2021-08-06 22:55
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
00:22
toma has joined #ste||ar
00:22
toma has left #ste||ar [Leaving]
00:49
jehelset has quit [Ping timeout: 250 seconds]
02:41
K-ballo has quit [Quit: K-ballo]
02:58
jehelset has joined #ste||ar
04:50
hkaiser has quit [Quit: Bye!]
05:41
jehelset has quit [Ping timeout: 250 seconds]
07:49
jehelset has joined #ste||ar
08:13
<
ms[m] >
diehlpk_work_: not aware of anything, and such a big performance decrease is unlikely to come from just piz daint updates
09:51
jehelset has quit [Ping timeout: 250 seconds]
11:13
Guest3737 has joined #ste||ar
11:14
Guest3737 has quit [Client Quit]
12:00
jehelset has joined #ste||ar
12:19
K-ballo has joined #ste||ar
12:33
hkaiser has joined #ste||ar
14:02
jehelset has quit [Ping timeout: 250 seconds]
14:32
hkaiser has quit [Quit: Bye!]
14:57
akheir has joined #ste||ar
15:54
diehlpk_work has joined #ste||ar
15:59
hkaiser has joined #ste||ar
16:09
<
diehlpk_work >
gonidelis[m], I could compile hpx on M1
16:09
<
diehlpk_work >
but all examples failed to run and segfaulted
16:09
<
gonidelis[m] >
Nice
16:10
<
gonidelis[m] >
That sounds like a gsoc project
16:10
<
diehlpk_work >
I do not think so
16:10
<
diehlpk_work >
the segfault was in one of the deps
16:12
jehelset has joined #ste||ar
16:26
<
K-ballo >
apple's M1, or something else?
16:27
<
diehlpk_work >
We had to change some of the ifdef to get hpx compiled
16:31
hkaiser_ has joined #ste||ar
16:33
hkaiser has quit [Ping timeout: 240 seconds]
16:42
<
hkaiser_ >
K-ballo: pls see #5766, thanks again for reporting
16:50
<
diehlpk_work >
That is amazing on Daint hpx is three times faster using system malloc instead of jemalloc
16:51
<
diehlpk_work >
Have we ever seen that?
16:55
<
hkaiser_ >
diehlpk_work: I have not seen tis - what about tcmalloc - that's usually better than jemalloc
16:57
<
diehlpk_work >
hkaiser_, We never used tcmalloc for a while with octo, since I had issue on Power9 with it
16:57
<
diehlpk_work >
I can try
17:09
<
gnikunj[m] >
ms: thanks! I'll take a look
17:31
hkaiser has joined #ste||ar
17:33
hkaiser_ has quit [Ping timeout: 240 seconds]
17:38
hkaiser has quit [Quit: Bye!]
18:43
hkaiser has joined #ste||ar
18:45
<
diehlpk_work >
hkaiser, Ok, on Daint the first issue is jemalloc and the second issue is MPI vs tcp
18:45
<
diehlpk_work >
MPI gets faster with system malloc
18:45
<
diehlpk_work >
However, tcp using system malloc is still 2.5 times faster
18:54
<
diehlpk_work >
Latest version of jemalloc seems not to help at all
19:02
<
hkaiser >
diehlpk_work: ok
19:02
<
hkaiser >
diehlpk_work: I would still disagree that hpx is generally 'broken' in distributed
19:14
<
diehlpk_work >
hkaiser, I agree but we need to fix things
19:19
<
hkaiser >
diehlpk_work: 100% agreed
19:32
<
diehlpk_work >
It works in rostam for the paper and we get decent results
19:32
<
diehlpk_work >
My guess it that we have some race condition which is occuring more often with newer MPI and hardware
19:35
weilewei has joined #ste||ar
19:35
<
weilewei >
hkaiser can you update the hpxc main branch?
