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/ | GSoC: https://github.com/STEllAR-GROUP/hpx/wiki/Google-Summer-of-Code-%28GSoC%29-2020
<Yorlik> Allright. I'm reading / digging / coding n the meantime :)
<bita> hkaiser, do you have a sec?
<hkaiser> sure
<bita> Usually in blaze we have start_row, start_col, size_row, size_col
<bita> Is it Okay if I change it?
<hkaiser> yes, please
<bita> thank you
<hkaiser> bita: please be careful not to miss any spots ;-)
<bita> I will
<hkaiser> well, thank _you_!
<hkaiser> bita: does distribute_vector needs changing as well?
<bita> no, it has only one dimension
<hkaiser> right
<bita> so it is fine :)
bita has quit [Quit: Leaving]
hkaiser has quit [Quit: bye]
wate123_Jun has quit [Remote host closed the connection]
akheir has quit [Quit: Leaving]
diehlpk_work_ has quit [Read error: Connection reset by peer]
diehlpk_work_ has joined #ste||ar
nikunj97 has joined #ste||ar
Hashmi has joined #ste||ar
<nikunj97> simbergm, is there any other way to visualize the trace output? Vampir isn't free to use and the demo version can't do more than 16 thread traces
karame_ has quit [Remote host closed the connection]
Hashmi has quit [Quit: Connection closed for inactivity]
nikunj97 has quit [Quit: Leaving]
<simbergm> nikunj97: good question
<simbergm> I don't know
<simbergm> khuck might know of alternatives, as might jbjnr
gonidelis has joined #ste||ar
gonidelis has quit [Ping timeout: 240 seconds]
gonidelis has joined #ste||ar
nikunj97 has joined #ste||ar
<nikunj97> simbergm, alright let me ask around
Hashmi has joined #ste||ar
<nikunj97> jbjnr, yt?
<heller1> There used to be one...
<heller1> But I forgot the name :(
<nikunj97> heller1, :/
<heller1> Kate Isaac wrote it....
<nikunj97> isn't there any other profiler that I can use with HPX?
<nikunj97> I have used vTune in the past, but I'm working on ARM
<heller1> You can always use perf
<nikunj97> this looks good
<nikunj97> I was thinking to use arm performance reports
<nikunj97> it's also available on the cluster
<nikunj97> btw why is Vampir so expensive!?
gonidelis has quit [Ping timeout: 240 seconds]
<heller1> Ask them
<heller1> I'm sure jsc has an installation of it somewhere
<heller1> Maybe in some module?
<nikunj97> heller1, I didn't see Vampir on my cluster
<heller1> But yeah, it's no fun using it over ssh
<nikunj97> may be on some other cluster
<nikunj97> I'll ask them if I could get a license or access to it as a module
<heller1> Yep
<heller1> Ravel/traveler might just work for you as well
<nikunj97> I just got access to Vampir through one of their head node
<nikunj97> I emailed him about the problems I was facing wrt Vampir. He was kind enough to provide me another cluster access :)
<nikunj97> people at JSC are really kind especially the prof I'm working with
Hashmi has quit [Quit: Connection closed for inactivity]
hkaiser has joined #ste||ar
nikunj97 has quit [Quit: Leaving]
nikunj97 has joined #ste||ar
<nikunj97> vampir with X11 forwarding is very very slow
<hkaiser> x11 forward is very slow
<nikunj97> hkaiser, yes definitely. I wish I had a license of vampir myself
<heller1> see if you can get access to a proper remote desktop
<hkaiser> nikunj97: talk to Kevin, he might be able to help
<nikunj97> is Kevin on irc?
<heller1> did you try out ravel/traveller?
<nikunj97> no I couldn't
<hkaiser> nikunj97: send an email
<nikunj97> I don't have his email id
<hkaiser> see pm
<nikunj97> hkaiser, thanks!
<nikunj97> I'll CC you in the email
<hkaiser> yes, please do
<nikunj97> hkaiser, I've sent him an email. Let's see what he has to say
nikunj97 has quit [Remote host closed the connection]
nikunj97 has joined #ste||ar
<nikunj97> heller1, I can't find ravel traveler through google search
<nikunj97> could you provide me it's link?
