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
<gonidelis[m]>
hkaiser: what partition do I use in rostam to run cuda?
<gonidelis[m]>
bhumit: let me check
<gonidelis[m]>
bhumit: generic histrogram is sth we still looking into implementing
<gonidelis[m]>
for sure you could give it a shot
<hkaiser>
bhumit[m]: please ask questions if you're unsure what this is all about
K-ballo has joined #ste||ar
<gonidelis[m]>
hkaiser: pm
jehelset has quit [Ping timeout: 250 seconds]
<bhumit[m]>
gonidelis: hkaiser will do, thank you
<gdaiss[m]>
@gonidelis Try diablo (or the cuda-V100 partition in general)
<gonidelis[m]>
Thanks man
<gnikunj[m]>
Didn't think we were still doing the histogram stuff
<gnikunj[m]>
gonidelis: optimizing cuda for V100s now? :P
<gonidelis[m]>
gnikunj: well projects are always open right? we just need people to take them on
<gnikunj[m]>
I guess Kamand has A100s
<gonidelis[m]>
gnikunj: why is V100 specificly interesting?
<gonidelis[m]>
;p
<gnikunj[m]>
gonidelis: fair point. Just that histogram project has been here for eternity. Thought hkaiser closed it or something.
<gnikunj[m]>
gonidelis: nothing specific. Just that it's too powerful for what you're testing :P
<gonidelis[m]>
lol
<bhumit[m]>
> gonidelis: fair point. Just that histogram project has been here for eternity. Thought hkaiser closed it or something.
<bhumit[m]>
Yes, the issue is closed as well which is why I confirmed it here once. Seems like it is so will be looking into it and maybe choose it as my GSoC project if it looks doable.
<bhumit[m]>
Also, GSoD 2022 has been announced. Are we participating?
<gonidelis[m]>
bhumit: most probably yes
<gnikunj[m]>
bhumit: did you try building HPX? We did set up a page to make it easier for GSoC participants to get used to HPX. gonidelis should have that
<bhumit[m]>
gonidelis[m]: Nice
<gonidelis[m]>
gnikunj: it's on the gsoc wiki on the top. it's not about building particularly
<gnikunj[m]>
hkaiser: do you know of a general distributed performant matrix multiplication algorithm? All of the one's I know of use square grids (either padded or non padded).
<gnikunj[m]>
(I've seen SUMMA, cannon and fox but I'm searching for more performance)
<bhumit[m]>
gnikunj[m]: Yup. Tried it on M1 too, btw. I have been making (documentation related) contributions since September '21 (IIRC)
<gnikunj[m]>
That's great!
<gonidelis[m]>
bhumit: nice!
<bhumit[m]>
s/(documentation related)//
<bhumit[m]>
* Yup. Tried it on M1 too, btw. I have been making contributions since September '21 (IIRC)
<bhumit[m]>
E: Oct '21 apparently
jehelset has joined #ste||ar
jbalint has quit [Ping timeout: 252 seconds]
jbalint has joined #ste||ar
<hkaiser>
gonidelis[m]: sorry was away, see pm, pls
<K-ballo>
hkaiser: what was decided wrt hpx-local?
<K-ballo>
I see the repo has been archived
<hkaiser>
K-ballo: it has been merged back
<hkaiser>
K-ballo: also, ms[m] forked hpx-local for internal use