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/
parsa[w] has quit [Read error: Connection reset by peer]
parsa[w] has joined #ste||ar
jbjnr has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
Matombo has joined #ste||ar
Matombo has quit [Ping timeout: 240 seconds]
Matombo has joined #ste||ar
<mcopik>
what is the expected outcome of share/ in build directory? is it supposed to contain a share-${HPX_VERSION} directory with docs and a symlink hpx to this version? does it make sense?
<github>
hpx/master e80522d atrantan: Integrate spmd multidimensionnal views for partitioned_vectors
<github>
hpx/master b6f6669 atrantan: Add missing includes
<github>
hpx/master 4b3adf0 atrantan: Make some private methods public
<heller>
wow. we have a big performance boost for octotiger
<heller>
on cori
<zao>
Neato. Is it still correct? :)
<heller>
*shrug*
<heller>
:P
<heller>
I assume so, at least
<hkaiser_>
heller: did Dominic get back to you yesterday?
<heller>
yes
<hkaiser_>
ok, thanks for looking into it
<heller>
as I projected ... now I have to do all the debugging ...
<hkaiser_>
do you have the same problems?
<hkaiser_>
(I assume not)
<heller>
I don't know ...
<heller>
I see a hang, but no OOM errors
<heller>
the hang is fixed when reverting to a known good revision
<heller>
bisecting ...
<hkaiser_>
I might have misunderstood what he said :/
<hkaiser_>
he said he even tried code from April and it failed in the same way...
<heller>
I still don't know in which way it failed for him
<hkaiser_>
grrr
<hkaiser_>
I really asked him to descirbe what's goin gon
denis_blank has joined #ste||ar
hkaiser_ has quit [Read error: Connection reset by peer]
bikineev has joined #ste||ar
EverYoung has joined #ste||ar
EverYoung has quit [Ping timeout: 255 seconds]
hkaiser has joined #ste||ar
<heller>
hkaiser: in the meantime, I got carried away with meta classes
<hkaiser>
yah, saw your email
<heller>
this solves all our boilerplate code generation issues
<hkaiser>
heller: time for you to wrap up the phd - too many shiny things left and right
<heller>
;)
<heller>
yes
<heller>
getting there...
<hkaiser>
yes, I know - with proper reflection support we can solve all the macro cruft
bikineev has quit [Ping timeout: 255 seconds]
<heller>
we can even get rid of explicit clients...
<heller>
auto generated actions ftw
<hkaiser>
right
<heller>
that was the motivation to work on, to give a small chapter on the "future" ;)
<hkaiser>
ok, ok
<heller>
anyway ...
bikineev has joined #ste||ar
<heller>
one of the remarks from sterling that got stuck in my head was: "yeah, you really need a new language to take proper advantage of all the ParalleX features"
<heller>
and look where we are heading to now...
<hkaiser>
not true
<hkaiser>
no new language ;)
<heller>
turns out C++ has always been that language :P
<hkaiser>
absolutely
<heller>
well, it is what sterling insisted on ... thus his "let's stick to a C interface, that way we can write compilers more easily!"
<hkaiser>
shrug
<hkaiser>
he admitted that we've been right all along
<heller>
sure, I am just reflecting
<zao>
heller: You wouldn't believe how much the metaclasses introduction smell like Template Haskell :)
<hkaiser>
lol
<hkaiser>
yah, indeed
<heller>
sure, but is this a bad or a good thing now?
<zao>
Cautiously optimistic.
K-ballo has joined #ste||ar
<zao>
Used responsibly, ducking out to manipulate the AST with native language and functions is quite powerful.
<zao>
Boost users and other modern C++ developers will probably ruin everything :)
<hkaiser>
let's write a C++ compile-time compiler ;)
<zao>
HPX seems to be death by a thousand templates :D
<heller>
Vir: ping
<K-ballo>
zao: looks useful
<zao>
Could probably use some friendlier output formats than text, but still, useful.
<zao>
Public domain, so go bananas.
<K-ballo>
that plain text output seems enough
bikineev has quit [Ping timeout: 260 seconds]
<Vir>
heller: pong
pree has quit [Ping timeout: 260 seconds]
<heller>
Vir: any status update on improved code gen for Vc2?
<Vir>
heller: it's turned into a rather involved refactoring. On the bright side, the changes are bringing down compile times considerably. It looks like I'm close to done with it. Then I'll recheck the code-gen in octotiger.
<zao>
K-ballo: I propose the new term refunctoring :P
<Vir>
heller: If there are still inliner issues I'll add a macro to turn on more [[ gnu::always_inline ]]
<Vir>
sorry
<Vir>
for some reason Travis started to segfault on all my AVX512 & KNL unit tests, though :-(
<heller>
:/
bikineev has joined #ste||ar
hkaiser has quit [Quit: bye]
pree has joined #ste||ar
bikineev has quit [Ping timeout: 240 seconds]
aserio has joined #ste||ar
<heller>
so meta classes give us operator. overload for free :D
<K-ballo>
you've been having some meta fun I see
<heller>
yeah ... I need the clang modfications now
<heller>
to play around further
<heller>
and figure out the correct syntax :P
<heller>
the paper doesn't really compile with cppx.godbolt.org ;)
hkaiser has joined #ste||ar
denis_blank has quit [Quit: denis_blank]
aserio has quit [Remote host closed the connection]
aserio has joined #ste||ar
Matombo has quit [Remote host closed the connection]
jbjnr has joined #ste||ar
Matombo has joined #ste||ar
eschnett has quit [Quit: eschnett]
ajaivgeorge has joined #ste||ar
<jbjnr>
diehlpk_work: taeguk[m] completed my gsoc evaluation. Pass. I'm not doing much mentoring, so criticize me if you need more.
