K-ballo 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/
hkaiser has joined #ste||ar
chuanqiu has joined #ste||ar
chuanqiu has quit [Client Quit]
hkaiser has quit [Quit: bye]
parsa has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
parsa has joined #ste||ar
parsa has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
parsa has joined #ste||ar
tiagofg[m] has quit [Ping timeout: 245 seconds]
tiagofg[m] has joined #ste||ar
<gonidelis[m]> ms[m]: i was lookinf for `--hpx::threads` i think
<gonidelis[m]> it's kind of confusing that their only difference is the double colon
hkaiser has joined #ste||ar
<gonidelis[m]> hkaiser: how do you use `hpx::threads` cmd line arg?
<gonidelis[m]> `--hpx::threads=16` does not work
<gonidelis[m]> neither `--hpx::threads 16`
<gonidelis[m]> oh, it's not a cmd line arg.... ?
<gonidelis[m]> hkaiser: With `--hpx:threads` I cannot more threads than my machine supports
<hkaiser> gonidelis[m]: it's --hpx:threads (on colon)
<hkaiser> one*
<gonidelis[m]> i thought there was another option that would allow me to invoke more threads
<hkaiser> more threads? more than what?
<hkaiser> you use --hpx:threads=N to create N threads
<gonidelis[m]> more threads than my laptop supports
<hkaiser> sure, there is no limit
<gonidelis[m]> I have a 16 core cpu
<hkaiser> N can be larger than the number of cores you have, things will slow down massively if you do, though
<gonidelis[m]> ahh strange, I get this when I try to "specified number of threads (32) is larger than number of available processing units (16): HPX(bad_parameter)"
<hkaiser> ahh, so we do check for that after all ;-)
<hkaiser> it does not make sense to do so anyways
<gonidelis[m]> yeah ok
<gonidelis[m]> I just wanted to show the bad performance beyond that num of cores
gonidelis2 has joined #ste||ar
gonidelis2 has quit [Client Quit]
gonidelis has joined #ste||ar
<gonidelis> hpx::max is quite slower than Cilk's max. They just use a simple `cilk_for` and a hyperobject to avoid races
<hkaiser> you mean max_element?
<gonidelis[m]> hkaiser: pm
<hkaiser> I never measured it's performance, frankly - so it has never been optimized in any ways
gonidelis has quit [Quit: Connection closed]
<hkaiser> gonidelis[m]: I would use a bigger array for measurements, at least factor of 10, if not 100
<gonidelis[m]> wow
<gonidelis[m]> say 160m
<gonidelis[m]> it's just that I am running 100 times each test and I am averaging it out
<gonidelis[m]> I guess for that high number of elements I could put test_count say equal to 5
<gonidelis[m]> hkaiser: ^^
gonidelis[m]23 has joined #ste||ar
gonidelis[m]23 has quit [Client Quit]
gonidelis[m] is now known as YourNick[m]
<hkaiser> YourNick[m]: please see here: https://github.com/STEllAR-GROUP/hpx/pull/5372, it's the best I could do in a rush (gives me 10-15% improvement for par with one thread)
<YourNick[m]> hkaiser: thank you so much
<YourNick[m]> let me check it
parsa has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
parsa has joined #ste||ar
<YourNick[m]> hkaiser: branch gives me this error
<YourNick[m]> the branch ^^
<hkaiser> uhh, that's unrelated
<hkaiser> you sure you have no hpx installed somewhere?
<hkaiser> no, it's a genuine problem if HPX_LOGGING=OFF
<hkaiser> YourNick[m]: try that PR again, I think I fixed it
<hkaiser> we don't test with LOGGING disabled, hrmm
<YourNick[m]> hkaiser: thanks!!!
<YourNick[m]> it works!!
<YourNick[m]> hkaiser: TWICE FASTER!!!!!!!!
<YourNick[m]> WHAT!? HOW?
<YourNick[m]> wow!
<hkaiser> nice
parsa has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
parsa has joined #ste||ar
parsa has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
parsa has joined #ste||ar