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/
eschnett has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
hkaiser has joined #ste||ar
quaz0r has quit [Ping timeout: 256 seconds]
hkaiser has quit [Quit: bye]
quaz0r has joined #ste||ar
Smasher has joined #ste||ar
Smasher has quit [Remote host closed the connection]
david_pfander has joined #ste||ar
heller has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
heller has joined #ste||ar
Vir has joined #ste||ar
hkaiser has joined #ste||ar
hkaiser has quit [Ping timeout: 252 seconds]
K-ballo has joined #ste||ar
K-ballo has quit [Read error: Connection reset by peer]
K-ballo has joined #ste||ar
aserio has joined #ste||ar
hkaiser has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
Smasher has joined #ste||ar
K-ballo has joined #ste||ar
<aserio> hkaiser: yt?
<aserio> see pm
stevenrbrandt has joined #ste||ar
<stevenrbrandt> Hello, I'm trying to get Phylanx to build and run. Typically, it fails when I run "python3 -c 'import phylanx'".
<stevenrbrandt> I'm trying to figure out how the CircleCI system builds everything.
<hkaiser> stevenrbrandt - talk to Rod
<stevenrbrandt> Toward that end, does anyone know where the source is for the docker image stellargroup/build_env:debian_clan
<stevenrbrandt> Rod doesn't know where the source to the above docker image is.
<hkaiser> ahh, that one - hold on
<aserio> akheir: yt?
<stevenrbrandt> thanks
<zao> stevenrbrandt: I would reckon that the import resolves there because it's installed into something that's in the sys.path for the system Python.
<zao> Or otherwise have the PYTHONPATH set up when running.
rtohid has joined #ste||ar
<zao> Also be vary of failures to load dependency shared libraries, which would be found via the stock search paths from ld.so.conf, and LD_LIBRARY_PATH.
<zao> Depends really on the kind of error you've got.
<stevenrbrandt> zao: The problem with running "import phylanx" is this error message: RuntimeError: partitioner::add_resource: Creation of 2 threads requested by the resource partitioner, but only 1 provided on the command-line.
<stevenrbrandt> For some reason, when CircleCI builds everything, this doesn't happen.
<hkaiser> stevenrbrandt - this I think is a problem in HPX not Phylanx, but I was not able yet to be imaginative enought to think of a reason
<hkaiser> jbjnr might have some insights or simbergm
<zao> What kind of resource constraints does CI run under? Is it single-cored?
<hkaiser> stevenrbrandt: I know that linux loads a module twice if you dlopen it with different but equivalent paths
<hkaiser> that could cause havoc
<hkaiser> can you list the full path of all loaded modules in gdb at the point of error
<hkaiser> ?
<K-ballo> zao: two cores
<stevenrbrandt> What I don't understand is why no one else but me sees this problem. I've tried building it on Ubuntu and Fedora with the same results.
hkaiser has quit [Quit: bye]
<stevenrbrandt> I'm looking at circle.yml. Where does the git clone of hpx happen?
eschnett has quit [Quit: eschnett]
quaz0r has quit [Ping timeout: 265 seconds]
<zao> Eeew... phylanx build installed into my homedir and not the prefix.
<zao> Installed /home/zao/.local/lib/python3.5/site-packages/phylanx-0.0.1-py3.5-linux-x86_64.egg
<zao> stevenrbrandt: If it makes you happier, I segfault on importing phylanx (in release build)
<zao> stevenrbrandt: What flavor do you build? Debug? RelWithDebInfo?
<zao> That's where I'm at when crashing on import.
akheir has quit [Remote host closed the connection]
<zao> This build is very fishy... phylanx python .so file points into the build tree and /usr/local/lib as the runpath, and has a reference to Shared library: [libhpxd.so.1]
<zao> Which is completely bullshit, as I've got a Release HPX.
quaz0r has joined #ste||ar
eschnett has joined #ste||ar
<zao> libhpx_phylanx.so also refers to "libhpx_iostreamsd.so.1"
<zao> Something is quite bogus in my container.
<stevenrbrandt> zao: I have a docker file. How do I send it to IRC?
