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/ | GSoC2018: https://wp.me/p4pxJf-k1
kisaacs has joined #ste||ar
kisaacs has quit [Ping timeout: 245 seconds]
nikunj has quit [Quit: Leaving]
kisaacs has joined #ste||ar
kisaacs has quit [Ping timeout: 264 seconds]
twwright_ has joined #ste||ar
twwright has quit [Ping timeout: 264 seconds]
twwright_ is now known as twwright
kisaacs has joined #ste||ar
hkaiser has quit [Quit: bye]
eschnett has joined #ste||ar
kisaacs has quit [Ping timeout: 252 seconds]
kisaacs has joined #ste||ar
mcopik has quit [Ping timeout: 245 seconds]
kisaacs has quit [Ping timeout: 252 seconds]
K-ballo has quit [Quit: K-ballo]
parsa has joined #ste||ar
kisaacs has joined #ste||ar
nanashi55 has quit [Ping timeout: 260 seconds]
nanashi55 has joined #ste||ar
kisaacs has quit [Ping timeout: 248 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
kisaacs has joined #ste||ar
kisaacs has quit [Ping timeout: 264 seconds]
jakub_golinowski has joined #ste||ar
jaafar_ has quit [Ping timeout: 248 seconds]
<jbjnr__>
who broke hpx?
<jbjnr__>
make[3]: *** No rule to make target 'lib/libhpx.so.1.2.0', needed by 'bin/contains_test_test'. Stop.
<jbjnr__>
etc etc for every binary
<github>
[hpx] msimberg opened pull request #3336: WIP: Only run docker push on master on CircleCI (master...circleci-docker-push-master) https://git.io/vhL1Y
mcopik has joined #ste||ar
twwright_ has joined #ste||ar
twwright has quit [Ping timeout: 264 seconds]
twwright_ is now known as twwright
<M-ms>
jbjnr__: huh, still broken? master is broken but not that broken, rostam and pycicle do at least successfully compile everything
<jbjnr__>
no. it was a misconfigured thingy at my end. building ok now
<github>
[hpx] hkaiser deleted init_logging_sequencing at c55f1ab: https://git.io/vhLAw
<M-ms>
jbjnr: with ^ merged things should be smoother again for pycicle, so wait and see?
<M-ms>
thanks hkaiser for the fix
<M-ms>
has anyone been able to reproduce the tcmalloc error that was showing up with gcc 7 and 8 before? (will most likely come up again on rostam now)
<jbjnr>
I think I screwed up and pycicle wasn't picking up the right settings. I'm about to restart it and do a force build to see if things time out again
<jbjnr>
skype is broken for me, but hangouts does work again if you can bear it
mcopik has joined #ste||ar
jakub_golinowski has quit [Ping timeout: 265 seconds]
anushi has quit [Ping timeout: 245 seconds]
jakub_golinowski has joined #ste||ar
<github>
[hpx] msimberg closed pull request #3330: Rename resource_partitioner test to avoid conflicts with pseudodependencies (master...rename-rp-test) https://git.io/vhk8E
zura has quit [Ping timeout: 268 seconds]
eschnett has quit [Quit: eschnett]
<jakub_golinowski>
M-ms, yt?
mcopik has quit [Ping timeout: 245 seconds]
mcopik has joined #ste||ar
galabc has joined #ste||ar
<M-ms>
jakub_golinowski: yep, here
<jakub_golinowski>
I was diving in the opecv parallelism today. I think now I have a bit better idea of what is going on. Interestingly the pthreads default parallelism was added and when there is no framework defined they have their own small implementation of parallelism
<jakub_golinowski>
And this is what the mandelbrot opencv example was using when I run it on my laptop before
<jakub_golinowski>
Now according to your previous suggestions I want to try out with TBB because I need to rebuild opencv anyway
<jakub_golinowski>
I wanted to use tbb from ubuntu package for convenience but opencv won't build in debug mode with tbb from package since it is in release mode
<jakub_golinowski>
I found some solution that suggests to change the value of CMakecache.txt
<jakub_golinowski>
but as I am developing within the opencv I am still interested in building it in debug mode
<M-ms>
jakub_golinowski: -DBUILD_TBB might be your friend, I would think that option means you don't have to worry about building tbb in the correct mode
<diehlpk>
We had a second interview at heise online
<github>
[hpx] hkaiser created global_object_destruction (+1 new commit): https://git.io/vhtR4
<github>
hpx/global_object_destruction fe14fde Hartmut Kaiser: Safeguard HPX against destruction in global contexts...
