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/
<hkaiser>
let me create a patch for the environ problem for you, sec
<hkaiser>
no sure, need to match those symbols against what's missing in your build
<zao>
Ah, nm had a demangle option too.
<hkaiser>
no worries I can demangle that myself, if needed
<zao>
Put the actual library in that directory too.
<zao>
Now actual sleep!
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
<zao>
tests/unit/parallel/segmented_algorithms/CMakeFiles/partitioned_vector_exclusive_scan_test_exe.dir/partitioned_vector_exclusive_scan.cpp.o: file not recognized: File format not recognized
<zao>
That's odd. (FreeBSD)
<zao>
Linker suddenly hates partitioned_vector_exclusive_scan.cpp.o, partitioned_vector_transform_binary.cpp.o, and partitioned_vector_transform_scan.cpp.o
<zao>
Not overly big either, 278M, 432M and 314M.
<zao>
They're the biggest three of the parallel segmented_algorithms tests tho.
EverYoung has quit [Remote host closed the connection]
<hkaiser>
the last sentence of this is correct, the first not
<jbjnr>
lol. thanks
<jbjnr>
hkaiser: last one - so get_next_thread is called when the scheduler wants some work, and wait_or_add_new is called for the same reason - what's the difference
<hkaiser>
get_next_thread just looks at the queues (including stealing), wait_or_add_new is called whenever it runs out of work and converts staged threads (thread descriptions) into real threads
<hkaiser>
so I guess the first sentence of that comment is not that wrong after all ;)
<hkaiser>
jbjnr: I have more refactorings to push, this time it shouldn't disrupt your work, if it does tell me so I can make it work again
aserio has quit [Ping timeout: 258 seconds]
<jbjnr>
what refactorings?
aserio has joined #ste||ar
<jbjnr>
hkaiser: if you push your refactorings to RP branch - I'll inspect them carefully before merging to my stuff
<hkaiser>
jbjnr: will push in a sec
<hkaiser>
ok - pls do
<jbjnr>
fingers crossed I won't have conflicts
<zao>
I need to work on HPX so I also can break jbjnr's code :)
<hkaiser>
I hope you don't
<hkaiser>
jbjnr: I pushed my changes
denis_blank has joined #ste||ar
<jbjnr>
you modified every file there is.
<jbjnr>
<sigh>
denis_blank2 has joined #ste||ar
denis_blank has quit [Disconnected by services]
denis_blank2 has quit [Client Quit]
denis_blank has joined #ste||ar
<hkaiser>
jbjnr: shouldn't conflict, though
<hkaiser>
jbjnr: was it intentional to change hpx such that in distributed runs (using the RP branch) now every locality executes hpx_main?
<hkaiser>
or is that not intentional (then I'll fix it)
<jbjnr>
hkaiser: I don't think it was intentional. and I did have conflicts.
<jbjnr>
and yes please - please do fix it
quantomb has quit [Quit: Leaving]
<patg[[w]]>
wash: did you update the authors ? If so please push you commit
<zao>
I'm overriding CMake linker flags... what on earth is CMAKE_MODULE_LINKER_FLAGS for?
<zao>
I understand EXE/SHARED/STATIC, but no docs exist on MODULE.
<github>
[hpx] Naios opened pull request #2799: Fix a unit test failure on GCC in tuple_cat (master...tuple_cat_fix) https://git.io/v7Vx9
<zao>
Neato, FreeBSD is working toward using LLD as system linker in the future. Currently default linker on arm64 HEAD and upcoming 11.1.
<zao>
I don't have to feel too bad about using ld.lld-devel then.
<zao>
Only affects those few tests anyway for people who don't want to install packages.
<wash>
working on it
<wash>
had to redo it because the previous authors block depended on the acm package
<patg[[w]]>
wash yeah I saw that
denis_blank has quit [Quit: denis_blank]
<heller>
wash: did we submit the abstract yet?
<wash>
yes
<hkaiser>
jbjnr: could you explain to me please why we have several overloads for the get_resource_partitioner() function?
<hkaiser>
jbjnr: is that part of a bigger plan?
