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/
<Yorlik> hkaiser: yt?
<hkaiser> here
<Yorlik> Hello! Up for another run on the Macros?
<Yorlik> Allright - gotta run.
* Yorlik heads out.
hkaiser has quit [Quit: bye]
vamatya has quit [Ping timeout: 248 seconds]
nikunj has quit [Remote host closed the connection]
nikunj has joined #ste||ar
Amy1 has quit [Quit: WeeChat 2.2]
Amy has joined #ste||ar
Amy has quit [Client Quit]
Amy has joined #ste||ar
Amy is now known as Guest88210
rori has joined #ste||ar
jbjnr_ has quit [*.net *.split]
tarzeau has quit [*.net *.split]
nikunj has quit [*.net *.split]
jbjnr has joined #ste||ar
nikunj has joined #ste||ar
hkaiser has joined #ste||ar
<hkaiser> simbergm: yt?
<simbergm> hkaiser: here
<hkaiser> hey
<hkaiser> yesterday I noticed a minor inconsistency in how we structure various modules
<hkaiser> some modules have their associated traits directly in the module directory, some have those in hpx/traits
<hkaiser> where should we put those?
<hkaiser> hpx/traits/? hpx/foo/? or peraps hpx/foo/traits/?
<hkaiser> simbergm: ^
<simbergm> mmh, true
<simbergm> we never really discussed or decided on anything on this
<simbergm> rori ^
<simbergm> I think hpx/modulename/traits sounds best
<simbergm> opinions?
<simbergm> hkaiser: ^
<hkaiser> I think so too
<hkaiser> hpx/modulename/traits sounds best to me
<hkaiser> second best option would be hpx/traits (i.e. leave the relative path as is, just move the file)
<rori> same !
<hkaiser> let's create a ticket
<hkaiser> to track the required changes
<simbergm> sure
<simbergm> you'll open it?
<hkaiser> I'm at fault here mostly, sorry for not discussing this earlier
<hkaiser> will do
<simbergm> no problem, we should do some thinking as well ;)
<simbergm> thanks for bringing it up
<simbergm> hkaiser: I replied on 4050, were there other tests on pycicle that you saw fail?
<simbergm> I think it looks okay to merge (after a bit of waiting...)
<hkaiser> simbergm: yah, I resolved some conflicts just now, let's see
<hkaiser> simbergm, rori: #4053
<simbergm> hkaiser: thanks!
<simbergm> btw, does the resiliency module need reviewing? nikunj is still dealing with c++11 compatibility?
<hkaiser> simbergm: yes, nikunj is working on the c++ compatibility, otherwise this is good to go
<simbergm> all right, nice
<rori> And any reason why we called our modules target hpx_<name> instead of <name>_module ?
<zao> Slightly less chance of collision with the consuming application, or are they scoped somehow?
<simbergm> zao is right, although it doesn't matter too much right now since we don't install the modules
<simbergm> but this way the libraries end up being called libhpx_modulename and when we eventually install (some) modules they'll be namespaced
<simbergm> adding a module suffix seems redundant since they already have a lib prefix (at least on linux)
<Yorlik> First graphical output in a really looong time I generated when playing with Hilbert Curves today: https://imgur.com/a/OFmziRy
<Yorlik> I really missed messing with graphic - should do it more - it's just so much fun :)
<rori> right
<hkaiser> Yorlik: nice
aserio has joined #ste||ar
hkaiser has quit [Quit: bye]
<nikunj> simbergm: I will try to finish everything by the weekend, I'm having some trouble making it C++11 compatible
<simbergm> nikunj: no rush! I was just curious about the state of that PR
hkaiser has joined #ste||ar
<hkaiser> simbergm: I did look at the serialization stuff yesterday
<hkaiser> simbergm: I think I know what needs to be done, so I'd suggest to first do some refactoring (akin what you're doing with threads) and only after that extract it into a module
<hkaiser> I believe the resulting module wouldn't depend on anything but config etc.
<hkaiser> also, on an unrelated note, I did the bost library detection for the program_options module in the module itself
<hkaiser> this might be doable for the filesystem module as well (i.e. remove it from SetupBoost.cmake)
<simbergm> yes, and yes! I saw your changes for program options and that's a really good direction
<simbergm> I can update the filesystem module to do the same
<simbergm> and thanks for having a look at serialization
<simbergm> how are you planning on getting rid of naming/agas/etc.?
