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/
Matombo has quit [Remote host closed the connection]
<jbjnr>
heller: yt?
<jbjnr>
nvm
<heller>
jbjnr: we need to skype ;)
<jbjnr>
when?
AnujSharma has joined #ste||ar
mcopik_ has joined #ste||ar
<jbjnr>
heller: when and for what - course?
<heller>
yes, course
<jbjnr>
numactl has saved us
<heller>
we could do it now
<heller>
why?
<jbjnr>
using interleaved memory brings our cholesky results up to match parsec :)
<jbjnr>
heller: how about skype 4pm?
<jbjnr>
or 3pm
bikineev has quit [Remote host closed the connection]
<heller>
jbjnr: 3pm it is
<heller>
jbjnr: ha! so it was memory bound all along!
<heller>
jbjnr: I told you so. measure the hardware perf counters ...
K-ballo has joined #ste||ar
bikineev has joined #ste||ar
hkaiser has joined #ste||ar
hkaiser_ has joined #ste||ar
hkaiser has quit [Ping timeout: 240 seconds]
bikineev has quit [Ping timeout: 248 seconds]
bikineev has joined #ste||ar
bikineev has quit [Ping timeout: 248 seconds]
pree has joined #ste||ar
<jbjnr>
heller: skype 3pm
zbyerly_ has joined #ste||ar
khuck has quit [Read error: Connection reset by peer]
khuck has joined #ste||ar
<heller>
jbjnr: gotcha
khuck has quit [Read error: Connection reset by peer]
khuck has joined #ste||ar
AnujSharma has quit [Ping timeout: 240 seconds]
hkaiser_ has quit [Quit: bye]
khuck has quit [Read error: Connection reset by peer]
khuck has joined #ste||ar
parsa has joined #ste||ar
khuck_ has joined #ste||ar
aserio has joined #ste||ar
khuck has quit [Ping timeout: 248 seconds]
eschnett has quit [Quit: eschnett]
diehlpk_work has joined #ste||ar
<aserio>
david_pfander, heller, jbjnr, khuck_, parsa: Operation Bell Meeting in 15mins
<heller>
aserio: sure thing!
zbyerly_ has quit [Ping timeout: 248 seconds]
hkaiser has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
<jbjnr>
aserio: sorry. my linux machine won't connect. tried chrome etc, just givems me html and won't run webex app stuff
<aserio>
sigh
<aserio>
thanks for trying
<jbjnr>
anyway, cscs is no longer involved in operation bell, so you don't need me any more :(
<diehlpk_work>
heller, zbyerly Added plenty of pictures from openhub and used two of them in Section 4
<hkaiser>
jbjnr: as I understood this decision was contengient on LSUs commitment in the future
<hkaiser>
did I misunderstand that?
<jbjnr>
I sent the email containing my opinions - but Thomas from CSCS is subscribed to the operation bell list still, and after he read it, he said it was time to move on, and I can't spend time on octotiger unless something changes. If that happens, please let me know
<hkaiser>
jbjnr: ok - thanks for clarifying
<jbjnr>
The problem is that we need a problem to solve and a scientist to drive it. But I wrote that in the message, so please send me a summary of discussion about it.
<hkaiser>
jbjnr: will do - I asked Dominic to respond to your email - if he doesn't, well that's a statement as well
<jbjnr>
thanks
<hkaiser>
jbjnr: I'm planning to do two things to support this
<diehlpk_work>
hkaiser, jbjnr We could use the fracture mechanics code as a problem
<hkaiser>
jbjnr: a) fund Gregor to come to LSU in November, and b) provide funding for a possible full-time position at CCT for Dominic starting next year
khuck_ has quit []
<zbyerly>
diehlpk_work, i saw that, looks good
<hkaiser>
jbjnr: not sure if I can do more
<hkaiser>
well, except general support for HPX for this
<zbyerly>
diehlpk_work, do we know approximately how long Section Two will be?
<diehlpk_work>
zbyerly, I will start to add content there now. Will let you know soon
rod_t has left #ste||ar [#ste||ar]
eschnett has joined #ste||ar
<zbyerly>
is Manycore a word
<diehlpk_work>
hkaiser, which publication do I use for agas and a general overview of hpx?
<hkaiser>
general overview: use the 'Parallelization for Local and Distributed Asynchronous Task-Based Programming'
StefanLSU has joined #ste||ar
rod_t has joined #ste||ar
<hkaiser>
agas: there is no paper yet, we just submitted one
<zbyerly>
~\cite{Kaiser:2015:HPL:2832241.2832244}
<diehlpk_work>
zbyerly, Section 2 will be approx 3 pages
RostamLog has joined #ste||ar
aserio has quit [Quit: aserio]
aserio has joined #ste||ar
StefanLSU has quit [Quit: StefanLSU]
Matombo has joined #ste||ar
rod_t has quit [Ping timeout: 246 seconds]
EverYoung has joined #ste||ar
parsa has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
pree has quit [Read error: Connection reset by peer]
<mbremer>
@hkaiser: Do you have time to chat today or are you slammed with deadlines?
<hkaiser>
mbremer: sorry, I have not forgotten...
<hkaiser>
mbremer: we can certainly talk, but I have not looked at your data yet
<mbremer>
No worries.
david_pfander has quit [Ping timeout: 264 seconds]
<mbremer>
Basically the scheduler is much more expensive than anticipated.
<hkaiser>
mbremer: only if you don't hav enough work
<hkaiser>
mbremer: but yes, I hear what you're saying
<hkaiser>
what's the average task length for you?
<mbremer>
Well so I made a bathtub plot to check for oversubscription
<hkaiser>
k
<mbremer>
and even at an average of 2000 microseconds it's still saying the 2x is the best oversubscription
<hkaiser>
what's 2000us?