<stevenrbrandt> It shows how I build my release build.
<zao> gist.github.com or a paste site, and show the link.
<zao> Looks sane at a glance.
<stevenrbrandt> sanity is what I strive for. :)
<zao> stevenrbrandt: Just built Debug-everything, phylanx loads properly.
<zao> In the exact same environment that Release-everything tanked.
<stevenrbrandt> What did you use for your memory allocator? tcmalloc?
<zao> Ah, it finds the HPX that apparently is present in the system directories in the docker image.
<zao> system.
<zao> Didn't mutate the phylanx image.
david_pfander has quit [Ping timeout: 256 seconds]
<stevenrbrandt> My debug builds hang.
<zao> Oh. blargh... misunderstood HPX_DIR.
<stevenrbrandt> Yeah, it has to point to the cmake thingy
<stevenrbrandt> if it's wrong, cmake searches your home directory tree for an HPX
<stevenrbrandt> and the first thing it finds, it uses
<zao> Much better with competently invoked build, dependencywise.
<zao> I'm sad to say that it doesn't crash in my container.
<zao> Which is docker://stellargroup/phylanx:dev
<zao> (on a laptop out in the wilderness, so not much bandwidth to play around on :D)
<zao> Notable may be that I'm building HPX and Phylanx with the Clang5 from the image.
Smasher has quit [Remote host closed the connection]
<aserio> Does anyone know how to get the type of a component from its client?
<stevenrbrandt> zao, do you have a dockerfile for what you are doing?
<stevenrbrandt> could you make one, maybe?
<zao> stevenrbrandt: Probably not, all I've done is pull down the HPX images verbatim from docker and building HPX and Phylanx inside: https://gist.github.com/zao/fc402f5a0c54d43409dbc22c24222c60
aserio has quit [Quit: aserio]
<heller> stevenrbrandt: is this running on a 4 socket machine?
<stevenrbrandt> 2 sockets
akheir has joined #ste||ar
<stevenrbrandt> OK, so @heller provided the clue. Running on a 1 socket machine works, running on 2 sockets gives me the error.
<stevenrbrandt> It had nothing to do with how I was building.
<zao> Heh.
<zao> Interesting note, Blaze detects the wrong compiler unless I explicitly set CMAKE_C/CXX_COMPILER.
<zao> Nevermind, forgot to load the magical buildenv module
hkaiser has joined #ste||ar
<zao> Good news, I'm crashing on startup on my cluster - https://gist.github.com/zao/c27f89154191a116d1ef399499922973
<zao> Crashes on nodes with -n1 -N1 -c1 too, so might be a completely disjoint problem on my side.
* zao returns to Lethal Weapon 3, quality holiday entertainment.
<zao> (still annoying that it installs to ~/.local, cluster nodes won't know crap about that)
<zao> (I forget who the author is of phylanx)
<stevenrbrandt> Phylanx is still pre-alpha and has many authors
<zao> Lovely :)
<heller> stevenrbrandt: good, do you know if you run with numa-balanced or any other bind option?
<zao> I'm thinking of the person I was testing the build for the other week, would have to refer to logs to see that.
<zao> Thought it was a one-person-thing :)
<stevenrbrandt> I don't know how to specify bind options
<stevenrbrandt> that was to @heller
appro_ has joined #ste||ar
<heller> Ok, I mean in the phylanx code
appro_ has quit [Client Quit]
<stevenrbrandt> @heller, I'm simply loading phylanx from python, so I have no opportunity to do that.
<stevenrbrandt> @heller, I get whatever hpx_init() gives me
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
<heller> stevenrbrandt: ok, I have complaints about --hpx:bind=numa-balanced on 4 socket systems, that's why I ask
akheir has quit [Remote host closed the connection]
rtohid has left #ste||ar [#ste||ar]
hkaiser_ has joined #ste||ar
hkaiser has quit [Ping timeout: 256 seconds]
hkaiser_ has quit [Quit: bye]
<zao> No ini either then, I reckon.
eschnett has quit [Quit: eschnett]
stevenrbrandt has quit [Quit: Page closed]