02:35
adityaRakhecha has quit [Ping timeout: 256 seconds]
02:47
dexhunter has quit [Quit: Ping timeout (120 seconds)]
02:47
dexhunter has joined #ste||ar
03:31
hkaiser has quit [Quit: bye]
06:42
zao has quit [*.net *.split]
06:43
zao has joined #ste||ar
06:48
jaafar_ has quit [Quit: Konversation terminated!]
07:48
<
jbjnr_ >
zao - thanks! Does that mean you actually watched it all the way through?
08:23
david_pfander has joined #ste||ar
10:10
<
zao >
2/3 through thus far
10:58
<
jbjnr_ >
don't punish yourself!
11:12
nikunj has quit [Ping timeout: 250 seconds]
11:20
nikunj has joined #ste||ar
11:36
nikunj has quit [Ping timeout: 268 seconds]
11:45
nikunj97 has joined #ste||ar
12:11
hkaiser has joined #ste||ar
13:38
K-ballo has quit [Ping timeout: 246 seconds]
13:55
K-ballo has joined #ste||ar
14:38
hello has joined #ste||ar
14:39
<
hello >
hello, everybody
14:40
time_ has joined #ste||ar
14:40
<
K-ballo >
hello hello, hello
14:41
<
time_ >
why does graph500 offical code segment fault when calculate scale >= 29?
14:41
<
time_ >
===================================================================================
14:41
<
time_ >
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
14:41
<
time_ >
= PID 120884 RUNNING AT server950-4
14:41
<
time_ >
= EXIT CODE: 11
14:41
<
time_ >
= CLEANING UP REMAINING PROCESSES
14:41
<
time_ >
= YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
14:41
<
time_ >
===================================================================================
14:41
<
time_ >
YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Segmentation fault (signal 11)
14:41
<
time_ >
This typically refers to a problem with your application.
14:41
<
time_ >
Please see the FAQ page for debugging suggestions
14:41
hello has quit [Remote host closed the connection]
14:41
time_ is now known as hello
14:42
<
zao >
See if you got a core dump, look at it with a debugger?
14:43
<
hello >
yes, I have done it.
14:44
<
zao >
Ah, it's a GSoC task to port some sort of benchmark to HPX?
14:44
<
hello >
warning: core file may not match specified executable file.
14:44
<
hello >
[New LWP 122104]
14:44
<
hello >
Failed to read a valid object file image from memory.
14:44
<
hello >
Cannot access memory at address 0x3d7ce21160
14:44
<
hello >
Cannot access memory at address 0x3d7ce21168
14:44
<
hello >
Core was generated by `./graph500_reference_bfs 30'.
14:44
<
hello >
Program terminated with signal SIGSEGV, Segmentation fault.
14:44
<
hello >
Python Exception <type 'exceptions.NameError'> Installation error: gdb.execute_unwindersfunction is missing:
14:44
<
hello >
#0 0x000000000040239a in fulledgehndl ()
14:44
<
hello >
Python Exception <type 'exceptions.ImportError'> No module named gdb.frames:
14:44
<
hello >
#0 0x000000000040239a in fulledgehndl ()
14:44
<
hello >
Backtrace stopped: Cannot access memory at address 0x7fff9401e2c8
14:44
<
zao >
Please use a gist or a pastesite for bulk output :)
14:45
<
hello >
ok, thanks for your remind.
14:45
<
hello >
zao: have you met this bug?
14:46
<
zao >
Never touched the software.
14:46
<
zao >
I wonder if it's memory-heavy, this benchmark.
14:46
<
hello >
do you know other irc channel that maybe using this software?
14:48
<
zao >
Don't know, but the mentors may have some clue about the software.
14:48
<
zao >
heller_: boop ^
14:48
<
zao >
What kind of compute node are you running on?
14:48
<
zao >
And how do you run/submit?
14:49
<
heller_ >
hello: hmm, isn't 29 quite a big input?
14:49
<
hkaiser >
hello: this is not an HPX application, is it?
14:49
<
hello >
I am a student, just interested in it.
14:50
<
heller_ >
hello: this is the MPI reference implementation?
14:51
<
hello >
yes, you are right.
14:51
<
heller_ >
I think you are just running out of memory
14:51
<
zao >
You'll also need 128G of disk space, according to the README.
14:51
<
hello >
My memory is 2T
14:52
<
hello >
both condition satisfied
14:52
<
zao >
I should try building this.
14:52
<
heller_ >
hello: so, scale == level, right?
14:53
<
heller_ >
sorry, yes
14:53
<
heller_ >
it's been a while ;)
14:53
<
zao >
Could you please list the command line you used?
14:53
<
heller_ >
so you want to run the mini problem
14:53
<
hello >
make clean;make
14:53
<
hello >
mpirun -n 64 ./graph500_reference_bfs 30
14:53
<
heller_ >
what's the default edge factor?
