hkaiser changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar-group.org | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | This channel is logged: irclog.cct.lsu.edu
toma has joined #ste||ar
toma has left #ste||ar [Leaving]
jehelset has quit [Ping timeout: 250 seconds]
K-ballo has quit [Quit: K-ballo]
jehelset has joined #ste||ar
hkaiser has quit [Quit: Bye!]
jehelset has quit [Ping timeout: 250 seconds]
jehelset has joined #ste||ar
<ms[m]> diehlpk_work_: not aware of anything, and such a big performance decrease is unlikely to come from just piz daint updates
<ms[m]> gnikunj: https://github.com/eth-cscs/COSMA may be interesting to you
jehelset has quit [Ping timeout: 250 seconds]
Guest3737 has joined #ste||ar
Guest3737 has quit [Client Quit]
jehelset has joined #ste||ar
K-ballo has joined #ste||ar
hkaiser has joined #ste||ar
jehelset has quit [Ping timeout: 250 seconds]
hkaiser has quit [Quit: Bye!]
akheir has joined #ste||ar
diehlpk_work has joined #ste||ar
hkaiser has joined #ste||ar
<diehlpk_work> gonidelis[m], I could compile hpx on M1
<diehlpk_work> but all examples failed to run and segfaulted
<gonidelis[m]> Nice
<gonidelis[m]> That sounds like a gsoc project
<diehlpk_work> I do not think so
<diehlpk_work> the segfault was in one of the deps
jehelset has joined #ste||ar
<K-ballo> apple's M1, or something else?
<diehlpk_work> Yes
<diehlpk_work> We had to change some of the ifdef to get hpx compiled
hkaiser_ has joined #ste||ar
hkaiser has quit [Ping timeout: 240 seconds]
<hkaiser_> K-ballo: pls see #5766, thanks again for reporting
<diehlpk_work> That is amazing on Daint hpx is three times faster using system malloc instead of jemalloc
<diehlpk_work> Have we ever seen that?
<hkaiser_> diehlpk_work: I have not seen tis - what about tcmalloc - that's usually better than jemalloc
<diehlpk_work> hkaiser_, We never used tcmalloc for a while with octo, since I had issue on Power9 with it
<diehlpk_work> I can try
<gnikunj[m]> ms: thanks! I'll take a look
hkaiser has joined #ste||ar
hkaiser_ has quit [Ping timeout: 240 seconds]
hkaiser has quit [Quit: Bye!]
hkaiser has joined #ste||ar
<diehlpk_work> hkaiser, Ok, on Daint the first issue is jemalloc and the second issue is MPI vs tcp
<diehlpk_work> MPI gets faster with system malloc
<diehlpk_work> However, tcp using system malloc is still 2.5 times faster
<diehlpk_work> Latest version of jemalloc seems not to help at all
<hkaiser> diehlpk_work: ok
<hkaiser> diehlpk_work: I would still disagree that hpx is generally 'broken' in distributed
<diehlpk_work> hkaiser, I agree but we need to fix things
<hkaiser> diehlpk_work: 100% agreed
<diehlpk_work> It works in rostam for the paper and we get decent results
<diehlpk_work> My guess it that we have some race condition which is occuring more often with newer MPI and hardware
weilewei has joined #ste||ar
<weilewei> hkaiser can you update the hpxc main branch?
