aserio 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/
kisaacs has joined #ste||ar
EverYoung has quit [Ping timeout: 245 seconds]
Anushi1998 has quit [Quit: Leaving]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
katywilliams has joined #ste||ar
kisaacs has quit [Ping timeout: 260 seconds]
katywilliams has quit [Ping timeout: 240 seconds]
EverYoun_ has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoun_ has quit [Read error: Connection reset by peer]
EverYoung has joined #ste||ar
kisaacs has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
eschnett has joined #ste||ar
daissgr has quit [Ping timeout: 268 seconds]
prashantjha has joined #ste||ar
hkaiser has quit [Quit: bye]
katywilliams has joined #ste||ar
katywilliams has quit [Ping timeout: 260 seconds]
eschnett has quit [Quit: eschnett]
K-ballo has quit [Quit: K-ballo]
katywilliams has joined #ste||ar
katywilliams has quit [Ping timeout: 240 seconds]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 264 seconds]
sharonhsl has joined #ste||ar
katywilliams has joined #ste||ar
EverYoung has joined #ste||ar
katywilliams has quit [Ping timeout: 240 seconds]
EverYoung has quit [Ping timeout: 276 seconds]
kisaacs has quit [Ping timeout: 260 seconds]
katywilliams has joined #ste||ar
hkaiser has joined #ste||ar
katywilliams has quit [Ping timeout: 264 seconds]
mcopik_ has quit [Ping timeout: 248 seconds]
kisaacs has joined #ste||ar
sharonhsl has left #ste||ar [#ste||ar]
Anushi1998 has joined #ste||ar
nanashi55 has quit [Ping timeout: 260 seconds]
nanashi55 has joined #ste||ar
Anushi1998 has quit [Quit: Leaving]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
hkaiser has quit [Quit: bye]
prashantjha has quit [Ping timeout: 245 seconds]
kisaacs has quit [Ping timeout: 264 seconds]
jaafar has quit [Ping timeout: 264 seconds]
anushi has quit [Ping timeout: 245 seconds]
anushi has joined #ste||ar
karan has joined #ste||ar
<karan> Hello Everyone, My name is karan and i am a 3rd year graduate from ggsipu. I have a good experience with c++ and python with familiarity with numpy and pandas and deep learning techniques. I would like to contribute to Augment CSV file or other interesting projects. Would someone tell me about about the next steps please or help me connect with potential mentors.
katywilliams has joined #ste||ar
katywilliams has quit [Ping timeout: 240 seconds]
<jbjnr> karan: "Augment CSV file" - I'm not familiar with that - is it one of our gsoc project ideas?
<jbjnr> thanks. That's on of our older suggestions. Pat who suggested it, is no longer active in HPX, so I suspect it might be tricky to find a mentor. It overlaps with some work I want on trace view generation etc. simbergm is also mentioned as a mentor - he should be online soon, so let's see if he replies with anything useful ...
<jbjnr> simbergm: btw - I upgraded the future_overhead test and tested my scheduler. it is 1.5x slower than the default one for empty tasks :(
<karan> Yes, It's under Classic Project Ideas. Apart from that I had also found "Applying Machine Learning Techniques on HPX Parallel Algorithms" under Vanguard Projects Ideas very interesting.
<jbjnr> karan: I think the machine learning project has had a lot of students enquiring. competition for that might be tough.
<karan> I would like to have your suggestion too. I am very good in languages like C/C++ and Python. https://github.com/karanagg262, My github link to get a bit insight about me
CaptainRubik has joined #ste||ar
kisaacs has joined #ste||ar
kisaacs has quit [Ping timeout: 245 seconds]
prashantjha has joined #ste||ar
david_pfander has joined #ste||ar
<simbergm> hi karan! that project *should* be fairly easy technically, so a large part of the project would be on exploring data/writing up examples... it's rather open ended, and you can change it to your liking
<simbergm> I'm a new mentor so I don't know as much about that project as I should, but for me the most important parts are working csv destination and useful examples to help future users analyze their programs
<simbergm> hkaiser is usually on later and he might be able to tell you exactly what doesn't work currently
katywilliams has joined #ste||ar
<simbergm> but if you'd like to get started there's a comprehensive section on performance counters in the documentation
<simbergm> building hpx and just playing around with the csv output, reading the source for those parts would also be good
<jbjnr> isn't there another project dedicated to perf counters? couldn't the two be combined?