14:53
<
hello >
thank you all very much, lol
14:54
<
hkaiser >
how many cores do you have on your machine?
14:54
<
heller_ >
hello: I have no idea why it segfaults, to be honest. We are not the maintainer of the reference implementation ;)
14:55
<
zao >
You might get slightly less bad debugging output if you add `-g` to the build flags I guess.
14:56
<
hello >
yeah, you are right
14:57
<
zao >
Hrm, maybe not, that talks about per-process.
14:57
<
hello >
you can run this program. then I will know whether beacause my machine physical configuration caused that.
14:58
<
hello >
yeah, I have read that website, but not solved.
15:00
<
hello >
I love you.
15:01
<
zao >
I don't have much memory per core on my cluster, so might not be the same kind of failure as yours.
15:01
<
hello >
could your compile this using -g -O0?
15:01
<
hello >
which gcc version do you use?
15:03
<
zao >
Oddly enough, -O0 -g fails to link.
15:03
<
hello >
do you know why not to link?
15:06
<
hello >
there are some warning...
15:08
<
zao >
Ah yes, all the "inline" functions need to be "static inline", probably a C standards thing.
15:19
<
hello >
yeah,you are right
15:19
<
zao >
graph_generation: 506.287539 s
15:19
<
zao >
AML: Fatal: non power2 groupsize unsupported. Define macro PROCS_PER_NODE_NOT_POWER_OF_TWO to override
15:20
<
zao >
Silly me, only asked SLURM for -n64 :)
15:23
<
zao >
Anyway, the reference code seems a bit brittle, I hope you manage to sort it out.
15:24
<
hello >
I know this.
15:24
<
hello >
I have learned the code.
15:25
<
hello >
almost understand and I have expand some code.
15:26
ste||ar-github has joined #ste||ar
15:26
<
ste||ar-github >
hpx/function_ref 1609077 Agustin K-ballo Berge: Implement util::function_ref
15:26
ste||ar-github has left #ste||ar [#ste||ar]
15:30
<
hello >
zao: how to gdb lauch multi MPI process?
15:31
<
zao >
No idea, we tend to recommend Allinea for our users when they need to debug MPI jobs.
15:31
<
hello >
I use attach
15:37
<
hello >
CFLAGS = -Drestrict=__restrict__ -O0 -g -gdwarf-2 -g3
15:38
<
hello >
zao: do you know why this not produce dump file when core dump?
15:38
<
hello >
my ulimit -c uliminted
15:39
<
zao >
`/proc/sys/kernel/core_pattern` might be sending the dumps somewhere else?
15:46
<
hello >
|/usr/libexec/abrt-hook-ccpp %s %c %p %u %g %t e
15:47
<
hello >
core_pattern's content is thi
15:47
<
zao >
A leading pipe means that core dumps are handled by the following command, so in your case the 'abrt' tool.
15:47
<
zao >
Which may ignore or stow them away somewhere else on the machine.
15:49
<
zao >
Backtraces on my other cluster, had more bigmem nodes free there.
15:49
<
hello >
I can not modify the core_pattern file.
15:50
<
hello >
So, what should I do to generate core files?
15:55
<
hello >
it seems I cant modify the path
16:11
<
zao >
Built with GCC 6.3.0 and OpenMPI 2.0.2, crashes at the same point as GCC 7.3.0 and OpenMPI 3.1.1.
16:12
<
zao >
No idea what to do from here. As for your core dumps, you probably have to involve a cluster admin.
16:20
ste||ar-github has joined #ste||ar
16:20
ste||ar-github has left #ste||ar [#ste||ar]
16:23
ste||ar-github has joined #ste||ar
16:23
<
ste||ar-github >
hpx/function_ref 167d29c Agustin K-ballo Berge: Implement util::function_ref
16:23
ste||ar-github has left #ste||ar [#ste||ar]
16:46
<
heller_ >
err, the expressions in the assertions are always evaluated?
17:22
adityaRakhecha has joined #ste||ar
17:23
ct-clmsn has joined #ste||ar
17:25
<
ct-clmsn >
hkaiser, did some more work on opencv_imread this weekend, the data loads but still getting blocking behavior
17:26
<
adityaRakhecha >
while executing one of the examples I am getting this, make:
*** No rule to make target 'fibonacci_local'. Stop.
17:26
<
ct-clmsn >
hkaiser, the dynamictensor<std::uint8_t> is working great for storing pixel data
17:38
<
adityaRakhecha >
anyone?
17:39
<
ct-clmsn >
adityaRakhecha, not sure about that issue - you get a clean build from cmake?
17:41
<
ct-clmsn >
will think about it, not sure i have a fix in mind
17:46
<
simbergm >
adityaRakhecha: are you on 1.2.0 or master? you might need `make fibonacci_local_exe`
17:48
<
adityaRakhecha >
1.2.0
17:49
<
adityaRakhecha >
Still same message(error)
17:57
hello has quit [Read error: Connection timed out]
18:02
<
simbergm >
adityaRakhecha: looks like a bug... could you check `make help | grep fibonacci`?
