hkaiser 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/ | GSoC: https://github.com/STEllAR-GROUP/hpx/wiki/Google-Summer-of-Code-%28GSoC%29-2020
hkaiser has quit [Quit: bye]
nikunj has quit [Remote host closed the connection]
nikunj has joined #ste||ar
nikunj97 has joined #ste||ar
jaafar has joined #ste||ar
Nikunj__ has joined #ste||ar
nikunj97 has quit [Ping timeout: 258 seconds]
ronniegandhi has joined #ste||ar
mdiers_ has quit [Remote host closed the connection]
mdiers_ has joined #ste||ar
Nikunj__ has quit [Quit: Leaving]
nikunj has quit [Remote host closed the connection]
nikunj has joined #ste||ar
ronniegandhi has quit [Ping timeout: 258 seconds]
hkaiser has joined #ste||ar
mdiers_ has quit [Quit: mdiers_]
mdiers_ has joined #ste||ar
<hkaiser> mdiers_: very high idle-rate - so not enough parallelism in your code for the number of cores you're running on
<hkaiser> try to increase the task sizes or reduce the number of cores, or both
<mdiers_> the percentage should be high?
<mdiers_> hkaiser: my idea was to calculate one block of work per numa domain to reduce the number of cores
<hkaiser> mdiers_: idle-rate should be low
<hkaiser> I think thinking about numa domain-related optimizations is premature, get things efficient first
<hkaiser> if you don't have sufficient work even numa domain can't help you
<mdiers_> hkaiser: is 0.01% low? :confused
<hkaiser> 0.01% is the unit of measure, the actual value is between 4180 (41.8%) and 8064 (80.64%)
<mdiers_> ahhhh, tkz ;-)
<mdiers_> how big should a task be? is a loop over 150 items with 10 calculations each too little?
<mdiers_> hkaiser: thank you very much for the idle counter ;-)
<hkaiser> about 300 microseconds at least
<hkaiser> mdiers_: there is another counter giving you the average task length measured
<hkaiser> I think /threads/time/average or somesuch
ronniegandhi has joined #ste||ar
<mdiers_> ok, the dist example is at 4.5 us, that explains it.
<mdiers_> thanks a lot for the 300 microseconds. i will see how i get it changed.
<Yorlik> If any of you template language and type erasure experts have a moment to look at this bit of madness: https://godbolt.org/z/ZuQvx2
<Yorlik> Question is at line 109 ff
<Yorlik> I am not sure if it is possible to do it at all, probably not the way how I wrote it in the comment.
<Yorlik> However - maybe someone has a good idea.
<Yorlik> Basically I want to pull out a component typed pointer through the type erasure.
<hkaiser> why do you type erase if you still need the type?
<hkaiser> for an example
<Yorlik> Because it kills any existing indirection.
<Yorlik> Thanks for that lnk - reading up ..
<Yorlik> BTW - the type is erased to have one single type to deal with at the top level, whioch can be stored in a vector. This top level type holds a type erased pointer to the polymorphic data.
nikunj97 has joined #ste||ar
nikunj has quit [Remote host closed the connection]
nikunj has joined #ste||ar
ronniegandhi has quit [Remote host closed the connection]
nikunj97 has quit [Ping timeout: 260 seconds]
Abhishek09 has joined #ste||ar
Abhishek09 has quit [Remote host closed the connection]
Yorlik has quit [Ping timeout: 240 seconds]