hkaiser changed the topic of #ste||ar to: Notice: We will move this channel to Libra.Chat (irc://irc.libera.chat:6667) soon. Please /join #ste||ar on that network. Please ask here if you need help. This channel here will stay available for a while.
RostamLog has joined #ste||ar
<hkaiser>
akheir: yay!
<akheir>
Testing the irclogger
<hkaiser>
thanks for reviving RostamLog
<akheir>
:D
<akheir>
yeah, works
<hkaiser>
I'll add the link back to the /topic here
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 | Colloquium on Artificial Intelligence Research and Optimization (cairo.stellar-group.org)
<jedi18[m]>
hkaiser: For the sentinel overload for exclusive scan, should it have a Sent2 variable denoting the end of the output range?
<hkaiser>
jedi18[m]: aren't there different overloads, let me have a look
<hkaiser>
jedi18[m]: well, the existing overloads all assume that the output sequence is large enough, no end iterator is supplied there
<hkaiser>
you probably want to have something like ranges::exclusive_scan(R&& r, O result, T init, BinOp op)
<jedi18[m]>
Oh ok thanks, the reason I’m asking is some algorithms like starts_with don’t have a last2 variable for the iterative overloads but have them for the sentinel overloads
<jedi18[m]>
Ok yeah I’ll use that, thanks!
<jedi18[m]>
Iterator*
<hkaiser>
jedi18[m]: let's do whatever makes sense, align ourselves to other algorithms
K-ballo has quit [Ping timeout: 258 seconds]
<gonidelis[m]>
hkaiser:
<gonidelis[m]>
1000000, 2, 0.181388
<gonidelis[m]>
500000, 1, 0.373336
<gonidelis[m]>
wtf
<hkaiser>
what are those numbers?
<gonidelis[m]>
it get's better ;p
<gonidelis[m]>
n.elements, cores, time[ms[
<gonidelis[m]>
[ms] *
<hkaiser>
hmmm, doesn't sound right
<gonidelis[m]>
it's not.... it's very funny because if I weak scale for 10k, 20k it's normal
<hkaiser>
that would mean a speedup of 4 when doubling the number of cores
<gonidelis[m]>
but just for 5k, 10k it gets better
<gonidelis[m]>
the thing is that I am doubling the n.elements too
<hkaiser>
ahh, it's weak scaling
<hkaiser>
cache effects?
<hkaiser>
doesn't sound right either way
<gonidelis[m]>
cache is what i suspected
<gonidelis[m]>
let me tweak it
<gonidelis[m]>
hkaiser: is it ok for you if I pass you the thesis document later in the day so you could check 1. the hpx paragraph (~4 pages) 2. the results paragraph (4algos x 3graphs = 12 diagrams)
<hkaiser>
ok
<gonidelis[m]>
hkaiser: thanks
K-ballo has joined #ste||ar
akheir has quit [Quit: Leaving]
<gonidelis[m]>
hkaiser: pm
RostamLog has joined #ste||ar
jedi18[m] has quit [Remote host closed the connection]
gnikunj[m] has quit [Remote host closed the connection]
srinivasyadav227 has quit [Read error: Connection reset by peer]
sestro[m] has quit [Read error: Connection reset by peer]
KordeJong[m] has quit [Remote host closed the connection]
Deepak1411[m] has quit [Write error: Connection reset by peer]
rachitt_shah[m] has quit [Remote host closed the connection]
mdiers[m] has quit [Remote host closed the connection]
heller[m] has quit [Remote host closed the connection]
gdaiss[m] has quit [Remote host closed the connection]
ms[m] has quit [Remote host closed the connection]
gonidelis[m] has quit [Remote host closed the connection]
pedro_barbosa[m] has quit [Remote host closed the connection]
rori[m] has quit [Remote host closed the connection]