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/
thundergroudon[m has quit [Read error: Connection reset by peer]
autrilla has quit [Read error: Connection reset by peer]
<github>
[hpx] hkaiser opened pull request #3094: Relax atomic operations on performance counter values (master...relaxed_atomic_operations) https://git.io/vNIYs
<hkaiser>
K-ballo: question is why wasn't that caught by the tests?
<K-ballo>
could be a number of reasons.. gcc never diagnosed that for instance
<K-ballo>
and at some point it was accepted as a defect, or it will be.. so a more recent clang wouldn't either
<hkaiser>
interesting
autrilla has joined #ste||ar
<zao>
This is Clang 3.8.1, so not exactly bleeding-edge indeed.
<hkaiser>
zao: could you add the missing {} to the object construction and try again, pls?
<zao>
Seems to have gotten it past that problem point.
<hkaiser>
k, thanks
<hkaiser>
I'll patch it up
<zao>
Seems like the rest of the build is good.
<hkaiser>
k
parsa has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
parsa has quit [Client Quit]
parsa has joined #ste||ar
hkaiser has quit [Quit: bye]
nanashi55 has quit [Ping timeout: 248 seconds]
nanashi55 has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
mcopik has quit [Ping timeout: 240 seconds]
V|r has quit [Ping timeout: 240 seconds]
Vir has joined #ste||ar
Vir has quit [Ping timeout: 256 seconds]
Vir has joined #ste||ar
Vir has quit [Changing host]
Vir has joined #ste||ar
david_pfander has joined #ste||ar
Vir has quit [Read error: Connection reset by peer]
Vir has joined #ste||ar
Vir has quit [Read error: Connection reset by peer]
kxkamil has joined #ste||ar
<github>
[hpx] msimberg opened pull request #3095: Add ciso646 header to get _LIBCPP_VERSION for testing inplace merge (master...fix-define-libcpp) https://git.io/vNIrl
<zao>
K-ballo: I'm on my third AMD CPU now, machine is still not stable with FreeBSD.
<zao>
Going to try feeding it HPX builds under Linux and see if it doesn't explode.
<K-ballo>
"see if it *doesn't* explode, confidence
<Zwei>
Oh, are AMD CPUs not good? I've heard something about that... not sure where the link is now.
<Zwei>
I was planning to build myself an 8 core ryzen for local parallel testing purposes...
kxkamil has quit [Quit: Leaving]
<hkaiser>
Zwei: zao has extensive experience in breaking ryzen machines
hkaiser has quit [Quit: bye]
<zao>
Zwei: I've had the classical segfaults when building GCC, a bunch of different machine hangs, and most annoyingly, hangs in FreeBSD when sending/receiving ZFS snapshot streams.
<zao>
Also managed to hang 13/16 cores under Linux when running way too many HPX tests simultaneously.
<zao>
I can mitigate the FreeBSD problem by disabling SMT, and have been talking to AMD support/RMA over email since August, for a machine I started building in April.
<zao>
Nice idea for a chip, but woefully unstable in execution.
hkaiser has joined #ste||ar
<Zwei>
Hmm... I guess I'll stick with intel then.
<Zwei>
Bummer, I wanted to support the underdog.
<zao>
They're probably fine for most people, but I've run into approximately all the problems.
<zao>
Including memory compat for the first few months.
<Zwei>
heh... wow.
<Zwei>
I'll mainly be using it for stress testing software.
<Zwei>
It's not much, but I dunno when they'll revoke my access...
<Zwei>
(they actually should have taken it away a year ago, idk why they haven't.... I keep logging in once a week and touching all the files, I think that might have something to do with it)
aserio has joined #ste||ar
diehlpk_work has quit [Remote host closed the connection]
<zao>
On my site we've finally gotten around disabling the default batch system account and expiring user accounts a few months after the project expires now.
<github>
[hpx] hkaiser closed pull request #3089: Fix a strange thing in parallel::detail::handle_exception. (Fix #2834.) (master...tg_fix_2834) https://git.io/vbxr3
<simbergm>
hkaiser: I was planning to write a "heads-up" email for the upcoming release, nothing complicated, do you have any requirements/wishes for it? I can send it to hpx-users, hpx-dev and the stellar lists?
<github>
hpx/master ed109f7 Hartmut Kaiser: Making older versions of clang happy (by supplying an explicit '{}' for the default construction of an object)
<hkaiser>
simbergm: no, and yes
<hkaiser>
there should be similar emails in the archive somewhere
<simbergm>
hkaiser: ok, thanks
<simbergm>
and yes, I saw the old ones from the archives, will take some inspiration from there
gentryx_ has quit [Remote host closed the connection]
<Zwei>
and it's in the topic.... I should have read that first... /o\
<Zwei>
I suck so much
<Zwei>
http://rostam.cct.lsu.edu/console <- I have no idea what I'm looking at. But I've been up for over 24 hours. I'll sleep now and learn more later :)
<zao>
Zwei: A lot of the development progress is driven by whether assorted CI passes.