<simbergm> yep, correct
<simbergm> there's a project for a histogram counter, and one to add some more arithmetic or statistical counters
<simbergm> these three can be moulded into something you like
<simbergm> if you have a concrete use case for yourself it can make the goal of the project easier to define
<simbergm> jbjnr: shame about the scheduler :/ I'm going to try what happens with a sliding semaphore or something like that
katywilliams has quit [Ping timeout: 260 seconds]
Viraj has joined #ste||ar
<jbjnr> the histogram perf counters would be quite useful. For example, I would like to see distributions of packet sizes sent during messaging, but oit's not enough work fo 3 months
<jbjnr> simbergm: I already upgraded the test to use slifding semaphore
<simbergm> jbjnr: ok, any difference?
<simbergm> it still waits for all futures at the end?
nikunj_ has joined #ste||ar
prashantjha has quit [Ping timeout: 245 seconds]
<github> [hpx] StellarBot pushed 1 new commit to gh-pages: https://git.io/vxqQu
<github> hpx/gh-pages 6010741 StellarBot: Updating docs
<simbergm> jbjnr: have you tried running a warmup iteration of cholesky before the actual run? on my possibly broken std::allocator, single numa node test I get almost twice the flops on subsequent runs
Viraj has quit [Quit: Page closed]
<jbjnr> simbergm: sorry. was on a call
<jbjnr> semaphoe - yes, much faster - on laptop sub microsecond (if we believe it)
<jbjnr> on daint it still crashes, even though I though I fixed it
CaptainRubik has quit [Quit: Page closed]
<jbjnr> cholesky - I do N runs each tim and don;t see any real variation bigger than the nornal noise level
<simbergm> but with the same runtime instance?
<jbjnr> ?
<jbjnr> you mean restart the app>
<simbergm> without stopping the runtime in between
<simbergm> yeah
<simbergm> do you run your binary N times or do you run the cholesky function N times?
<jbjnr> ok. I run a new copy of the test each time, so the app restarts, but I do not believe this will change anything
<jbjnr> we get 1TF
<jbjnr> it's not going to double!
<simbergm> no, this was with 128 blocksize
<jbjnr> on daint?
<simbergm> yeah
<jbjnr> I don't believe you :)
<simbergm> it might be bogus, but kind of makes sense if stacks etc get reused
<simbergm> first time they have to be allocated fresh
<simbergm> I don't believe it either, so take it with a grain of salt
<jbjnr> aha. I see what you're getting at
<jbjnr> yes. interesting
<jbjnr> (clean up the stack reuse code btw)
<simbergm> sure :)
<simbergm> this is on master btw
<jbjnr> ask heller_ what he broke - he said it wasn't as efficient as before
<simbergm> heller_: what did you break?
<jbjnr> I did not consider that. I will rerun with N iterations in single exe and see what happens
<simbergm> (again?)
<simbergm> ...
<jbjnr> so, with my semaphore based future test, the default scheduler gives 0.8 sec for 1million and my scheduler about 1.2 for 1 million
<jbjnr> If we are losing 0.5 microseconds per task, then my scheduler is worth paying for, BUT
<simbergm> how many threads?
<jbjnr> if we are really losing 60microseconds on daint in a real app, then we have a problem
<simbergm> yeah
<jbjnr> on laptop, just a couple, can't run on daint yet due to crashy crash
<simbergm> ok, sure
<jbjnr> fix it please!
<jbjnr> I did not modify the actio test yet
<jbjnr> ^action
<jbjnr> Note that the test is not 100% reliable - because the Nth future might run before the (N-1)th
<jbjnr> so the semaphore could be signalled before all have completed, but for very large N, this should not be a big deal
<jbjnr> to fix that we should use a proper condition var and an atomic counter
<jbjnr> probably still much much cheaper than wait_all on a million futures
<heller_> jbjnr, simbergm: nothing is broken on master
<heller_> but it could be better
<simbergm> heller_: no, nothing broken there :) question was what related to caches is worse in your lazy thread init branch?
EverYoung has joined #ste||ar
<heller_> jbjnr: did the wait_or_add_new thing improve things? I think it will be much better once we properly reuse stacks and not just thread_data
<jbjnr> heller_: small improvements, nothing ground breaking, but we are chipping away at the iceberg and it is shrinking
<simbergm> that's probably it, so the stack caching is removed/unused in that branch?
<jbjnr> ha!
<heller_> nono
<jbjnr> fix it, fix it, fix it
<simbergm> ok
<heller_> it's just different ;)
<simbergm> right...
<heller_> but yeah, needs to be fixed
<simbergm> how is it different?