19:46
weilewei has quit [Quit: Ping timeout (120 seconds)]
19:46
ms[m] has quit [Ping timeout: 240 seconds]
19:46
srinivasyadav227 has quit [Ping timeout: 240 seconds]
19:46
KordeJong[m] has quit [Ping timeout: 240 seconds]
19:46
gonidelis[m] has quit [Ping timeout: 240 seconds]
19:46
mdiers[m] has quit [Ping timeout: 240 seconds]
19:47
bhumit[m] has quit [Ping timeout: 250 seconds]
19:47
gnikunj[m] has quit [Ping timeout: 240 seconds]
19:47
jedi18[m] has quit [Ping timeout: 252 seconds]
19:47
dkaratza[m] has quit [Ping timeout: 252 seconds]
19:47
rori[m] has quit [Ping timeout: 250 seconds]
19:47
pedro_barbosa[m] has quit [Ping timeout: 250 seconds]
19:47
heller[m] has quit [Ping timeout: 268 seconds]
19:47
gdaiss[m] has quit [Ping timeout: 268 seconds]
19:47
deepak[m] has quit [Ping timeout: 268 seconds]
20:03
jedi18[m] has joined #ste||ar
20:11
bhumit[m] has joined #ste||ar
20:13
weilewei has joined #ste||ar
20:15
<
hkaiser >
weilewei: done
20:20
weilewei has quit [Quit: Ping timeout (120 seconds)]
20:24
ms[m] has joined #ste||ar
20:41
KordeJong[m] has joined #ste||ar
20:41
srinivasyadav227 has joined #ste||ar
20:42
gonidelis[m] has joined #ste||ar
20:42
mdiers[m] has joined #ste||ar
20:43
dkaratza[m] has joined #ste||ar
20:43
rori[m] has joined #ste||ar
20:43
pedro_barbosa[m] has joined #ste||ar
20:43
gnikunj[m] has joined #ste||ar
20:43
gdaiss[m] has joined #ste||ar
20:43
heller[m] has joined #ste||ar
21:13
deepak[m] has joined #ste||ar
21:15
jehelset has quit [Ping timeout: 250 seconds]
21:28
weilewei has joined #ste||ar
21:35
weilewei has quit [Quit: Ping timeout (120 seconds)]
21:36
deepak[m] has quit [Read error: Connection reset by peer]
21:36
gnikunj[m] has quit [Read error: Connection reset by peer]
21:36
gdaiss[m] has quit [Read error: Connection reset by peer]
21:36
dkaratza[m] has quit [Write error: Connection reset by peer]
21:36
bhumit[m] has quit [Read error: Connection reset by peer]
21:36
heller[m] has quit [Read error: Connection reset by peer]
21:36
ms[m] has quit [Read error: Connection reset by peer]
21:36
pedro_barbosa[m] has quit [Read error: Connection reset by peer]
21:36
rori[m] has quit [Read error: Connection reset by peer]
21:36
mdiers[m] has quit [Read error: Connection reset by peer]
21:36
KordeJong[m] has quit [Read error: Connection reset by peer]
21:36
jedi18[m] has quit [Read error: Connection reset by peer]
21:36
srinivasyadav227 has quit [Read error: Connection reset by peer]
21:36
gonidelis[m] has quit [Read error: Connection reset by peer]
21:38
srinivasyadav227 has joined #ste||ar
21:40
weilewei has joined #ste||ar
21:41
<
weilewei >
cmake -DHPX_DIR=/home/weile/install/hpx/lib64/cmake/HPX/ ..
21:41
gonidelis[m] has joined #ste||ar
21:41
<
weilewei >
hkaiser is that a correct CMake command to build hpxc?
21:41
gnikunj[m] has joined #ste||ar
21:41
pedro_barbosa[m] has joined #ste||ar
21:41
<
weilewei >
CMake Error at /home/weile/install/hpx/lib64/cmake/HPX/HPX_AddPseudoDependencies.cmake:29 (add_dependencies):
21:41
<
weilewei >
The dependency target "attr_stacksize_exe" of target
21:41
<
weilewei >
"examples.threads.attr_stacksize" does not exist.
21:41
jedi18[m] has joined #ste||ar
21:42
gdaiss[m] has joined #ste||ar
21:42
<
weilewei >
same as create_thread_exe join_thread_exe and many other targets not exist
21:42
heller[m] has joined #ste||ar
21:42
ms[m] has joined #ste||ar
21:42
rori[m] has joined #ste||ar
21:43
KordeJong[m] has joined #ste||ar
21:43
dkaratza[m] has joined #ste||ar
21:44
bhumit[m] has joined #ste||ar
21:44
deepak[m] has joined #ste||ar
21:44
mdiers[m] has joined #ste||ar
21:51
<
hkaiser >
weilewei: ok
21:51
<
hkaiser >
didn't look
21:52
<
weilewei >
hkaiser yeah, how to correctly build hpxc?
21:52
<
hkaiser >
no idea ;-)
21:52
<
weilewei >
does it happen in your Windows machine? I am trying it on the rostam
21:52
<
hkaiser >
worked for me
21:52
<
hkaiser >
let me try again
21:53
<
hkaiser >
weilewei: why is HPX_AddPseudoDependencies.cmake complaining while configuring hpxc?
21:53
<
weilewei >
I don't know
21:53
<
hkaiser >
have you HPX installed/build somewhere?
21:54
<
weilewei >
yes I installed it in /home/weile/install/hpx/
21:54
<
weilewei >
and use cmake -DHPX_DIR=/home/weile/install/hpx/lib64/cmake/HPX/ .. to build hpxc
21:54
<
hkaiser >
give me a sec
21:56
<
hkaiser >
weilewei: I just tried, all works for me (I configured against a built version of HPX, not an installed one, however)
22:00
<
weilewei >
hkaiser in rostam?
22:16
<
weilewei >
I tried the build version, same error
22:31
<
hkaiser >
weilewei: pls change that to hpx::future
22:32
<
hkaiser >
and rename the struct on_exit to something else
22:33
<
hkaiser >
weilewei: let me fix all of this
22:54
weilewei has quit [Quit: Ping timeout (120 seconds)]
23:23
jehelset has joined #ste||ar