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/
rtohid has left #ste||ar [#ste||ar]
EverYoung has joined #ste||ar
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 265 seconds]
EverYoun_ 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
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
EverYoung has joined #ste||ar
EverYoun_ has quit [Ping timeout: 248 seconds]
daissgr has quit [Quit: WeeChat 1.4]
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Read error: Connection reset by peer]
EverYoung has joined #ste||ar
Smasherr has quit [Quit: Connection reset by beer]
Smasher has joined #ste||ar
Smasher has quit [Client Quit]
EverYoun_ has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
vamatya has joined #ste||ar
Smasher has joined #ste||ar
EverYoun_ has quit [Ping timeout: 240 seconds]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 240 seconds]
jaafar has quit [Ping timeout: 248 seconds]
vamatya has quit [Quit: Leaving]
EverYoung has joined #ste||ar
hkaiser has quit [Quit: bye]
EverYoung has quit [Remote host closed the connection]
eschnett has quit [Quit: eschnett]
EverYoung has joined #ste||ar
hkaiser has joined #ste||ar
eschnett has joined #ste||ar
EverYoung has quit [Ping timeout: 276 seconds]
hkaiser has quit [Quit: bye]
Smasherr has joined #ste||ar
Smasher has quit [Ping timeout: 240 seconds]
Smasherr has quit [Ping timeout: 240 seconds]
Smasher has joined #ste||ar
anushi has quit [Ping timeout: 265 seconds]
Shikhar has joined #ste||ar
Shikhar has quit [Quit: Page closed]
nanashi55 has quit [Ping timeout: 260 seconds]
nanashi55 has joined #ste||ar
CaptainRubik has joined #ste||ar
Smasherr has joined #ste||ar
Smasher has quit [Ping timeout: 240 seconds]
CaptainRubik has quit [Read error: Connection reset by peer]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
AndroUser has joined #ste||ar
anushi has joined #ste||ar
Smasher has joined #ste||ar
AndroUser has quit [Ping timeout: 260 seconds]
Smasher has quit [Remote host closed the connection]
EverYoung has quit [Read error: Connection reset by peer]
simbergm has quit [Ping timeout: 248 seconds]
simbergm has joined #ste||ar
simbergm has quit [Client Quit]
simbergm has joined #ste||ar
Nikunj has joined #ste||ar
<Nikunj>
Hi
<Nikunj>
My name is Nikunj Gupta
<Nikunj>
I have quite a bit of working knowledge on many programming languages and have worked on projects but haven't worked on the open source front yet. I would like to start my open source journey through hpx.
<Nikunj>
I'm currently a Freshman at IIT (Indian Institute of Technology) Roorkee
<anushi>
Also you can take a look on open issues currently :) on git channel
AndroUser2 has joined #ste||ar
AndroUser2 has quit [Read error: Connection reset by peer]
<Nikunj>
I'd like to work on issue #3124 in hpx. If no one else is working on that issue, I'd like to take it up.
<Nikunj>
Can I get a general idea on the template of CMakeLists.txt that you expect
<simbergm>
hi Nikunj, I don't believe anyone is working on that one at the moment, so you can also comment on the issue on github to say that you will take it
<Nikunj>
Also I'm unaware of the initial copyrights and license that you provide in the first few lines of CMakeLists.txt
<Nikunj>
Hey Simberg! Ok I will provide a comment on that issue stating that I'd be taking up the issue
<simbergm>
you should be able to take a CMakeLists.txt from one of the other examples and modify it for that example
<simbergm>
for the license header simply substitute in the correct year and name and you're good
<Nikunj>
Ok
david_pfander has joined #ste||ar
<simbergm>
note that you should update examples/hello_world_component/README.rst as well in that case
<Nikunj>
In that case I'd start working right away. What is the general deadline for an issue that you provide?
Smasherr has quit [Ping timeout: 240 seconds]
<simbergm>
there is no deadline! if you think you won't have time to work on it for a while just comment again in case someone else wants to take it up
<Nikunj>
Yes examples/hello_world_component/README.rst should also be updated, also I feel that Makefile and hello.sh should be deleted so as to reduce redundancy
Smasher has joined #ste||ar
<simbergm>
well, seeing as it's meant to demonstrate how to build a component without cmake I believe they should stay
<simbergm>
the issue is simply to prevent us from breaking the example and not noticing, i.e. to have it in our CI
<Nikunj>
Ok
<Nikunj>
I feel that I've understood the basics of the issue. I will provide the commit as soon as it's done. In case I feel trapped I'll raise a question here.