<jbjnr> thread local stacks for the stacks
<heller_> yes
<heller_> so, what we do now is to allocate thread_data (which includes the stack)
<heller_> this gets reused
<jbjnr> heller_: I did not rerun vtune yet bcos I had to give a talk yesterday, but I'm back on the case now
<jbjnr> simbergm: I will push a new branch with the lazy init and wait_or_add_new on without all my other crap.
<heller_> what could happen, is that we have *lots* of thread_data objects lying around, with stacks allocated, but we don't really reuse that
<simbergm> jbjnr: okay, whichever way you like
<jbjnr> we can try to get it into the release if you want to work on it
<heller_> after the release...
<jbjnr> I cann do it now. 2 mins hopefully
<simbergm> later
<heller_> talking about the release ... shouldn't we just do it right now?
<simbergm> heller_: pretty much, yes
<simbergm> your libfabric PR needs to go in, hwloc PR seems to be ready, and then a week of waiting until the 10 year anniversary ;)
EverYoung has quit [Ping timeout: 276 seconds]
<heller_> :P
<heller_> I want to prepare my ARM builders...
<simbergm> very good
<simbergm> do they belong to FAU or your own?
<heller_> FAU
<heller_> 8 core ARM64
kxkamil has joined #ste||ar
<jbjnr> heller_: who is giving you all these raspberry pi-like boards?
<heller_> jbjnr: allscale ;)
<jbjnr> cool.
<jbjnr> Can we use them for fun stuff?
<jbjnr> I'd like to run a crypto bot-trader on one :)
<heller_> sure thing :P
<jbjnr> (I'm setting one up at home for that)
<heller_> they'll idle most of the time anyways, they are just bought as a demonstrator for the final review
<jbjnr> next year's gsoc : convert bot trader to use hpx threading for improved response speed!
<heller_> I maintain them, and the evil plan is to have a small test cluster I can use for funny things ;)
<jbjnr> PYCICLE!
<heller_> exactly!
heller_ has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
heller has joined #ste||ar
<heller> jbjnr: my own little botnet ;)
Anushi1998 has joined #ste||ar
victor_ludorum has joined #ste||ar
heller____ has joined #ste||ar
<heller____> hello
<heller> hi
heller____ has left #ste||ar [#ste||ar]
Anushi1998 has quit [Read error: Connection reset by peer]
Dalvan has joined #ste||ar
<Dalvan> Hi
<heller> hi Dalvan
<heller> you made it ;)
DalvanGriebler has joined #ste||ar
<DalvanGriebler> I think now it works in my client
<Dalvan> teste
Dalvan has quit [Quit: Page closed]
<DalvanGriebler> das
<heller> great
<heller> Hi
<heller> hi
K-ballo has joined #ste||ar
<zao> heller: Building HPX with 2GB of RAM? :D
<heller> zao: ever heard of cross compiling ;)?
<zao> Fair enough.
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
katywilliams has joined #ste||ar
katywilliams has quit [Ping timeout: 264 seconds]
WCMATUZimsky has joined #ste||ar
WCMATUZimsky has quit [Client Quit]
Anushi1998 has joined #ste||ar
karan has quit [Quit: Page closed]
hkaiser has joined #ste||ar
eschnett has joined #ste||ar
victor_ludorum has quit [Ping timeout: 260 seconds]
Anushi1998 has quit [Ping timeout: 245 seconds]
ct-clmsn has joined #ste||ar
mcopik_ has joined #ste||ar
aserio has joined #ste||ar
hkaiser has quit [Quit: bye]
victor_ludorum has joined #ste||ar
diehlpk_work has joined #ste||ar
DalvanGriebler has quit [Ping timeout: 264 seconds]
DalvanGriebler has joined #ste||ar
<heller> DalvanGriebler: diehlpk_work is your GSOC admin guy you need to talk to
katywilliams has joined #ste||ar
eschnett has quit [Quit: eschnett]
katywilliams has quit [Ping timeout: 265 seconds]
mbremer has quit [Quit: Page closed]
katywilliams has joined #ste||ar
nikunj_ has quit [Ping timeout: 260 seconds]
jakub_golinowski has joined #ste||ar
eschnett has joined #ste||ar
Anushi1998 has joined #ste||ar
anushi_ has joined #ste||ar
DalvanGriebler has quit [Quit: Ex-Chat]
akheir has joined #ste||ar
Anushi1998 has quit [Ping timeout: 240 seconds]
anushi_ is now known as Anushi1998
katywill1ams has joined #ste||ar
katywilliams has quit [Ping timeout: 260 seconds]
<simbergm> jbjnr: still planning on splitting off the wait_or_add_new into a PR? I could start doing it now...
