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
<K-ballo>
a more pressing reason to do so are ref-like types
diehlpk_work has quit [Remote host closed the connection]
<gonidelis[m]>
const & is a ref like type
<gonidelis[m]>
?
<srinivasyadav227>
hkaiser: Did SIMD Meeting got postponed to 4th of March ?
<hkaiser>
srinivasyadav227: yes, sorry
<srinivasyadav227>
hkaiser: cool, no worries, just want to confirm it :)
<hkaiser>
srinivasyadav227: btw, even if the meeting next tuesday is cancelled on the calendar (it's a holiday here), I'd be happy to meet anyways
<srinivasyadav227>
thanks!
<hkaiser>
even if it's just for a couple of minutes to touch bases
<srinivasyadav227>
hkaiser: yes, I want to meet if possible :)
<hkaiser>
ok, I'll put it back on, then
<srinivasyadav227>
I got list of things to say xD
<srinivasyadav227>
thanks :)
<hkaiser>
done
<hkaiser>
srinivasyadav227: did your exam go well?
<srinivasyadav227>
hkaiser: yup, i did well, all of them! :)
<hkaiser>
nice!
<srinivasyadav227>
I am back on working on EVE since few days, and made some progress with scalar types as-well, and I am working on benchmarks now!
<hkaiser>
cool!
<hkaiser>
srinivasyadav227: the meeting tomorrow was cancelled because Joel asked to postpone it
<hkaiser>
I had to move it to 10CST however, that's one hour later than usual
<hkaiser>
just for this one meeting
<srinivasyadav227>
hkaiser: sure, np, It works for me :)
<hkaiser>
thanks
K-ballo has quit [Quit: K-ballo]
hkaiser has quit [Quit: Bye!]
hkaiser has joined #ste||ar
heller[m] has joined #ste||ar
K-ballo has joined #ste||ar
<mdiers[m]>
<mdiers[m]> "The asynchronous execution..." <- After a few more tests I think it has something to do with mpi.
<mdiers[m]>
With `--hpx:ini=hpx.parcel.mpi.enable!=0 --hpx:ini=hpx.parcel.tcp.enable!=1` it works, with `--hpx:ini=hpx.parcel.mpi.enable!=1 --hpx:ini=hpx.parcel.tcp.enable!=0` there is partly no is_ready.
<hkaiser>
interesting
<hkaiser>
can we reproduce your problem somehow?
<mdiers[m]>
Is there maybe a tip where I can put some logs for the mpi stuff?
<hkaiser>
I find it to be unlikely that mpi should add problems, but - hey - everything is possible
<mdiers[m]>
hkaiser: Yes, we have found a flow with which it is reproducible. We have had the problem for a while but have never been able to reproduce it reliably.
<hkaiser>
mdiers[m]: can you share that?
<mdiers[m]>
hkaiser: Ohh sorry, I read over the we.
<mdiers[m]>
I am glad that I was able to reproduce it at all.
<mdiers[m]>
It is very runtime dependent, as soon as I change something at the data or parameters it no longer occurs so reliably.
<mdiers[m]>
Stupid question: How can I turn on the mpi_debug for logging?
<mdiers[m]>
* for logging? `--hpx:debug-parcel-log` ?
<hkaiser>
in mpi itself?
<mdiers[m]>
hkaiser: No, I only found a mpi_debug in mpi_future.
<mdiers[m]>
Is there already a date for a version 1.8?
<hkaiser>
mdiers[m]: somewhere in March/April timeframe
<mdiers[m]>
hkaiser: thanks
<hkaiser>
mdiers[m]: are you testing with HPX master? or do you use V1.71?
<mdiers[m]>
<hkaiser> "mdiers: are you testing with HPX..." <- Currently 1.7.1. Should I update to the master?
diehlpk_work has joined #ste||ar
<hkaiser>
mdiers[m]: might be a good idea to at least test with master to avoid surprises after the release
<mdiers[m]>
ok, I start the build of the environment on monday and test it.
nanmiao has joined #ste||ar
diehlpk_work has quit [Ping timeout: 256 seconds]
diehlpk_work has joined #ste||ar
<gonidelis[m]>
hkaiser: does openMP expose in any way data dependency graphs?
<hkaiser>
openmp has a tooling API that allows for externl tools to hook into it, APEX is one of those tools (works without HPX), not sure though if it can produce dependency graphs
<gonidelis[m]>
hkaiser: I am trying to figure out whether openMP does actually rely on a dep. graph logic or if it tries to hide it underneath