Smasher has quit [Ping timeout: 240 seconds]
Vir has quit [Read error: Connection reset by peer]
<github>
hpx/master b2f3758 cogle: Implemented count/count_if for containers
<github>
hpx/master c060c33 cogle: Merge branch 'master' into count_update
<github>
hpx/master aeb813d cogle: Updated the documentation
sam29 has joined #ste||ar
Smasher has quit [Ping timeout: 240 seconds]
Smasher has joined #ste||ar
Smasher has quit [Ping timeout: 240 seconds]
Smasher has joined #ste||ar
sam29 has quit [Ping timeout: 260 seconds]
Smasherr has joined #ste||ar
Smasher has quit [Ping timeout: 240 seconds]
Anushi1998 has joined #ste||ar
<Anushi1998>
When I am running make tests && make test it stuck at 45% when Building CXX object tests/unit/parallel/segmented algorithms/CMakeFiles/partitioned_vector_transform_binary_test_exe.dir/partitioned_vector_transform_binary.cpp.o
<Anushi1998>
I have tried this 2-3 time & every time leaving laptop for 2 hrs but there is no progress!
<heller_>
how much RAM do you have?
<Anushi1998>
4GB
Anushi1998 has quit [Client Quit]
Anushi1998 has joined #ste||ar
<Anushi1998>
heller_: What can I do? Can I avoid this particular file to continue?
<heller_>
edit the cmake lists
<heller_>
for reference, it is not stuck, just takes a while
<heller_>
probably swaps heavily
<Anushi1998>
Okay, thanks :)
Anushi1998 has quit [Quit: Leaving]
mcopik has quit [Ping timeout: 256 seconds]
hkaiser has joined #ste||ar
<K-ballo>
question about templated components at cpplang slack
Smasherr is now known as Smasher
AndroUser2 has joined #ste||ar
<simbergm>
K-ballo: btw, the overload on format_to for hpx::iostreams::ostream worked, thanks
<simbergm>
something is obviously missing though
<K-ballo>
yeah, the hpx streams... but there must be no way to implement them sanely
<simbergm>
possible... this is already better than before
<simbergm>
actually... hkaiser: yt?
<hkaiser>
here
<simbergm>
so the wacky hello_world output
<simbergm>
it's because hpx::util::format_to takes a std::ostream and then somehow the hpx ostream << overloads don't get called
<simbergm>
so the locking and so on is bypassed
<zao>
Ooh.
<hkaiser>
interesting
AndroUser2 has quit [Read error: Connection reset by peer]
<simbergm>
but do you know what might be missing from the hpx ostream? possible to fix it?
AndroUser2 has joined #ste||ar
<hkaiser>
the locking is supposed to happen inside the virtual functions that are called from the ostream
<hkaiser>
simbergm: but I'll have a look
<simbergm>
okay, great
<simbergm>
otherwise it works by adding a format_to(hpx::iostreams::ostream, ...) overload
<hkaiser>
simbergm: nice - should we do that? ;)
<hkaiser>
at least for the time being
<simbergm>
I would do it if there doesn't seem to be another way
<simbergm>
I don't know how to fix the ostream properly at the moment
<hkaiser>
need to look closer, as K-ballo said, there doesn't seem to be a way to sanely implement those
victor_ludorum has joined #ste||ar
<simbergm>
okay, but then I think this is already infinitely better than how it is now, so I'll do a PR
<hkaiser>
k
<simbergm>
thanks
<hkaiser>
well, thank you!
AndroUser2 has quit [Remote host closed the connection]
<Nikunj>
Hey! I'm back. So, I was able to create the CMakeLists.txt which ran fine on running cmake however running make afterwards prints out the following error statement:
<Nikunj>
CMakeFiles/hello_world_component.dir/hello_world_component.cpp.o: In function `examples::server::hello_world::invoke()':
<Nikunj>
hello_world_component.cpp:(.text+0x31d): undefined reference to `hpx::cout'
<Nikunj>
collect2: error: ld returned 1 exit status
<hkaiser>
Nikunj: you need to link with th eiostreams component
<hkaiser>
add 'COMPONENT_DEPENDENCIES iostreams' to add_hpx_executable() for your target
<K-ballo>
isn't there a CMakeLists.txt already?