<heller1> nikunj97: i gave you the github link earlier (kisaacs)
<nikunj97> aah that one
<nikunj97> found it, my bad
diehlpk_mobile[m has quit [Ping timeout: 246 seconds]
diehlpk_mobile[m has joined #ste||ar
diehlpk_mobile[m has quit [Ping timeout: 246 seconds]
diehlpk_mobile[m has joined #ste||ar
<nikunj97> can anyone explain how to use vncserver?
<nikunj97> I'm getting a time out or host not found
<nikunj97> on my cluster node: I do `vncserver :1`
<nikunj97> which gives the following output:
<nikunj97> Desktop 'TurboVNC: juwelsvis02.ib.juwels.fzj.de:1 (gupta2)' started on display juwelsvis02.ib.juwels.fzj.de:1 (juwelsvis02.ib.juwels.fzj.de:1)
<nikunj97> Starting applications specified in /etc/turbovnc/xstartup.turbovnc
<nikunj97> Log file is /p/home/jusers/gupta2/juwels/.vnc/juwelsvis02.ib.juwels.fzj.de:1.log
<nikunj97> zao ^^
<zao> You run your applications by having DISPLAY set on the remote machine, to the X display number exposed.
<zao> Then you connect to the VNC side of the thing, either directly or via a bunch of tunneling depending on how locked down it is.
<nikunj97> what do you mean by running application?
<nikunj97> the script they have runs an xfce4 session
<zao> Oh, then "just" connect with a VNC viewer to the address mentioned.
<nikunj97> it says no such host is known
<zao> The IP address may be an internal one to their network.
<nikunj97> is there a way to bypass that?
<zao> What OS are you on?
<nikunj97> on windows rn
<nikunj97> but can switch OSes
<zao> You can connect via SSH to the usual head nodes, and forward the port to your host.
<nikunj97> didn't quite get you
<zao> ssh -L 5901:juwelsvis02.ib.juwels.fzj.de:5901 you@head-node
<nikunj97> I connect to the head node
<nikunj97> let me try that
<zao> You can then connect to localhost:1 in your VNC viewer on your Windows machine, and it'll go through the SSH tunnel and connect to the address mentioned.
<nikunj97> aah
<nikunj97> let me try
<zao> It's -L localport:remotehost:remoteport
<zao> Looking at the address, it should be "obvious" that it's not a public one as it contains the term "ib" which I guess is InfiniBand :D
<nikunj97> ooo, didn't know that
<nikunj97> it says "the connection was refused by the computer"
<nikunj97> it worked
<zao> woo!
<nikunj97> \o/
<nikunj97> zao, thanks!
<nikunj97> btw my connection gets reset in ssh
<nikunj97> how do I keep it alive till I exit it myself?
<zao> SSH connections normally keep themselves alive. You can say -N to just set up the tunnel and not bother with a shell, but not sure if that changes anything.
<nikunj97> so I do ssh -N 5901:juwelsvis02.ib.juwels.fzj.de:5901 you@head-node?
<zao> ssh -N -L
<nikunj97> zao, I still get Connection reset by [ip] port 22
<zao> You may want to try juwelsvis02.fz-juelich.de:1 in your VNC client too.
<nikunj97> that didn't work for me
<nikunj97> aah you mean removing ib
<nikunj97> I haven't tried that
<zao> They mention those different fqdns at https://apps.fz-juelich.de/jsc/hps/juwels/visualization.html
<zao> Note the different top-level domain too.
<nikunj97> yes
nikunj97 has quit [Quit: Leaving]
maxwellr96 has joined #ste||ar
<maxwellr96> I'm trying to figure out how to test if an operation throws an exception in HPX, and I know I'm supposed to use the HPX_TEST_THROW macro, but I have to specify the exception, rather than the error code, and I can't find how to translate the error code to the precise exception being thrown
<maxwellr96> I know the error code from the source I want to test
<maxwellr96> And I can't find an example in the Phylanx codebase to serve as a template
nikunj97 has joined #ste||ar
<nikunj97> hkaiser, heller1 this is how the trace output looks like: https://imgur.com/a/zUsrDXQ
<nikunj97> there's quite a lot of white space in ther
<nikunj97> *there
<hkaiser> maxwellr96: you can do it yourself, just look at the HPX_TEST_THROW macro and create your own
<hkaiser> nikunj97: yah, that's how it normally looks like
<nikunj97> hkaiser, what should I be looking at?
<hkaiser> nikunj97, more parallelism or less cores
nikunj97 has quit [Remote host closed the connection]
nikunj97 has joined #ste||ar
nikunj97 has quit [Ping timeout: 260 seconds]