18:03
<
simbergm >
if there are other fibonacci examples there but not `fibonacci_local` I'd be very grateful if you could open an issue
18:04
hello has joined #ste||ar
18:05
<
heller_ >
K-ballo: regarding your template instantiation count ... I guess most of the enable_if instantiations could be eliminated with if constexpr
18:07
<
heller_ >
and then the question is which one is faster ;)
18:10
<
K-ballo >
we already shouldn't be doing enable_if for something tag dispatching could handle
18:16
jaafar has joined #ste||ar
18:21
ste||ar-github has joined #ste||ar
18:21
<
ste||ar-github >
hpx/function_ref 6a4e7be Agustin K-ballo Berge: Implement util::function_ref
18:21
ste||ar-github has left #ste||ar [#ste||ar]
18:45
ste||ar-github has joined #ste||ar
18:45
<
ste||ar-github >
hpx/function_ref 6539b09 Agustin K-ballo Berge: Implement util::function_ref
18:45
ste||ar-github has left #ste||ar [#ste||ar]
18:50
<
heller_ >
K-ballo: I don't think so. IIRC we don't use tag dispatching that often
18:50
<
heller_ >
or we have a different definition for it ;)
18:54
<
K-ballo >
we might not be doing it, I haven't looked for it, but we should
18:54
<
K-ballo >
there's a greater gain in going from sfinae to tag dispatching than from tag dispatching to constexpr if
19:03
<
K-ballo >
and we actually do do it, maybe not everywhere but I have seen plenty of them
19:07
<
simbergm >
adityaRakhecha: I know, I must've messed up the cmake config somehow
19:08
<
simbergm >
feel free to see if you can fix it, but I'll try to fix it tomorrow otherwise
19:09
<
adityaRakhecha >
I would love to work on it :D
19:11
<
adityaRakhecha >
Could you please suggest how should I proceed for it ?
19:13
adityaRakhecha_ has joined #ste||ar
19:14
adityaRakhecha has quit [Quit: Page closed]
19:15
<
adityaRakhecha_ >
Tomorrow after my college I will open the issue and work on it. Please let me do it.
19:17
<
simbergm >
adityaRakhecha_: yeah, no problem, I'm very happy if you look into it
19:18
<
simbergm >
I don't know of a debugger for cmake, so your best bet is sprinkling message here and there
19:22
<
heller_ >
it's a little beyond me why fibonacci_local shouldn't be there but the others are
19:23
<
zao >
Which one is it that has a name collision between examples and tests?
19:23
<
zao >
Also, do we build tests/examples by default now?
19:23
<
heller_ >
since ever, yes
19:24
<
zao >
I never remember and thus explicitly specify them.
19:25
<
simbergm >
adityaRakhecha_, heller_: hrm, I had an old checkout without fibonacci_local and thought it didn't work because of that
19:25
<
simbergm >
adityaRakhecha_: could you double-check with latest master and/or a fresh build directory
19:49
ste||ar-github has joined #ste||ar
19:49
ste||ar-github has left #ste||ar [#ste||ar]
20:35
ct-clmsn has quit [Quit: Leaving]
21:28
K-ballo has quit [Read error: Connection reset by peer]
21:29
K-ballo has joined #ste||ar
22:11
nikunj97 has quit [Ping timeout: 246 seconds]
23:11
ste||ar-github has joined #ste||ar
23:11
<
ste||ar-github >
hpx/modular_assert 2bc06e1 Thomas Heller: Adding HPX library "PP"...
23:11
<
ste||ar-github >
hpx/modular_assert 34113aa Thomas Heller: Adding HPX Library "Config"...
23:11
<
ste||ar-github >
hpx/modular_assert 2291ace Thomas Heller: Adding Assert module...
23:11
ste||ar-github has left #ste||ar [#ste||ar]
23:15
ste||ar-github has joined #ste||ar
23:15
ste||ar-github has left #ste||ar [#ste||ar]
23:17
ste||ar-github has joined #ste||ar
23:17
<
ste||ar-github >
hpx/modular_assert 26f3c38 Thomas Heller: Adding Assert module...
23:17
ste||ar-github has left #ste||ar [#ste||ar]
23:35
ste||ar-github has joined #ste||ar
23:35
<
ste||ar-github >
hpx/modular_assert 404e9d4 Thomas Heller: Adding Assert module...
23:35
ste||ar-github has left #ste||ar [#ste||ar]
23:52
ste||ar-github has joined #ste||ar
23:52
<
ste||ar-github >
hpx/modular_assert c1a8db4 Thomas Heller: Making inspect happy...
23:52
ste||ar-github has left #ste||ar [#ste||ar]