<Nikunj>
@k-ballo, for hello_world_component there isn't
<Nikunj>
That's raised as an issue #3124 in hpx repo
<K-ballo>
I'm confusing it with the external example
<K-ballo>
no, that's not what the issue is about (assuming that one is mine)
<Nikunj>
Yes, I'm working on your issue. I guess I misunderstood the issue in that case.
<simbergm>
K-ballo: one can either call the hello.sh script manually in the CI scripts on rostam, or one can add a CMakeLists.txt, no? hello.sh and co can still stay to show how to build it manually
<K-ballo>
somewhat recently we made some changes that broke that example, and we didn't noticed because it's never built by our CI builds
<K-ballo>
"This directory intentionally avoids providing a CMakeLists.txt file because it is intended to illustrate building by command line." so we should excercise the script in CI builds, regardless of whether there is or isnt a cmakelist (if there is one it ought to be excercised by CI as well)
<hkaiser>
we could add a cmakelists that launches the script ;)
<simbergm>
I noticed the readme but what's the harm in having the CMakeLists.txt for CI? makes it easier, don't need a special case for it
<K-ballo>
I have no opinion on having or not having a cmakelist.txt
<hkaiser>
K-ballo: I agree the CI should exercise the script
<K-ballo>
but CI must at least test the script, since that's what the example is about
<hkaiser>
nod
<simbergm>
that's a good point
<simbergm>
so Nikunj, sorry that I led you slightly on the wrong path
<hkaiser>
will not work for appveyor, though - but that's not a big problem, I think
<K-ballo>
looks like a linux only example
<github>
[hpx] msimberg opened pull request #3174: Fix format to (master...fix-format-to) https://git.io/vAnUg
<Nikunj>
Sorry for replying late (make -j5 takes all the juice that my processor has). So what should I exactly be aiming for?
<zao>
Nikunj: I had to add additional swap to build reliably with high -j :)
<Nikunj>
Additional swap is a nice idea! I'll add one myself then
<zao>
You computer will still of course suck mud, but you won't OOM at least :)
<zao>
I got slightly better build times doing -j16 (with hyperthreading and swap) over -j8 on my Ryzen :)
<zao>
Unusable interactively tho, of course.
<Nikunj>
ya right :)
mcopik has joined #ste||ar
eschnett has quit [Quit: eschnett]
<zao>
435G builds/
<zao>
Soon time to figure out how to archive/cull old builds :)
<K-ballo>
435G?
<zao>
(just 26 trees, some with rwdi and debug)
<zao>
A compressed build tree is around 4 gigabytes.
victor_ludorum has quit [Quit: Page closed]
<zao>
K-ballo: I still need to figure out how to be more like jbjnr and do miracles in implementing working CI infrastructure.
<Nikunj>
K-ballo: I still need learn implementing CI infrastructure, in case we need to hello
<Nikunj>
in case we need to add hello_world_component, I do not find myself fit for the job
victor_ludorum has joined #ste||ar
<hkaiser>
Nikunj: don't give up, it's not difficult
<simbergm>
it can be confusing at first, but it gets better
<simbergm>
you need three things:
<Nikunj>
hkaiser: I will learn a bit about CI integration and how to write cmakes specific to CI
<simbergm>
1. check that you can run hello.sh manually
<hkaiser>
simbergm: btw, do you have a non-log-log plot for the data shown in #3137?
<simbergm>
2. find a way to run commands from cmake
<simbergm>
3. run the command when building :)
<hkaiser>
well, it's just log, but anyways
<simbergm>
hkaiser: I don't but I can add one
<simbergm>
it's just not very easy to read
<hkaiser>
log plots hide the important things ;)
<simbergm>
sure, give me a sec
<hkaiser>
no worries, no rush
eschnett has joined #ste||ar
<Nikunj>
I see
<simbergm>
and Nikunj: there are no stupid questions so always ask if you get stuck
<Nikunj>
sure :)
<simbergm>
hkaiser: added a couple of plots to #3137
<hkaiser>
simbergm: thanks!
hkaiser has quit [Quit: bye]
aserio has joined #ste||ar
simbergm has quit [Ping timeout: 260 seconds]
simbergm has joined #ste||ar
<diehlpk_work>
Who is the student interested in the OpenCV project?