<mbremer>
I think it's still better than 1x.
<hkaiser>
task length?
<mbremer>
Yeah
<hkaiser>
mbremer: have you built your app with vectorization enabled?
<mbremer>
no so that's still in the works
<hkaiser>
also, configure hpx to use upto 512 cores, this will vectorize all bitmask operations in the scheduler
<mbremer>
does Vc/1.3 support avx512 or is that coming only in version 2?
<hkaiser>
mbremer: not sure, Vir might be a better source of information about the current state of affairs
<hkaiser>
mbremer: send Vir and email if he doesn't respond here
<hkaiser>
an* email even
<mbremer>
Would you think that the vectorization would help out the scheduler a lot?
<hkaiser>
mbremer: a bit
<hkaiser>
mbremer: you don't need Vc for that, just tell the compiler to vectorize HPX
<hkaiser>
and configure hpx for max 512 cores
<hkaiser>
that gave us some speedup for pour application, don't remember how much, though
<hkaiser>
heller: do you remember?
<aserio>
twwright: yt?
<aserio>
I have a question about SC
<aserio>
nvm
mbremer has quit [Quit: Page closed]
<github>
[hpx] hkaiser closed pull request #2898: Adding missing #include and missing guard for optional code section (master...fixing_2893) https://git.io/v51wz
<zbyerly>
does anyone know what a thread pool is?
<jbjnr>
lol
<jbjnr>
zbyerly: if you leave notes in the doc, then I will add stuff later. I can devote tomorrow to writing
rod_t has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
aserio has quit [Read error: Connection reset by peer]
aserio has joined #ste||ar
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
<diehlpk_work>
jbjnr, Section 2 should be checked too
<jbjnr>
diehlpk_work: ok. I am going out for a few hours, but later and tomorrow I'll write/edit
<diehlpk_work>
Ok, enjoy. I try to finish section 2 this day
<diehlpk_work>
hkaiser, Do we have a publication for performance counters?
<hkaiser>
yes, look for Pats papers
<diehlpk_work>
atricia Grubel, Hartmut Kaiser, Kevin Huck, Jeanine Cook: Using Intrinsic Performance Counters to Assess Efficiency in Task-based Parallel Applications
aserio1 has joined #ste||ar
<diehlpk_work>
This one?
<hkaiser>
yes
aserio has quit [Ping timeout: 252 seconds]
aserio1 is now known as aserio
<heller>
jbjnr: see pm please
<diehlpk_work>
hkaiser, Any publication for parcel port?
hkaiser has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
<hkaiser>
diehlpk_work: none, we always wanted to write one, though
<diehlpk_work>
Ok, just asking for what to cite in Section 2
<hkaiser>
bibek_desktop: submitted a paper about parcel coalescing to one of the sc workshops this year
<hkaiser>
diehlpk_work: some active messages paper perhaps explaining that parcels are a better version of those
<hkaiser>
or reference sterlings ParalleX white paper, the citation should be in one of the recent bibtex files
bikineev has quit [Remote host closed the connection]
<pree>
hkaiser : I'm looking for writing a series blog post about HPX in "XRDS - crossroads, The ACM magazine for students". I have collected the requirements to take membership to write blog in XRDS, I'm looking for some interesting topics for that : ) I hope that it will be too useful for HPX to get to the vibrant student community over the world.
<pree>
Any suggestions for this ?
<pree>
Thank you
<hkaiser>
pree: some general overview first, perhaps, later you can pick specific topics
aserio has quit [Ping timeout: 246 seconds]
<pree>
Okay I will let you know when I get membership : )
<pree>
thanks
<hkaiser>
pree: cool!
<pree>
: )
aserio has joined #ste||ar
pree has quit [Ping timeout: 260 seconds]
Matombo has joined #ste||ar
StefanLSU has joined #ste||ar
<diehlpk_work>
hkaiser, Could you may have a look to Section 2.2
<hkaiser>
can do a bit later today
<hkaiser>
about to have a meeting
<diehlpk_work>
and 2.1 Lightweight User Level Tasks? I do not know how to shorten the content there
<hkaiser>
does it need shortening?
<diehlpk_work>
Yes, AGAS has serveral subsections and we have one pahe left.
<hkaiser>
k
<hkaiser>
can do later
<diehlpk_work>
1.5 in total but we have to finsih the conclusion and write an abstract too
<diehlpk_work>
Cool, thanks
Matombo has quit [Remote host closed the connection]
mbremer has joined #ste||ar
Matombo has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
Matombo has quit [Quit: Leaving]
Matombo has joined #ste||ar
Matombo has quit [Client Quit]
StefanLSU has quit [Quit: StefanLSU]
Matombo has joined #ste||ar
aserio has quit [Ping timeout: 248 seconds]
aserio has joined #ste||ar
aserio1 has joined #ste||ar
StefanLSU has joined #ste||ar
Matombo has quit [Remote host closed the connection]
Matombo has joined #ste||ar
aserio has quit [Ping timeout: 240 seconds]
aserio1 is now known as aserio
parsa has joined #ste||ar
parsa has quit [Client Quit]
Matombo has quit [Remote host closed the connection]
parsa has joined #ste||ar
parsa has quit [Client Quit]
Matombo has joined #ste||ar
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
rod_t has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rod_t has joined #ste||ar
<zbyerly>
aserio, how many studnets have we had in GSoC total
<hkaiser>
zbyerly: ask diehlpk_work
<diehlpk_work>
17
rod_t has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rod_t has joined #ste||ar
hkaiser has quit [Quit: bye]
bikineev has joined #ste||ar
akheir has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
hkaiser has joined #ste||ar
<heller>
hkaiser: you got a link failure for the partitioned vector changes