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/ | GSoD: https://developers.google.com/season-of-docs/
hkaiser has quit [Ping timeout: 250 seconds]
hkaiser has joined #ste||ar
<hkaiser> Yorlik: you easily create your own HPX perf counters
<hkaiser> that's essentially just a simple function that needs to be registered with hpx
<Yorlik> Yes, I know that. The question is how do I get them ASAP out of the app to the logging / monitoring backend
<hkaiser> perf counters are hpx components so they are available over the wire
<Yorlik> Like - if I want to instrument a certain amount I want to minimize any negative performance impact.
<hkaiser> do you need logging or performance information?
<Yorlik> I guess the HPC perf counters are not very lowlevel
<Yorlik> Yes
<Yorlik> I want life monitoring and drive my development and architecture by what I find
<Yorlik> You know I'm a performance junkey
<Yorlik> So - I want to measure really a lot
<hkaiser> perf counters need to be queried actively
<Yorlik> But not only for development
<hkaiser> i.e. it's a pull model
<hkaiser> logging is more of a push model
<Yorlik> prometheous allows both: pull and push
<Yorlik> So i could just use the pull library to get the counters out
<hkaiser> I don't know anything about this
<Yorlik> But I'd really like to decouiple that from any backend being up
<Yorlik> My idea is make this a log
<Yorlik> And analyze it later
<Yorlik> Or to allow both - live and offline analysis
<hkaiser> how often do you need this information to be sampled?
<Yorlik> It depend s a lot
<Yorlik> If I want to log function calls at certain points it might be very often - that would be development only ofc
<Yorlik> Stuff like amount of objects created / destroyed
<Yorlik> bandwidth of the update loop / per object type
<hkaiser> that's a typical perf counter
<Yorlik> event statistics
<Yorlik> I'm pretty sure the perf counters will bring me a long eay
<hkaiser> you sample it every other couple of ms or so
<Yorlik> But for development I'll also do finction traces
<hkaiser> that's a different thing
<Yorlik> yep
<Yorlik> It would be a macro conditional compile
K-ballo has quit [Quit: K-ballo]
hkaiser has quit [Quit: bye]
JClave has joined #ste||ar
JClave has quit [Remote host closed the connection]
quaz0r has quit [Ping timeout: 248 seconds]
quaz0r has joined #ste||ar
K-ballo has joined #ste||ar
Yorlik has quit [Read error: Connection reset by peer]
Yorlik has joined #ste||ar
Yorlik has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
K-ballo1 has joined #ste||ar
K-ballo has quit [Read error: Connection reset by peer]
K-ballo1 has quit [Ping timeout: 244 seconds]
K-ballo has joined #ste||ar
nikunj has joined #ste||ar
nikunj has quit [Ping timeout: 276 seconds]
nikunj has joined #ste||ar
hkaiser has quit [Quit: bye]