anushi_ has joined #ste||ar
Anushi1998 has quit [Ping timeout: 240 seconds]
anushi_ has quit [Client Quit]
Anushi1998 has joined #ste||ar
EverYoung has joined #ste||ar
zao- has joined #ste||ar
prashantjha has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
victor_ludorum has quit [Ping timeout: 260 seconds]
EverYoung has joined #ste||ar
Smasher has joined #ste||ar
zao_ has quit [*.net *.split]
quaz0r has quit [*.net *.split]
prashantjha has quit [Ping timeout: 276 seconds]
quaz0r has joined #ste||ar
katywill1ams has quit [Ping timeout: 240 seconds]
katywilliams has joined #ste||ar
<Anushi1998> heller: I am working towards "Implementation of Faster Associative Container for GIDs" for GSoC 2018
<Anushi1998> So, I was looking this file hpx/src/runtime/agas/addressing_service.cpp
<Anushi1998> I am thinking to implement Trie or some other data structure in separate file (to make implementation clean) and then we need to have storage of Vector of tuples.Is it correct?
<Anushi1998> simbergm: Can I ask you a question regarding Issue #1668 Adapt all parallel algorithms to Ranges TS ?
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
prashantjha has joined #ste||ar
EverYoun_ has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
jaafar has joined #ste||ar
victor_ludorum has joined #ste||ar
<heller> anushi:
<heller> Anushi1998: Sounds good
<Anushi1998> So shall I proceed with implementation of Trie?
hkaiser has joined #ste||ar
david_pfander has quit [Ping timeout: 260 seconds]
<Anushi1998> hkaiser: I am trying to adapt adjacentfind.hpp to Ranges TS.I want to how should I decide to choose projection or not?....I am sorry if it sounds a silly question :(
<hkaiser> Anushi1998: look at the standards document: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/n4569.pdf
<Anushi1998> Okay
<hkaiser> there might be a newer version of this document available
katywilliams has quit [Ping timeout: 264 seconds]
<Anushi1998> hkaiser: thanks
katywilliams has joined #ste||ar
<heller> Anushi1998: yes sounds like a plan
<Anushi1998> heller: Okay,thanks:)
katywilliams has quit [Ping timeout: 240 seconds]
daissgr has joined #ste||ar
katywilliams has joined #ste||ar
aserio has quit [Ping timeout: 276 seconds]
kxkamil has quit [Quit: Leaving]
aserio has joined #ste||ar
aserio1 has joined #ste||ar
daissgr has quit [Ping timeout: 245 seconds]
aserio has quit [Ping timeout: 252 seconds]
aserio1 has quit [Ping timeout: 260 seconds]
aserio has joined #ste||ar
aserio has quit [Client Quit]
aserio has joined #ste||ar
daissgr has joined #ste||ar
aserio has quit [Ping timeout: 240 seconds]
aserio has joined #ste||ar
Anushi1998 has quit [Read error: Connection reset by peer]
Anushi1998 has joined #ste||ar
jakub_golinowski has quit [Ping timeout: 246 seconds]
aserio has quit [Read error: Connection reset by peer]
aserio has joined #ste||ar
ct-clmsn has quit [Quit: Leaving]
anushi_ has joined #ste||ar
Anushi1998 has quit [Ping timeout: 252 seconds]
aserio has quit [Ping timeout: 260 seconds]
anushi_ has quit [Remote host closed the connection]
anushi_ has joined #ste||ar
anushi_ is now known as Anushi1998
nikunj has joined #ste||ar
mcopik_ has quit [Ping timeout: 240 seconds]
prashantjha has quit [Ping timeout: 256 seconds]
zao- has quit [Ping timeout: 240 seconds]
victor_ludorum has quit [Quit: Page closed]
eschnett has quit [Quit: eschnett]
katywilliams has quit [Ping timeout: 264 seconds]
<nikunj> @hkaiser: I have a question regarding my GSoC project, good time to ask?
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
hkaiser has quit [Quit: bye]
katywilliams has joined #ste||ar
katywilliams has quit [Ping timeout: 240 seconds]
aserio has joined #ste||ar
katywilliams has joined #ste||ar
EverYoun_ has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
Dalvan has joined #ste||ar
Dalvan has quit [Client Quit]
hkaiser has joined #ste||ar
mcopik_ has joined #ste||ar
nikunj has quit [Quit: Page closed]
akheir has quit [Remote host closed the connection]
<github> [hpx] biddisco created remove_wait_or_add_new (+4 new commits): https://git.io/vxYcG
<github> hpx/remove_wait_or_add_new b6dc6ce John Biddiscombe: Merge branch 'remove-schedulers' into lazy_thread_init
<github> hpx/remove_wait_or_add_new fbc1871 Thomas Heller: Changing the coroutine implementations to do a lazy init...
