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/
<hkaiser>
thanks zao
Smasher has quit [Ping timeout: 276 seconds]
Smasher has joined #ste||ar
<github>
[hpx] hkaiser created fixing_2800 (+1 new commit): https://git.io/v7PjT
<hkaiser>
the user explicitly creates the rp in user land and passes it to init()
<jbjnr>
pleas go ahead
<jbjnr>
(sa long as it still initializes a singleton that we can access from everywhere else)
<hkaiser>
that forces a certain protocol onto the API which can't be abused
<jbjnr>
ok
<hkaiser>
jbjnr: yes, the singleton will refer to the instance in user land
<jbjnr>
if the user doesn't need the RP and doesn't use it - then?
<hkaiser>
hpx will create one as before
<jbjnr>
default one still created behind the scenes
<jbjnr>
great
<hkaiser>
right
<hkaiser>
I also want to clean the public API of the RP to the absolute minimum needed for the user
<hkaiser>
move all the stuff into the private parts and let hpx have access to them, if needed
<hkaiser>
jbjnr: note though that this will require for you to change your startup code for the apps where you use the RP
<hkaiser>
(slightly)
<jbjnr>
no problem. I only have one so far
<hkaiser>
k
<jbjnr>
and it will only take 2 mins
<hkaiser>
ok, I'll do this, then
<hkaiser>
thanks
<hkaiser>
taeguk[m]: yt?
<github>
[hpx] taeguk opened pull request #2808: Fix compile errors on some environments of parallel::partition. (master...tg_partition_fix) https://git.io/v7X9l
eschnett has joined #ste||ar
pree has joined #ste||ar
taeguk has joined #ste||ar
<taeguk>
hkaiser: I submitted PR which fixes compile errors in parallel::partition.
<taeguk>
But, I can't test correctness on the environments that you suggested.
<taeguk>
In rostam, with clang/3.8.1 and boost/1.60.0, I encountered cmake errors.
<zao>
Do you have a gist with the cmake output, and the command line you used?
akheir has quit [Remote host closed the connection]
aserio has joined #ste||ar
mbremer has joined #ste||ar
<mbremer>
@hkaiser: Regarding the issue, it seems to compile now. I'm trying to make the unit tests however, I'm running into a separate error. It seems to be an internal compiler error with gcc-6.1, which is not getting tested in the buildbot.
<heller>
mbremer: do you build all of HPX including the examples?
<mbremer>
I believe so.
<mbremer>
HPX_WITH_TESTS and HPX_WITH_EXAMPLES are both on
<mbremer>
this showed up when I ran make tests
vamatya has joined #ste||ar
<heller>
mbremer: can you give the exact error and with which test/example it occured?
<mbremer>
I also have gcc/4.9 5.2 which I can load as modules
<heller>
ok
pree has quit [Ping timeout: 260 seconds]
bikineev has quit [Remote host closed the connection]
<heller>
mbremer: the PR should only affect the core
<mbremer>
@heller: So you think it's a different issue? I'm trying to build with gcc/5.2
hkaiser has quit [Quit: bye]
<heller>
mbremer: the PR you linked only concerns builds without hwloc, right?
<mbremer>
yup, and those issues seem to have gotten fixed.
<heller>
great
<mbremer>
kk. Will you approve the merge request then? Otherwise, I can keep hacking and try to get the tests to build, although that just might be an entirely different issue
<heller>
yeah, the ices you see are entirely different
<heller>
you can approve the PR yourself
<mbremer>
kk, will do. Thanks!
hkaiser has joined #ste||ar
bikineev has joined #ste||ar
<aserio>
hkaiser: yt?
<hkaiser>
aserio: here
<aserio>
hkaiser: I thought of two other alternatives
<aserio>
introspect
<aserio>
conditional
<hkaiser>
hmmm
<aserio>
sentient is on the list but it is too loaded
<aserio>
introspect assumes too much
<aserio>
is conditional too loaded?
<hkaiser>
conditional suggests it's a either/or decision
<hkaiser>
which it is not
<aserio>
yea..
<aserio>
elective
EverYoung has joined #ste||ar
<hkaiser>
heh
<heller>
what are you up to?
<heller>
the async_if thingy?
<aserio>
trying to name it
<hkaiser>
heller: yah
<heller>
in the end, it is a callable which returns the launch policy, right?
<heller>
launch::select(...)?
<hkaiser>
yes
<heller>
launch::dynamic?
<hkaiser>
I currently have launch::lazy
<heller>
hmm
<heller>
that is too close to deferred, i think
EverYoun_ has quit [Ping timeout: 240 seconds]
<zao>
mbremer: What build configuration was the ICE with?
<hkaiser>
heller: I'm nnot happy with launch::lazy myself
<hkaiser>
launch::select sounds like a possible option
<zao>
I did a Debug build now with a freshly build GCC 6.1.0, no crash. Any other flags?
<zao>
(C++14)
<aserio>
launch::defined
<aserio>
launch::handcrafted :p
<aserio>
better: artisanal
<mbremer>
zao: I ahve CXX14=On, HWLOC=Off, PARCELPORT_MPI=ture, and MALLOC="jemalloc"
<zao>
Hrm, linker error about hwloc, but otherwise nothing yet, let's see with the PP too.
<zao>
CMAKE_BUILD_TYPE?
<mbremer>
Release
<zao>
Ah, I don't even get to that test when building, libhpx.so fails to link on master.
<zao>
Is this on a branch?
<hkaiser>
fixing_2807
<zao>
Ah!
<zao>
builds fine, sadly.
<hkaiser>
lol
<zao>
Upside of things, I've figured out how to build individual compilers with EasyBuild without too much pain.