K-ballo 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/
patrick22 has joined #ste||ar
patrick22 has quit [Ping timeout: 272 seconds]
hkaiser has quit [Quit: bye]
bita has quit [Ping timeout: 240 seconds]
akheir has quit [Quit: Leaving]
shahrzad has quit [Quit: Leaving]
patrick22 has joined #ste||ar
<gonidelis[m]>
ms[m]: i am getting this `unknown warning option '-Wno-sync-nand'` when using clang
<gonidelis[m]>
any idea where that might come from?
<ms[m]>
gonidelis[m]: we set it, the warning about the warning option being unknown is harmless, but I'd be interested to see what your cmake configuration says about that
<ms[m]>
we check if it's available and enable it only if it is, so either somethings strange in your configuration or I've broken the check
<gonidelis[m]>
hm let me see
<gonidelis[m]>
is there any place where i could retrieve my cmake config?
<gonidelis[m]>
because I built the stuff like one week ago
<gonidelis[m]>
ms[m]:
<ms[m]>
if you look in CMakeCache.txt I'd expect to see HPX_WITH_CXX_FLAG_WNO_SYNC_NAND:INTERNAL=1
<ms[m]>
can you try a clean build directory and see what it does?
<K-ballo>
did you switch your existing gcc build into a clang build?
<K-ballo>
like, just overwriting CMAKE_CXX_COMPILER ?
<ms[m]>
dirty :/
<ms[m]>
I sure hope not...
<gonidelis[m]>
K-ballo: yes i think i tried building with gcc at first and because it failed I switched to clang
<gonidelis[m]>
that's correct
<K-ballo>
that's no bueno
<K-ballo>
I though we had talked about setting CXX instead?
<gonidelis[m]>
although note that the gcc build never completed
<K-ballo>
you've effectively configured your build for gcc, but use clang
<gonidelis[m]>
what do you mean that's no bueno ;p ?
<K-ballo>
you are supposed to set the CXX environment variable on a clean build directory
<gonidelis[m]>
K-ballo: ok let me try
<gonidelis[m]>
Do I just `rm -r *` in my build dir first?
<K-ballo>
sure
<gonidelis[m]>
so just to be sure `CC=clang CXX=clang++ cmake [instal_path] [source_path]` is going to do the job ?
<K-ballo>
that's going to set CC and CXX env vars, then call cmake with the given arguments
<gonidelis[m]>
ok that's what we want. I think...
<gonidelis[m]>
K-ballo: btw did the porblem emrged because i used `CMAKE_CXX_COMPILER` rather than setting the env var `CXX`?
<K-ballo>
yes
<K-ballo>
the first cmake configuration pass is "special"
<gonidelis[m]>
wow didn't know that
<K-ballo>
you can't change the toolchain afterwards
<K-ballo>
if you do, you end up with a project using one toolchain but configured for a different one
<gonidelis[m]>
aren't they doing the same job though?
<gonidelis[m]>
also why is the first pass "special" >
<gonidelis[m]>
?
<K-ballo>
because the first pass configures the toolchain
<K-ballo>
they do a close enough job.. if you had set CMAKE_CXX_COMPILER directly on the first run it would have done effectively the same thing for the default toolchain
<K-ballo>
because the way CMAKE_CXX_COMPILER is specified to behave on the first run is to take its value from CXX
<gonidelis[m]>
thanks a lot :)
<K-ballo>
in general, you (as a user) are not supposed to mess with toolchain variables
<ms[m]>
since hpx::parallel::transform is deprecated and the segmented algorithm dispatch is done internally, using hpx::transform does (or is meant to) the right thing
hkaiser has joined #ste||ar
akheir has joined #ste||ar
shahrzad has joined #ste||ar
<diehlpk_work_>
HPX is has been funded by:
<diehlpk_work_>
This sounds akward or?
<diehlpk_work_>
ms[m], Is this a typo in the Readme?
<K-ballo>
definitely
<diehlpk_work_>
I will fix it
<K-ballo>
should it be is or has been? or both?
<K-ballo>
it was changed from HPX is currently funded by to HPX is has been funded by:
<ms[m]>
definitely a typo, can't comment on which one it should be...
<diehlpk_work_>
I did a pr with HPX has been funded by
<ms[m]>
thanks!
<ms[m]>
is it urgent to get in or can it wait for a comment from hkaiser?
<diehlpk_work_>
No, just found it while opening HPX on Github
<ms[m]>
👍️
<parsa>
hkaiser: finished with the summary. emailed it to you
<parsa>
let me know if you want me to add anything else to it
<hkaiser>
parsa: thanks
bita has joined #ste||ar
<bita>
hkaiser, Nan and I fixed a bug last night on the NumPy version and now we can train notMNIST. I am trying to find a good initialization point for the second-order
<hkaiser>
\o/
<diehlpk_work_>
akheir, Can I have a old rostam ndoe for one week?
<gonidelis[m]>
well... has anyone noticed before how superior HPX is, compared to other runtime systems?