<hkaiser> externalize it
<simbergm> wait, does that mean something specific?
<hkaiser> we already have the actual functionality in the preprocess helper
<hkaiser> no, I want to separate the serialization from a) handling futures and b) handling id_types
<hkaiser> all of this can be done externally to the base serialization
<hkaiser> all we need is a opaque pointer stored in the archive
jaafar has quit [Ping timeout: 276 seconds]
<simbergm> okay, sounds like you're on top of things
<simbergm> this is why I appreciate your help on those parts of the code base!
<rori> hkaiser: if you have time, could you have a look at the travisci error on windows for the cmake refactoring PR, cause I think it almost ready to go, and I couldn't find the reason for this error on windows from my linux env :/
aserio has quit [Ping timeout: 276 seconds]
<heller> hkaiser: I'd rather get rid of the special handling entirely
<heller> The special handling for the corner case when gids are being split isn't only in the serialization layer but also bubbles up in the parcelport
Yorlik has quit [Read error: Connection reset by peer]
aserio has joined #ste||ar
heller has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
heller has joined #ste||ar
K-ballo1 has joined #ste||ar
K-ballo has quit [Ping timeout: 244 seconds]
K-ballo1 is now known as K-ballo
aserio has quit [Ping timeout: 264 seconds]
USTOBRE has quit [Quit: Connection closed for inactivity]
<K-ballo> hkaiser: yt?
rori has quit [Quit: bye]
aserio has joined #ste||ar
<hkaiser> K-ballo: here now
<K-ballo> hlaiser: see pm
Yorlik has joined #ste||ar
<Yorlik> hkaiser: yt?
hkaiser has quit [Quit: bye]
aserio has quit [Ping timeout: 258 seconds]
aserio has joined #ste||ar
aserio has quit [Quit: aserio]
hkaiser has joined #ste||ar
<Yorlik> hkaiser: yt?
<hkaiser> Yorlik: here
<Yorlik> Up for some tedious work? ;)
<Yorlik> hkaiser: ^
<hkaiser> Yorlik: not now, sorry
<Yorlik> OK. I'll work on my Quadtree then :)
<Yorlik> hkaiser: Just realized you were involved in the creation of the Boost R-Tree. Would it be hard to modify this R-Tree to utilize Hilbert Curves to speed up insertions and relocation of entries (moving objects)?
<Yorlik> I'm a bit torn wether to use Quadtrees or R-Trees for my spatial search.
<hkaiser> quadtrees are difficult to use if you have anything but point data
<hkaiser> Yorlik: ^^
<Yorlik> So - Since Objects have bounding boxes or capsules --> R-Tree? What about using Hilbert Curves?
<Yorlik> Since many objects will move there will be lots of changes
<hkaiser> why hilbert curves?
<Yorlik> I was thinking about dividing the objects into grounded (= 2.5D ) and flying (full 3D) ones.
<hkaiser> they may give you proximity, but only in certain directions
<Yorlik> Hilbert Curves for faster reindexing / insertion
<hkaiser> nah
<hkaiser> nobody has shown that hilbert curves are actually faster
<hkaiser> they are just cooler ;-)
<Yorlik> lol
<Yorlik> How would you optimize an R-Tree for frequent reindexing of moving objects?
<Yorlik> My idea still is to use fixed size, relatively small buckets at the lowest level / leaf level
<Yorlik> Like 16*16 m or so
<Yorlik> maybe smaller
<Yorlik> Imagine a crowd of moving bounding boxes = players/animals/monsters
<hkaiser> rtrees are fairly fast
<zao> Spatial hashes are nice too, especially if you have a good idea of the granularity you want to query with.
<hkaiser> sure, but quadtrees are bad if you have bboxes
<Yorlik> I think we will use bboxes and capsiles mostly
<hkaiser> sure
<hkaiser> long time ago I developed a special quadtree-like structure for handling bboxes
<hkaiser> essentially a 4d quadtree
<hkaiser> extremely fast
<hkaiser> in 4d a bbox is representable by a 4d point
<Yorlik> 4d Quadtree sounds like a generalization beyond the Octree
<hkaiser> yes