<hkaiser>
diehlpk: nice!
<diehlpk>
hkaiser, aserio I will add it to our website later
<hkaiser>
nicde
<hkaiser>
nice
<jakub_golinowski>
M-ms, thanks for the hint
<jakub_golinowski>
M-ms, I have a question about cmake
<jakub_golinowski>
Up to now I was using add_hpx_executable with the opencv apps and everythin was nice and running. Now I want to have an opencv app that is not using hpx explicitly
mcopik has quit [Ping timeout: 245 seconds]
<jakub_golinowski>
when I use standard add_executable() the opencv is complaining about lack of C++11
nikunj has joined #ste||ar
<jakub_golinowski>
So I guess the add_hpx_executable is somehow preventing that but I do not know exactly where
<jakub_golinowski>
well Ok
<jakub_golinowski>
set (CMAKE_CXX_STANDARD 11) did the trick
kisaacs has quit [Ping timeout: 252 seconds]
kisaacs has joined #ste||ar
mcopik has joined #ste||ar
<anushi>
Hi
zura has quit [Quit: Leaving]
<M-ms>
jakub_golinowski: yep, add_hpx_executable takes care of that for you
<M-ms>
You'll have to be careful later when adding hpx to opencv that you use the same standard, you might want to recompile with c++11 and stick to that if you didn't already change it
<jakub_golinowski>
M-ms, you mean recompile HPX?
<jakub_golinowski>
btw now I am in the difficult stage -> I am trying to add HPX as the dependecy to opencv but it is not all plain sailin
<M-ms>
Yeah, exactly
<jakub_golinowski>
currently I am going through all the cmakes in opencv
jaafar_ has joined #ste||ar
<jakub_golinowski>
in the CMakeLists.txt in the root of opencv the have all this build options
<jakub_golinowski>
For example:
<jakub_golinowski>
OCV_OPTION(WITH_TBB "Include Intel TBB support" OFF IF (NOT IOS AND NOT WINRT) )
<M-ms>
Mmh, I know...
<jakub_golinowski>
adn then suddenly
<jakub_golinowski>
they use IF HAVE_TBB THEN "TBB (ver ${TBB_VERSION_MAJOR}.${TBB_VERSION_MINOR} interface ${TBB_INTERFACE_VERSION})"
<jakub_golinowski>
my point is that HAVE_TBB seems to be not defined anywhere'
<M-ms>
You can either try hpx_setup_target on opencv targets or add $HPX_LIBRARIES to some opencv variable
<jakub_golinowski>
when I searched the whole repo it has only one occurance
<jakub_golinowski>
no it is not true
<jakub_golinowski>
it is used in src files as #define but I cannot find where it is declared
<M-ms>
Just a minute
<jakub_golinowski>
So my plan was to add find_package(HPX REQUIRED) (of course when -DWITH_HPX=ON is passed)
<jakub_golinowski>
hmm
<jakub_golinowski>
I think that a header file is generated from this config
<jakub_golinowski>
and this is the place where all the HAVE_... defines are defined
<M-ms>
so the HAVE_TBB may be generated by some cmake code that transforms WITH_TBB
<M-ms>
ok, that seems to be it then
<jakub_golinowski>
So I wanted to find this code
<jakub_golinowski>
but I can also rebuild it and see if HAVE_HPX is defined
<M-ms>
and btw, you will never like cmake but you will learn to deal with it...
<M-ms>
will have dinner now, but leave your questions here
<jakub_golinowski>
ok, thanks!
aserio has quit [Ping timeout: 276 seconds]
nikunj97 has joined #ste||ar
nikunj has quit [Ping timeout: 260 seconds]
nikunj97 has quit [Ping timeout: 245 seconds]
nikunj has joined #ste||ar
<M-ms>
so jakub_golinowski I will be semi-unreachable until sunday (might have internet shortly tomorrow afternoon)
<jakub_golinowski>
M-ms, ok I think I am done with the HAVE_HPX business in a sense that I know how to translate the -DWITH_HPX=ON to the #define HAVE_HPX in
<jakub_golinowski>
M-ms, so maybe just one important Q
<jakub_golinowski>
now I am at the stage that I should make sure that the opencv app can do stuff like
<jakub_golinowski>
#include <hpx/...>
<M-ms>
cmake can be a pain sometimes, so don't worry too much if you can't find the best way to do something with it right away
<jakub_golinowski>
so far I was using add_hpx_executable
<jakub_golinowski>
is there a standard way to deal with libraries?