weilewei has quit [Quit: Ping timeout (120 seconds)]
ms[m] has quit [Ping timeout: 240 seconds]
srinivasyadav227 has quit [Ping timeout: 240 seconds]
KordeJong[m] has quit [Ping timeout: 240 seconds]
gonidelis[m] has quit [Ping timeout: 240 seconds]
mdiers[m] has quit [Ping timeout: 240 seconds]
bhumit[m] has quit [Ping timeout: 250 seconds]
gnikunj[m] has quit [Ping timeout: 240 seconds]
jedi18[m] has quit [Ping timeout: 252 seconds]
dkaratza[m] has quit [Ping timeout: 252 seconds]
rori[m] has quit [Ping timeout: 250 seconds]
pedro_barbosa[m] has quit [Ping timeout: 250 seconds]
heller[m] has quit [Ping timeout: 268 seconds]
gdaiss[m] has quit [Ping timeout: 268 seconds]
deepak[m] has quit [Ping timeout: 268 seconds]
jedi18[m] has joined #ste||ar
bhumit[m] has joined #ste||ar
weilewei has joined #ste||ar
<hkaiser> weilewei: done
weilewei has quit [Quit: Ping timeout (120 seconds)]
ms[m] has joined #ste||ar
KordeJong[m] has joined #ste||ar
srinivasyadav227 has joined #ste||ar
gonidelis[m] has joined #ste||ar
mdiers[m] has joined #ste||ar
dkaratza[m] has joined #ste||ar
rori[m] has joined #ste||ar
pedro_barbosa[m] has joined #ste||ar
gnikunj[m] has joined #ste||ar
gdaiss[m] has joined #ste||ar
heller[m] has joined #ste||ar
deepak[m] has joined #ste||ar
jehelset has quit [Ping timeout: 250 seconds]
weilewei has joined #ste||ar
weilewei has quit [Quit: Ping timeout (120 seconds)]
deepak[m] has quit [Read error: Connection reset by peer]
gnikunj[m] has quit [Read error: Connection reset by peer]
gdaiss[m] has quit [Read error: Connection reset by peer]
dkaratza[m] has quit [Write error: Connection reset by peer]
bhumit[m] has quit [Read error: Connection reset by peer]
heller[m] has quit [Read error: Connection reset by peer]
ms[m] has quit [Read error: Connection reset by peer]
pedro_barbosa[m] has quit [Read error: Connection reset by peer]
rori[m] has quit [Read error: Connection reset by peer]
mdiers[m] has quit [Read error: Connection reset by peer]
KordeJong[m] has quit [Read error: Connection reset by peer]
jedi18[m] has quit [Read error: Connection reset by peer]
srinivasyadav227 has quit [Read error: Connection reset by peer]
gonidelis[m] has quit [Read error: Connection reset by peer]
srinivasyadav227 has joined #ste||ar
weilewei has joined #ste||ar
<weilewei> cmake -DHPX_DIR=/home/weile/install/hpx/lib64/cmake/HPX/ ..
gonidelis[m] has joined #ste||ar
<weilewei> hkaiser is that a correct CMake command to build hpxc?
gnikunj[m] has joined #ste||ar
pedro_barbosa[m] has joined #ste||ar
<weilewei> CMake Error at /home/weile/install/hpx/lib64/cmake/HPX/HPX_AddPseudoDependencies.cmake:29 (add_dependencies):
<weilewei>   The dependency target "attr_stacksize_exe" of target
<weilewei>   "examples.threads.attr_stacksize" does not exist.
jedi18[m] has joined #ste||ar
gdaiss[m] has joined #ste||ar
<weilewei> same as create_thread_exe join_thread_exe and many other targets not exist
heller[m] has joined #ste||ar
ms[m] has joined #ste||ar
rori[m] has joined #ste||ar
KordeJong[m] has joined #ste||ar
dkaratza[m] has joined #ste||ar
bhumit[m] has joined #ste||ar
deepak[m] has joined #ste||ar
mdiers[m] has joined #ste||ar
<hkaiser> weilewei: ok
<hkaiser> didn't look
<weilewei> hkaiser yeah, how to correctly build hpxc?
<hkaiser> no idea ;-)
<weilewei> does it happen in your Windows machine? I am trying it on the rostam
<hkaiser> worked for me
<hkaiser> let me try again
<weilewei> sure
<hkaiser> weilewei: why is HPX_AddPseudoDependencies.cmake complaining while configuring hpxc?
<weilewei> I don't know
<hkaiser> have you HPX installed/build somewhere?
<weilewei> yes I installed it in /home/weile/install/hpx/
<weilewei> and use     cmake -DHPX_DIR=/home/weile/install/hpx/lib64/cmake/HPX/ ..  to build hpxc
<hkaiser> ok
<hkaiser> give me a sec
<weilewei> good
<hkaiser> weilewei: I just tried, all works for me (I configured against a built version of HPX, not an installed one, however)
<weilewei> hkaiser in rostam?
<weilewei> I tried the build version, same error
<weilewei> hkaiser
<weilewei> If I commented out the examples cmake config, the cmake config passed. https://github.com/STEllAR-GROUP/hpxc/blob/master/CMakeLists.txt#L39-L40
<weilewei> but then the build process gave me error: https://gist.github.com/weilewei/21dfda84a36fa0e9cc696d498f840ca2
<hkaiser> weilewei: pls change that to hpx::future
<hkaiser> and rename the struct on_exit to something else
<hkaiser> weilewei: let me fix all of this
<weilewei> thanks!
weilewei has quit [Quit: Ping timeout (120 seconds)]
jehelset has joined #ste||ar