<github> hpx/remove_wait_or_add_new 7fe9741 John Biddiscombe: Remove staged queue from thread map and run_now param from create_thread api...
<github> [hpx] msimberg pushed 1 new commit to circle_2: https://git.io/vxYcZ
<github> hpx/circle_2 1375ed7 Mikael Simberg: Use make instead of ninja for circleci
<K-ballo> simbergm: why make instead of ninja?
<jbjnr> simbergm: I pushed a wait or add new branch - it depends on the remove-schedulers and also the lazy init branches
<jbjnr> bt it's 'clean' and can be used as a startpoint for stack cleanup
<jbjnr> sorry - it took longer than expected due to a disk crash and also a harder than expected branch cleanup and merge etc
<github> [hpx] biddisco opened pull request #3243: Remove wait or add new (master...remove_wait_or_add_new) https://git.io/vxYC2
wash has quit [*.net *.split]
wash has joined #ste||ar
aserio has quit [Quit: aserio]
diehlpk_work has quit [Quit: Leaving]
Anushi1998 has quit [Remote host closed the connection]
Anushi1998 has joined #ste||ar
katywilliams has quit [Ping timeout: 260 seconds]
jbjnr has quit [Remote host closed the connection]
Anushi1998 has quit [Remote host closed the connection]
Anushi1998 has joined #ste||ar
Anushi1998 has quit [Remote host closed the connection]
Anushi1998 has joined #ste||ar
nikunj has joined #ste||ar
<parsa[w]> hkaiser: do we have anything like define(x, nil)?
nikunj has quit [Ping timeout: 260 seconds]
<hkaiser> parsa[w]: use define(x, '())
<hkaiser> or define(x, block())
<hkaiser> the first returns an empty list, the second nil
<parsa[w]> is equal(x, nil) true for define(x, block())?
EverYoun_ has joined #ste||ar
<parsa[w]> (or for both)
<hkaiser> for the second
<hkaiser> parsa[w]: I was planning to add nil directly to the compiler
<parsa[w]> why?
nikunj has joined #ste||ar
EverYoung has quit [Ping timeout: 245 seconds]
EverYoun_ has quit [Ping timeout: 252 seconds]
prashantjha has joined #ste||ar
<hkaiser> parsa[w]: why not?
<parsa[w]> do we need it?
<hkaiser> we can use block(), but that's counter-intuiyive, isn't it?
<parsa[w]> would we be able to send nil as an argument to primitives if you add it to the compiler?
<hkaiser> yes
<hkaiser> nil is one of the variant options
<parsa[w]> it is needed then :)
<zao> If this is C++, remember that Apple still likes to `#define nil`
EverYoung has joined #ste||ar
<parsa[w]> hkaiser: did more sketchy things in #311. check it out whenever you have time
<parsa[w]> zao: no worries, it's a DSL
<zao> Ah, great.
<heller> parsa[w], hkaiser: wouldn't it be better to have it named 'None', would be more pythonic
<parsa[w]> a) you came up with the name, b) nil is shorter and i personally like it
<parsa[w]> oh didn't read who sent it
<heller> :P
<parsa[w]> heller: it's a node in an expression tree, nil was an intuitive choice
<heller> I know where it comes from
<heller> it's just a name
<parsa[w]> right
<heller> in python, those things are called None
<parsa[w]> i know, still doesn't matter... PhySL wasn't supposed to be Python
<heller> ok
<heller> so PhySL is more lisp like, right?
<parsa[w]> yeah, more on the Scheme side though
<heller> ok, then nil makes much more sense
<hkaiser> zao: do they?
<hkaiser> uggh
<zao> They did back in the days with Obj-C++ and whichever Boost lib it was, Proto?
<heller> fusion
<heller> list
<heller> cons and nil
<hkaiser> parsa[w]: #315
<hkaiser> heller: but fusion does not #define nil, does it?
<heller> no, it has a symbol named nil
<hkaiser> sure, that's fine
<heller> obj-c++ defines nil
<K-ballo> nil is an obj-c keyword
<heller> which renders that code unusable
<hkaiser> ok
<K-ballo> check, verify, and some other are the ones that they define
<parsa[w]> hkaiser: that was really fast... and it works
<hkaiser> ;)
<hkaiser> was a two-liner
<heller> the emperor at work
<parsa[w]> indeed
<parsa[w]> 2 lines plus 94 more :)
<heller> :P