<zao>
»[VS2017] Disabled notifications of changes to the open solution to work around a problem in Visual Studio that caused errant "Specified cast is not valid" compilation errors. Learn more. (case=104119)»
<zao>
(Visual Assist X)
<diehlpk_work>
jbjnr, No, perfect, but we should not announce the result in public :)
<jbjnr>
really? why not? if we ail a student, then they'll find out eventually!
<jbjnr>
^fail
jakemp has joined #ste||ar
ajaivgeorge_ has joined #ste||ar
<Vir>
david_pfander, heller: If I want to build octotiger with a -D flag that changes the Vc ABI, do I have to rebuild all of hpx?
aserio1 has joined #ste||ar
<diehlpk_work>
jbjnr, Google guidelines
<jbjnr>
ok
ajaivgeorge has quit [Ping timeout: 240 seconds]
<david_pfander>
Vir: can work, have been doing that myself, but no guarantees :) (likely exception: you're using the datapar algorithms that hpx provides)
<Vir>
david_pfander: so, all Vc code in HPX is inline/template code, I guess?
aserio has quit [Ping timeout: 246 seconds]
aserio1 is now known as aserio
<david_pfander>
Vir: probably yes, but only hkaiser knows for sure
<hkaiser>
yes
<Vir>
alright, let's see if the GCC inliner groks it now
<mcopik>
ajaivgeorge_: Hi Ajai, how it's going today? Any progress with segmented find?
<david_pfander>
Vir: BTW I'm working on AVX2/AVX512 vectorization, there are quite a few codegen problems in pretty critical places (even for AVX2)
<Vir>
david_pfander: have you also looked at AVX2 codegen with the avx ABI tag?
<david_pfander>
Vir: I've been doing that today for some time, I'm working with "Vc::datapar_abi::avx"
<Vir>
david_pfander: good, I'm more confident about that one, than about fixed_size
akheir has joined #ste||ar
<akheir>
aserio: are you coming to pxfs meeting?
<aserio>
akheir: yes, We will have to have it in your office
<david_pfander>
Vir: Am I right in thinking that through all the layers of abstraction, the datapar.h impl::<functions> still end up in the arithmetics.h (oder the <vectorISA>.h files)?
<Vir>
david_pfander: if I understand your question correctly, then yes
<heller>
david_pfander: does the tave or cori build work for you now?
<Vir>
datapar.h forwards via a detail::traits specialization to static member functions in a datapar_impl class
<Vir>
and those use free functions from e.g. detail/x86/arithmetics.h
eschnett has joined #ste||ar
<david_pfander>
heller: tave still no, cori still yes. I've gotten debug jobs through the queue on cori with moderate waiting time, so there is some productivity. Not doing much on tave right now, I've already spend too much time getting that fixed.
<david_pfander>
Vir: perfect, just wanted to be sure :)
aserio has quit [Ping timeout: 276 seconds]
aserio has joined #ste||ar
Matombo has quit [Ping timeout: 276 seconds]
Matombo has joined #ste||ar
Matombo has quit [Remote host closed the connection]
bikineev has joined #ste||ar
Matombo has joined #ste||ar
pree has quit [Ping timeout: 260 seconds]
pree has joined #ste||ar
vamatya has joined #ste||ar
david_pfander has quit [Ping timeout: 240 seconds]
bikineev has quit [Ping timeout: 255 seconds]
bikineev has joined #ste||ar
bikineev has quit [Remote host closed the connection]
Matombo has quit [Ping timeout: 260 seconds]
Matombo has joined #ste||ar
<K-ballo>
hkaiser: didn't I implement array_view some years ago?
<K-ballo>
I see coordinate.hpp, which was part of it IIRC
aserio has quit [Ping timeout: 276 seconds]
aserio has joined #ste||ar
pree has quit [Quit: AaBbCc]
Matombo has quit [Remote host closed the connection]
jgoncal has quit [Ping timeout: 246 seconds]
jgoncal has joined #ste||ar
bikineev has joined #ste||ar
EverYoung has joined #ste||ar
<hkaiser>
K-ballo: yah
<hkaiser>
it never made it in
aserio has quit [Ping timeout: 260 seconds]
hkaiser has quit [Quit: bye]
<K-ballo>
I remember not liking it.. it implemented the proposal as specified and the spec was horrible
EverYoung has quit [Ping timeout: 246 seconds]
aserio has joined #ste||ar
<github>
[hpx] atrantan opened pull request #2719: Add documentation for partitioned_vector (master...partitioned_vector_docs) https://git.io/vQnx2
aserio has quit [Quit: aserio]
Matombo has joined #ste||ar
eschnett has quit [Quit: eschnett]
zbyerly_ has joined #ste||ar
hkaiser has joined #ste||ar
EverYoung has joined #ste||ar
hkaiser_ has joined #ste||ar
hkaiser_ has quit [Read error: Connection reset by peer]
hkaiser has quit [Read error: Connection reset by peer]
EverYoung has quit [Ping timeout: 276 seconds]
hkaiser has joined #ste||ar
jakemp has quit [Ping timeout: 276 seconds]
akheir has quit [Remote host closed the connection]
eschnett has joined #ste||ar
Matombo has quit [Remote host closed the connection]
<github>
[hpx] hkaiser closed pull request #2704: Add a synchronous mapping API (master...pack_traversal) https://git.io/vQvZG