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
weilewei has quit [Remote host closed the connection]
nanmiao11 has quit [Remote host closed the connection]
kale[m] has quit [Ping timeout: 244 seconds]
kale[m] has joined #ste||ar
hkaiser has quit [Quit: bye]
bita_ has quit [Ping timeout: 258 seconds]
kale[m] has quit [Ping timeout: 260 seconds]
kale[m] has joined #ste||ar
nikunj97 has joined #ste||ar
Yorlik has quit [Ping timeout: 264 seconds]
kale[m] has quit [Ping timeout: 272 seconds]
kale[m] has joined #ste||ar
Nikunj__ has joined #ste||ar
akheir has quit [Quit: Leaving]
nikunj97 has quit [Ping timeout: 256 seconds]
Nikunj__ has quit [Client Quit]
kale[m] has quit [Ping timeout: 260 seconds]
kale[m] has joined #ste||ar
<gonidelis[m]> https://github.com/STEllAR-GROUP/hpx/pull/4809 6 fails here... is any of that my fault?
<ms[m]> freenode_gonidelis[m]: the pycicle errors no, the windows github actions failures do look like they're caused by your changes though...
<ms[m]> you may be missing a HPX_EXPORT or something like that in one of your files
<gonidelis[m]> ms[m]: hmm What's the purpose of HPX_EXPORT?
hkaiser has joined #ste||ar
<ms[m]> freenode_gonidelis[m]: it tells the linker that that symbol should be public in a shared library, otherwise the symbol will be in the shared library but hidden for linking (it can still be referenced within the shared library)
<ms[m]> (note that I'm not sure that's your actual problem, I was just guessing based on the error message and the fact that windows tends to be stricter with exporting things correctly)
<hkaiser> ms[m]: I don't think this is a import/export problem
<hkaiser> I'd assume those types are header only
<hkaiser> I rather think that this is caused by a mismatch between the partitioned and non-partitioned overloads for for_each_() (note the '_')
<hkaiser> or a mismatch between declaration of said function and its definition
kale[m] has quit [Ping timeout: 272 seconds]
kale[m] has joined #ste||ar
<hkaiser> ms[m]: would you mind if we merged #4527 at some point?
<ms[m]> hkaiser: no, not at all
<ms[m]> I was just waiting for it to be ok to merge
<hkaiser> I fixed the conflict now
<ms[m]> sorry I didn't comment anything on it
<hkaiser> no worries
<ms[m]> ok, thanks
<ms[m]> then let's merge it once ci cycles
<hkaiser> ok
<gonidelis[m]> hkaiser: I can see two overloads of `for_each_()`: the segmented and the non-segmented one. Do you think that I should fix sth here in order for the tests to pass https://github.com/STEllAR-GROUP/hpx/blob/2d5e9a1877fdf1b5a7a9ece03361ae6c74263d85/libs/algorithms/include/hpx/parallel/algorithms/for_each.hpp#L230 ?
<hkaiser> gonidelis[m]: not sure, I have not looked closely - all I know is that that the overloads should match (except for the is_segmented argument)
<hkaiser> also, the function is declared and defined in different places, those should match as well
<gonidelis[m]> ok should I try change the segmented version(s) and see if the tests pass? If that does not work we could rewind back to the last existing commit?
<hkaiser> gonidelis[m]: I don't see an implementation for for_each_() (non-segmented) anymore
<hkaiser> ahh, got it now
<gonidelis[m]> I have not touched the segmented one
nanmiao11 has joined #ste||ar
<gonidelis[m]> hkaiser: Plus, I think I need to use `is_sentinel_for`. Seems like I missed it. Is that right?
<hkaiser> gonidelis[m]: is_sentinel_for would just properly constrain the arguments, it's not causing the linker error
<gonidelis[m]> yeah I know. I am just asking wether I should add it in general, despite the errors.
<hkaiser> gonidelis[m]: wrt the segmented version of for_each_: you might want to add the second template argument there (without changing the implementation), that could resolve the linker problem
<gonidelis[m]> hkaiser: will do it right now
<gonidelis[m]> hkaiser: ok just pushed. let's wait for the tests and see
<gonidelis[m]> question: Is `is_sentinel_for` needed here https://github.com/STEllAR-GROUP/hpx/blob/b847d94f562d23f968a9542972a8f3b6211bcac9/libs/algorithms/include/hpx/parallel/algorithms/reduce.hpp#L92 since `reduce_` is being called by `reduce` (#line 342) who already carries the `is_sentinel_for` constraint?
karame_ has quit [Remote host closed the connection]
nikunj97 has joined #ste||ar
<hkaiser> gonidelis[m]: we don't need to protect our internal implementation functions
<hkaiser> only the user-facing API functions
<gonidelis[m]> hkaiser: thanks
<gonidelis[m]> hkaiser: Just added `is_sentinel_for`. I don't think that there is any other fix that needs to be done. Should I work on some other algo as we wait for the tests to be completed?
<hkaiser> as you wish
<gonidelis[m]> hkaiser: any suggestions?
<hkaiser> any iteration based algo: copy, transform, etc.
<gonidelis[m]> I pick from here
<gonidelis[m]> right?
nanmiao11 has quit [Remote host closed the connection]
karame_ has joined #ste||ar
nanmiao11 has joined #ste||ar
rtohid has joined #ste||ar
bita_ has joined #ste||ar
diehlpk__ has joined #ste||ar
<ms[m]> hkaiser: do you mind if I merge the jenkins PR?