<M-ms>
sure
<jakub_golinowski>
I found in the docs
<jakub_golinowski>
hpx_setup_target(target)
<M-ms>
yeah, hpx_setup_target is I guess the official way of doing it
<jakub_golinowski>
for the TBB for example the key commands I guess are:
<M-ms>
but opencv might have some variables where they keep their dependencies/includes etc and then it might be easier to use HPX_LIBRARIES and HPX_INCLUDE_DIRS (or DIRECTORIES, don't remember)
<M-ms>
yes, and something more to add tbb as a dependency to opencv itself I think
<jakub_golinowski>
like this: list(APPEND extra_libs tbb)
<jakub_golinowski>
?
<M-ms>
for hpx you should not even need an add_library, I think that is for the case when tbb is built by opencv
<M-ms>
mmh, yes, looks reasonable
<M-ms>
I guess that's from the opencv cmake=
<M-ms>
?
<jakub_golinowski>
I am going over the various CMakeLists.txt
<jakub_golinowski>
in the opencv repository to make sense out of it
<M-ms>
and probably extra_libs is at some point used for target_link_libraries(opencv extra_libs) or something like that
<jakub_golinowski>
but disturbingly it seems to be the same
<jakub_golinowski>
(also it is longer but I just pasted firs errors)
<hkaiser>
diehlpk: see pm, pls
<K-ballo>
that error message still shows something being built in C++11 mode
<K-ballo>
see how the function is declared as "HPX_CXX14_CONSTEXPR void reset()" but the error message refers to a const qualified function
<K-ballo>
that was a C++11 rule, dropped in C++14
<jakub_golinowski>
but what is bugging me
<jakub_golinowski>
since the hpx is already built then why do I see the build errors in the first place?
<K-ballo>
you are seeing a header being parsed
<jakub_golinowski>
because these are from headers
<jakub_golinowski>
?
<jakub_golinowski>
ah yea
<K-ballo>
what's that warning about precompiled headers being invalid?
<K-ballo>
are you using precompiled headers for opencv?
<jakub_golinowski>
indeed this option is set
<jakub_golinowski>
-- Precompiled headers: YES
<K-ballo>
did you do a clean opencv build?
<jakub_golinowski>
what do you mean by clean?
<K-ballo>
from scratch, not carrying around leftovers from previous builds
<K-ballo>
that warning makes me question whether you are actually still compiling for C++11, or if you changed that and yet are still reparsing some stale precompiled headers or something
<jakub_golinowski>
I deleted the directory
<jakub_golinowski>
and created it again
<K-ballo>
ok, and setting flags for C++14 presumably ?
<jakub_golinowski>
by the export
<K-ballo>
does opencv respect the export?
<jakub_golinowski>
and then -std=c++14 is listed in cmake messages
<diehlpk>
K-ballo, i think that cmake on linux respect export to env
<jakub_golinowski>
I am still thinking that the error might be caused by the wrong cmake configuration
<K-ballo>
cmake respects CXXFLAGS everywhere, as long as no explicit toolchain is specified
<K-ballo>
but the project might override it, projects are quite bad at handling flags
<K-ballo>
people tend to bundle their own requirements as users with the requirements of the project, with those of the toolchain
<nikunj97>
hkaiser: should I write a unit test for my pr?
<nikunj97>
hkaiser: I've tested it with hpx debug options on my pc
<nikunj97>
and everything is running fine as of now
<hkaiser>
nikunj97: a unit test is certainly appreciated
aserio has quit [Quit: aserio]
aserio has joined #ste||ar
aserio has quit [Client Quit]
<nikunj97>
hkaiser: ok, so I'll work on a unit test as well
<nikunj97>
hkaiser: also what should I name my branch?
hkaiser has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
<hkaiser>
nikunj97: what_ever_you_like ;-)
nikunj97 has quit [Ping timeout: 256 seconds]
<K-ballo>
release
<jakub_golinowski>
diehlpk, K-ballo after building HPX in c++11 the error did not occur but now I get linking error which sounds more like wrong cmake config
<K-ballo>
involving hpx symbols?
<jakub_golinowski>
K-ballo, you mean in the error?