<heller>
neato
<jbjnr>
as I recall - its because if the user doesn't use the resource manager, then hpx::init will do the calls to initialize it, and we need one with argc,argv, and one without for general access after initialization. Can't remember others, but feel free to remove some if they are surplus
hkaiser has quit [Quit: bye]
<heller>
jbjnr: good luck. I am off now. vacation ahead
eschnett has quit [Quit: eschnett]
<jbjnr>
heller: enjoy your vacation
<K-ballo>
heller: typo, ahaed
<wash>
enjoy
<wash>
patg[[w]]: check your email, I sent a copy with authors
<diehlpk_work>
Blog post with the status update of our students
hkaiser has joined #ste||ar
bikineev has quit [Ping timeout: 248 seconds]
bikineev has joined #ste||ar
eschnett has joined #ste||ar
<jbjnr>
hkaiser: you missed a reply earlier I think
<jbjnr>
as I recall - its because if the user doesn't use the resource manager, then hpx::init will do the calls to initialize it, and we need one with argc,argv, and one without for general access after initialization. Can't remember others, but feel free to remove some if they are surplus
<jbjnr>
^^^
<jbjnr>
about resource manager get api
<hkaiser>
jbjnr: ok, thanks
<hkaiser>
I think we need only one of those
<diehlpk_work>
I have vacation next week and will be not available
<hkaiser>
diehlpk_work: have fun!
<diehlpk_work>
Thanks
<hkaiser>
diehlpk_work: thanks for the blog post!
<diehlpk_work>
You are welcome
<diehlpk_work>
I will write a final one too
<hkaiser>
great!
<diehlpk_work>
I also was thinking about a statistic for gsoc
<diehlpk_work>
One post with diagramms of amount of students per year, passed fail, and so on
<diehlpk_work>
What do you think
<hkaiser>
sounds good to me!
<hkaiser>
and number of PRs
<diehlpk_work>
Yes
<hkaiser>
if we can reproduce this...
<diehlpk_work>
If not we start with this year for the pr
<hkaiser>
good thinking
aserio has quit [Quit: aserio]
<diehlpk_work>
gtg, bye
diehlpk_work has quit [Quit: Leaving]
<jbjnr>
hkaiser: if you fix the run_main in distributed, then please let me know. I would like to do an mpi test over the weekend and get some new perf graphs of the distributed matrix stuff. if hpx_main is broken (and I didn't know about it), then all my results would be nonsense
<hkaiser>
jbjnr: almost done
<jbjnr>
great. thanks
<jbjnr>
almost fixed all the new errors you gave me with your earlier commit!'
<hkaiser>
jbjnr: if you gave me access to your branch I'd be able to do that for you - by no means I want to cause more work on your end
<hkaiser>
and I apologize for any trouble this might cause
<jbjnr>
I have the same branch, but many uncommitted tweaks. I am working on the same material that you are changing
EverYoung has quit [Remote host closed the connection]
mcopik has quit [Ping timeout: 255 seconds]
bikineev has quit [Remote host closed the connection]
<hkaiser>
jbjnr: just commit your tweaks, no problem if that changes again
<hkaiser>
jbjnr: is it ok to push the fix now?
<jbjnr>
hkaiser: yes please. my stuff i working again and I'm off to bed.
<jbjnr>
tomorrow I'll try an mpi run
<jbjnr>
thanks very much for the fixes
<hkaiser>
jbjnr: pushed, should break anything this time
<github>
[hpx] hkaiser pushed 1 new commit to resource_partitioner: https://git.io/v7wca
<github>
hpx/resource_partitioner f54f94c Hartmut Kaiser: Fix running hpx_main only on locality 0 by default
<hkaiser>
should not *
<jbjnr>
thanks - and goodnight
jbjnr_ has joined #ste||ar
jbjnr has quit [Ping timeout: 258 seconds]
jbjnr_ is now known as jbjnr
<hkaiser>
heller: yt?
<hkaiser>
chapter 1 page two you flipped the meaning of intra- and inter- node parallelism
<hkaiser>
intra- is used for 'inside', inter- is used for 'in-between'
<hkaiser>
you're using it the other way around
parsa has joined #ste||ar
vamatya_ has quit [Read error: Connection reset by peer]
vamatya_ has joined #ste||ar
patg[[w]] has quit [Ping timeout: 240 seconds]
<github>
[hpx] hkaiser force-pushed resource_partitioner from f54f94c to 7a1dfbf: https://git.io/v7lfK
<github>
hpx/resource_partitioner 7a1dfbf Hartmut Kaiser: Fix running hpx_main only on locality 0 by default