<hkaiser> ms[m]: pls go ahead
<ms[m]> thanks
<ms[m]> let's see how overwhelmed daint becomes from that ;)
<hkaiser> :D
weilewei has joined #ste||ar
<K-ballo> why is `any` in the serialization module?
<hkaiser> K-ballo: ask ms[m]
<ms[m]> just because the serializable variants of it are there, and it depends on the datastructures module
<ms[m]> it could just as well have been in the include module
<ms[m]> K-ballo ^
<ms[m]> i.e. a bit out of necessity and a bit for lack of a better place
diehlpk__ has quit [Remote host closed the connection]
diehlpk__ has joined #ste||ar
kale[m] has quit [Ping timeout: 256 seconds]
kale[m] has joined #ste||ar
Yorlik has joined #ste||ar
diehlpk__ has quit [Ping timeout: 260 seconds]
akheir has joined #ste||ar
rtohid has quit [Ping timeout: 245 seconds]
<weilewei> So on Daint, DCA+hpx runs fine on Debug mode, while Release mode failed
<weilewei> I used an old commit from May which can supports HPX MPI Future, because current hpx master doesn't compile when enable hpx mpi future
<weilewei> commit d5627bec7efa8688f61bb450e41edc24b8d316d4
<weilewei> This commit works fine on Summit with Release mode though...
<hkaiser> weilewei: please report problems with master
<weilewei> hkaiser ok
<hkaiser> it doesn't help anybody if you silently work around issues
<weilewei> Sure... I will report it
<hkaiser> thanks
<hkaiser> thansk
wash[m] has joined #ste||ar
nikunj has quit [Ping timeout: 244 seconds]
nikunj97 has quit [Ping timeout: 272 seconds]
nikunj has joined #ste||ar
zao_ has joined #ste||ar
nikunj97 has joined #ste||ar
nikunj97 has quit [Read error: Connection reset by peer]
hkaiser has quit [Quit: bye]
kale[m] has quit [Ping timeout: 246 seconds]
<weilewei> ok, so after replace openblas with netlib-lapack in DCA on Piz Daint, hpx runs fine. I suspect some threading issues behind it. So netlib-lapack might be our good friend. Similarly, essl is not working with hpx on Summit
kale[m] has joined #ste||ar
nikunj97 has joined #ste||ar
Nikunj__ has joined #ste||ar
nikunj97 has quit [Ping timeout: 260 seconds]
nikunj has quit [Read error: Connection reset by peer]
nikunj has joined #ste||ar
kale[m] has quit [Ping timeout: 260 seconds]
kale[m] has joined #ste||ar
<bita_> is there a flag to suppress hpx deprecated warnings?
<K-ballo> depends on the deprecated thing
<bita_> headers?
<weilewei> HPX_WITH_DEPRECATION_WARNINGS ?
<bita_> warning: "The header hpx/runtime/get_num_localities.hpp is deprecated, please include hpx/runtime.hpp instead" [-W#warnings]
<bita_> #warning "The header hpx/runtime/get_num_localities.hpp \
<bita_> weilewei, I tried that, but didn't work :o
<K-ballo> can't you just fix the code?
<weilewei> bita_ if it doesn't stop compilation, does it matter?
<bita_> it stops compilation when I use ninja :/
<K-ballo> HPX_RUNTIME_LOCAL_HAVE_DEPRECATION_WARNINGS
<K-ballo> it only stops compilation if you ask for compilation to stop on warnings
<K-ballo> the cmake var is HPX_RUNTIME_LOCAL_DEPRECATION_WARNINGS
<bita_> K-ballo, how can i fix that?
<weilewei> lol, never use ninja, only tried cmake... sorry
<K-ballo> "please include hpx/runtime.hpp instead"
Nikunj__ has quit [Ping timeout: 256 seconds]
<bita_> thank you
<bita_> weilewei, thanks :)
Nikunj__ has joined #ste||ar
<K-ballo> weilewei: one would use ninja as a cmake generator
<K-ballo> you should give it a try, it better exploits build parallelism
hkaiser has joined #ste||ar
<weilewei> K-ballo ok, that's good to know
<weilewei> I don't think cmake can build things in parallel
<K-ballo> cmake doesn't build things
<weilewei> :') it configures things
<hkaiser> bita_: that is a warning generated by blaze
<K-ballo> ninja is a generator like make is, or visual studio / xcode
<bita_> how can I get rid of it?
<hkaiser> the best is to disable warnings from that module, as K-ballo said, define HPX_RUNTIME_LOCAL_DEPRECATION_WARNINGS=Off
<weilewei> K-ballo oh, I see
<bita_> :+1
<K-ballo> so you cmake -DGENERATOR="Ninja" .., then run ninja instead of make
<K-ballo> that's not right.. what's the right incantation?
<K-ballo> cmake -G"Ninja" ..
<weilewei> just tried, hmm it seems it generates very clean build message, only one line
<weilewei> it seems pretty fast in building same library
<K-ballo> it overwrites uninteresting output
<weilewei> K-ballo thanks, it is a good tool, I like it
nikunj97 has joined #ste||ar
Nikunj__ has quit [Ping timeout: 258 seconds]
Nikunj__ has joined #ste||ar
nikunj97 has quit [Ping timeout: 244 seconds]
nikunj97 has joined #ste||ar
zao_ is now known as zao
nikunj97 has quit [Client Quit]
Nikunj__ has quit [Read error: Connection reset by peer]
Yorlik_ has joined #ste||ar
Yorlik has quit [Ping timeout: 272 seconds]