<zao>
(at least in a wishful manner)
<zao>
So these dashboards/consoles show how well the software builds across a spectrum of toolchains, libraries and OSes.
<zao>
As we pretend to support several releases back in compilers and libraries like Boost, it helps to have boxen that actually build with them and find the silly oversights you make when you develop yourself on a single configuration.
<zao>
G'night :)
aserio has joined #ste||ar
<K-ballo>
...
<heller_>
what's wrong?
<zao>
I should work on my phrasing some time :)
aserio has quit [Ping timeout: 240 seconds]
Smasher has joined #ste||ar
mcopik has joined #ste||ar
daissgr has joined #ste||ar
ITassin has quit [Ping timeout: 256 seconds]
ITassin has joined #ste||ar
diehlpk_work has joined #ste||ar
<K-ballo>
that reminds me.. hkaiser: memory_order_consume? are you sure?
akheir has joined #ste||ar
aserio has joined #ste||ar
Vir has joined #ste||ar
<hkaiser>
K-ballo: why not?
<K-ballo>
hkaiser: because not even SG1 knows what it means
<K-ballo>
and so it has deprecated it
<hkaiser>
hmm, then I screwed up, I meant to use the proper one ;)
<hkaiser>
what was it - memory_order_acquire?
<K-ballo>
acquire is the non-deprecated one
<hkaiser>
nod, I meant to use that one
<hkaiser>
thanks for the heads-up
rtohid has quit [Quit: rtohid]
rtohid has joined #ste||ar
ITassin has quit [Quit: Leaving]
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
bibek has quit [Quit: Konversation terminated!]
bibek has joined #ste||ar
<K-ballo>
most performance tests don't link against HPX, what's the reason for that?
<heller_>
they might not use any hpx features?
<heller_>
like the "baseline" tests
<heller_>
or the ones using OpenMP or TBB
<heller_>
aserio: by your judgement, is it really worth it to eliminate all sentences that use the passive form or nominalization?
<heller_>
aserio: also, based on a given metric, how "readable" must a dissertation be? college undergrad or highschool junior year?
<heller_>
also, I am not sure how much I should trust those metrics ... for me, my text improved...
<heller_>
but who am I to judge?
<K-ballo>
they don't use any hpx features that would require linking
<K-ballo>
..but what if they did?
<heller_>
linking it against HPX wouldn't hurt, I guess
<heller_>
or split up libhpx ...
<heller_>
I guess it is specifically about your new fmtlib?
<heller_>
but that's probably out of scope for the change ... just linking against libhpx doesn't hurt
<K-ballo>
yes, specifically about the changes that make the existing features the benchmarks are already using require linking
<K-ballo>
I could always make the format bits header only as before, and they would remain unaffected
<hkaiser>
K-ballo: I think that's done just because we were too lazy to have them not link agains libhpx
<K-ballo>
but I wonder if that defeats the purpose, maybe they weren't even supposed to be using hpx features before
<heller_>
creating a libhpx_util.so probably requires a lot of refactoring and untangling to avoid cycling dependencies
<hkaiser>
that shouldn't however affect anything
<heller_>
depends on what you consider "hpx features"
<K-ballo>
hkaiser: they seem to be *explicitly* made not link against hpx
<heller_>
they were
<hkaiser>
the tests yes, the build system no
<heller_>
the build systems was amended to suite the needs for the tests
<heller_>
I think at some point in time, someone thought by merely linking against libhpx, you'd get a performance impact or interfere in some other strange ways with openmp or tbb
<heller_>
which I am pretty sure is not the case
<K-ballo>
I figured that could be one possible rationale, globals and such
<K-ballo>
anyways.. it's extremely easy for me to keep the new bits header-only, so I'd rather not mind the rationale and just do as told
<hkaiser>
will libhpx even be loaded if while the app was linked against it does not consume any symbols from it?
hkaiser has quit [Quit: bye]
aserio has quit [Ping timeout: 276 seconds]
<heller_>
global ctors, maybe
<heller_>
but I am not sure
<heller_>
but if it'll get loaded?
<heller_>
I guess it will ... how should the dynamic linker decide if the program needs it or not? the static linker said it did, so it is
hkaiser has joined #ste||ar
akheir has quit [Remote host closed the connection]