<Nikunj>
Hey! So I've been googling about CI integrations and how to implement it. I was easily able to reproduce step 1 that simbergm gave above. However, I don't understand as to what exactly do I do next.
<diehlpk_work>
I think that we can replace the main in this example with the hpx_main and it should still work
<zao>
diehlpk_work: Sure... I may be misunderstanding the problem statement but I thought it was to make HPX a backend for any application using OpenCV.
<CaptainRubik>
Hi. Reg Concurrent Data Structures- Lockfree implementations require atomic hardware primitves. How does hpx currently manage the differences in various architectures given that it is a library meant to work on clusters? Which files should I look into for the same.
<jbjnr>
CaptainRubik: we use std::atomic and rely on the compiler writers to make sure it works on different archtectures
sam29_ has joined #ste||ar
<jbjnr>
for clusters, each node should have a binary compiled locally, no need to do anything 'special'
parsa[[w]] has joined #ste||ar
<jbjnr>
meaning - that if you had some x86 and some arm and some powerpc nodes in a cluster, you'd have to cpmpile HPX on each independently
<CaptainRubik>
ok. So what should I start looking into for Concurrent DS project?
<jbjnr>
then persuade them to relicense it under the bost license :)
<jbjnr>
then persuade them to relicense it under the bost license :)boost
<jbjnr>
oops
<jbjnr>
boost - I was trying to say
<jbjnr>
we really want stuff like that that we can use.
parsa[w] has quit [Ping timeout: 255 seconds]
sam29_ has quit [Ping timeout: 260 seconds]
<jbjnr>
they have "several implementation of unordered set/map - lock-free and fine-grained lock-based
<jbjnr>
" those would be very useful in hpx
<jbjnr>
(for examle)
<CaptainRubik>
Well, I don't have much knowledge about license stuff. But I am sure this will be a great resource to understand the algorithm and then implement it ourselves.
<jbjnr>
yes
<CaptainRubik>
as a part of the project.
<jbjnr>
I'll send them an email and ask about boost license ...
<aserio>
and what is the difference between PR and experimental?
<jbjnr>
if a build gets killed by see the config/build entriesefore the tests run, then we onl
<jbjnr>
grrr. my keyboard
<jbjnr>
if a build gets killed before the tests run, then we only see the config/build entries
<aserio>
for instance when a new change set is pushed?
<jbjnr>
yes, or when I restart pycicle to test why the entries are not appearing!
<aserio>
lol
<jbjnr>
sorry. a bit of a mess today after the cdash upgrade
<aserio>
and the PR vs Experimental?
<CaptainRubik>
The license for libcds says "Redistribution and use in source and binary forms, with or without modification, are permitted ." Does that work for us then?
<jbjnr>
CaptainRubik: we need the ability to "relicense" because we have to use boost <sigh>
<jbjnr>
I'll contact them and ask.
<jbjnr>
CaptainRubik: the main thing to understand is that multithreaded code is hard. locks stop things from clobbering each other, but cause delays. Lockfree containers are great, but can be worse because multiple threads can access internals and cause cache invalidation.
<jbjnr>
the project really is to gather as much stuff as possible and assimilate it into hpx
<jbjnr>
because writing it from scratch is super hard
<CaptainRubik>
Okay so that means
<jbjnr>
so I'll send an email to the libcds people
<CaptainRubik>
I must exhaust all resources available
<CaptainRubik>
and compile a list
<jbjnr>
yes
<jbjnr>
good place to start
<CaptainRubik>
Alright thanks.
<jbjnr>
you'll need to modify what's out there because the locks and mutexes etc in hpx are specialized for hpx, so your tasks ought to be #1 find what;s out there, #2 see if it can be used, #3 modify it to work with hpx, #4 integrate it into the hpx code base. But there is still the option of writing it from scratch if necessary - or if you want the learnign experience
vamatya has joined #ste||ar
eschnett has joined #ste||ar
anushi has joined #ste||ar
kisaacs has joined #ste||ar
aserio has quit [Ping timeout: 240 seconds]
<kisaacs>
Since I'm not familiar with the code base, where should I go to check out the list of supported numpy primitives in Phylanx? Thanks!
<bibek>
kissacs: Hey kate, this directory has the hpp files for all the primitives that we have added so far