurm-12673918.out
aserio1 has joined #ste||ar
aserio has quit [Ping timeout: 250 seconds]
aserio1 is now known as aserio
nikunj has quit [Remote host closed the connection]
nikunj has joined #ste||ar
<nikunj> hkaiser, yt?
nikunj has quit [Read error: Connection reset by peer]
nikunj has joined #ste||ar
<hkaiser> hey nikunj
<nikunj> are you free?
<nikunj> hkaiser ^^
<hkaiser> yes, let's talk now
<nikunj> great let me grab the essentials
aserio1 has joined #ste||ar
<hkaiser> bibek: yt?
<diehlpk_work> jbjnr, Which version of libfabric did you use?
<hkaiser> simbergm: I think the cache-line PR is fine now, it's absolutely crutial as it fixes a couple of things along the lines
aserio has quit [Ping timeout: 268 seconds]
aserio1 is now known as aserio
<nikunj> hkaiser, I'm ready
<nikunj> shall I call?
hkaiser has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
<simbergm> hkaiser: ok, including the timeouts in the resource tests? If not, you can still merge and I'll look at it this week
<hkaiser> simbergm: yes, that PR fixes those as well
<nikunj> hkaiser, I think I've understood the gist of my project. I'll go through the repository in depth and ask questions if I come up with any
<simbergm> hkaiser: ok, awesome and thanks
<hkaiser> nikunj: marvelous!
<diehlpk_work> jbjnr, Same for hpx? Can you provide me with the short hash?
aserio has quit [Ping timeout: 250 seconds]
aserio has joined #ste||ar
zao has quit [Read error: Connection reset by peer]
zao has joined #ste||ar
aserio1 has joined #ste||ar
daissgr has joined #ste||ar
aserio has quit [Ping timeout: 250 seconds]
aserio1 is now known as aserio
<daissgr> aserio: Will the joining by number work this time? :)
<aserio> daissgr: it should
<daissgr> Okay! Currently, it still showing "It is not yet time to join the meeting"
<aserio> ?
<daissgr> one second
<daissgr> we have to move the link to our conference device
<daissgr> not sure why it does not like the join-by-number feature anymore - it was quite handy
<aserio> jbjnr: Will you be joining the call today
aserio1 has joined #ste||ar
aserio has quit [Ping timeout: 258 seconds]
aserio1 is now known as aserio
hkaiser has quit [Quit: bye]
aserio has quit [Ping timeout: 250 seconds]
aserio has joined #ste||ar
hkaiser has joined #ste||ar
daissgr has quit [Ping timeout: 264 seconds]
<K-ballo> we have some weird indentation situation in basic_function.hpp
<K-ballo> some kind of clang-format that misfired?
hkaiser has quit [Quit: bye]
<parsa> K-ballo: i have a curious case for you. compare the output of https://wandbox.org/permlink/E6N2CraHqVz1e1ut and https://wandbox.org/permlink/MLxWcxN8fWJHzOzj
<parsa> no idea what Clang is doing to the first bit of the string
<parsa> after the double
daissgr has joined #ste||ar
nikunj has quit [Quit: Leaving]
aserio has quit [Quit: aserio]
hkaiser has joined #ste||ar
daissgr has quit [Quit: WeeChat 1.9.1]
<diehlpk_work> hkaiser, I combined sub grids per second and speed up in one figure with two y-axis and people are confused
<hkaiser> lol
<diehlpk_work> You can find the figure in the current version
<diehlpk_work> I think we should go back to having them separated
<hkaiser> diehlpk_work: what's the point of having to identical graphs?
<hkaiser> diehlpk_work: but I'd like to leave that to Gregor, I'm not the one having to determine how things should look
<diehlpk_work> Do not plot the speedup or do not have both in log scale
<hkaiser> that's what I was suggesting
<diehlpk_work> I talked to Kevin and we will discuss again in tomorrow's meeting
eschnett has joined #ste||ar
<K-ballo> parsa: looking
<K-ballo> broken libc++ ?