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/
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
bikineev has joined #ste||ar
ajaivgeorge_ has quit [Ping timeout: 255 seconds]
ajaivgeorge_ has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
ajaivgeorge_ has quit [Ping timeout: 248 seconds]
ajaivgeorge_ has joined #ste||ar
patg has joined #ste||ar
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
<K-ballo>
wash[m]: could you take a look at the std-atomic branch when you have the chance? it's choking on some boost::lockfree replacement code you wrote (latest build https://circleci.com/gh/STEllAR-GROUP/hpx/7388)
Matombo444 has joined #ste||ar
Matombo has quit [Ping timeout: 260 seconds]
eschnett has joined #ste||ar
parsa has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
Matombo444 has quit [Remote host closed the connection]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
<wash[m]>
K-ballo I am traveling right now, I'll email myself a reminder to take a look
ajaivgeorge_ has quit [Quit: ajaivgeorge_]
hkaiser has quit [Quit: bye]
vamatya has joined #ste||ar
vamatya has quit [Ping timeout: 260 seconds]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
parsa has quit [Quit: *yawn*]
patg has quit [Quit: This computer has gone to sleep]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
mars0000 has joined #ste||ar
mars0000 has quit [Ping timeout: 240 seconds]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
vamatya has joined #ste||ar
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
vamatya has quit [Ping timeout: 268 seconds]
taeguk has joined #ste||ar
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
Matombo has joined #ste||ar
zao has quit [Ping timeout: 248 seconds]
zao has joined #ste||ar
Matombo has quit [Remote host closed the connection]
david_pfander has joined #ste||ar
bikineev has quit [Remote host closed the connection]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
david_pfander has quit [Quit: david_pfander]
mcopik has joined #ste||ar
david_pfander has joined #ste||ar
david_pfander has quit [Client Quit]
david_pfander has joined #ste||ar
david_pfander has quit [Client Quit]
david_pfander has joined #ste||ar
david_pfander has quit [Client Quit]
david_pfander has joined #ste||ar
david_pfander has quit [Client Quit]
david_pfander has joined #ste||ar
david_pfander has quit [Client Quit]
david_pfander has joined #ste||ar
<mcopik>
parsa[w]: yt?
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
hkaiser has joined #ste||ar
bikineev has quit [Ping timeout: 240 seconds]
K-ballo has joined #ste||ar
Matombo has joined #ste||ar
<jbjnr>
Wow! I got my first one of these "Stack overflow in coroutine at address 0x0000000000000008."
<hkaiser>
so it works!
<jbjnr>
not sure. it happens during shutdown
<jbjnr>
abd ubcreasubg the stack size does not make it go away!
<jbjnr>
increasing^
<jbjnr>
<what were my fingers doing then>
<hkaiser>
address 0x8 looks suspiciously like a call to a not implemented virtual function or a jump through a nullptr
<heller>
doesn't look good
<jbjnr>
you're back!
<heller>
I am always here ;)
<heller>
just not writing much
<heller>
procrastination ftw
<heller>
I am gonna fix the libfabric/verbs stuff now
<jbjnr>
thanks.
<jbjnr>
I would volunteer, but until I fix this matrix stuff, I'm in trouble.
<heller>
yeah ... focus on that for now
<jbjnr>
stack overflow does not happen in debugger, but does on commandline
<jbjnr>
very annoying
<hkaiser>
does it happen in release mode?
<jbjnr>
hmmm. would need to recompile everything
<hkaiser>
just asking
<hkaiser>
I mean relwithdebinfo
<jbjnr>
still need a recompile
<jbjnr>
sorry
<jbjnr>
only just found it
<hkaiser>
do you do your measurements in debug mode, then?
<jbjnr>
I'm debugging on the laptop, doing real runs on the cray
zbyerly_ has quit [Remote host closed the connection]
<hkaiser>
k
<jbjnr>
just fixed one error, only to get the stack overflow one
zbyerly_ has joined #ste||ar
<jbjnr>
If I disable io threads, will normal stuff stop working
<heller>
hkaiser: if all tests pass when just replacing boost.regex with std.regex using the default regex syntax, do you think we are safe?
<jbjnr>
most code will not put any work on them, but maybe there is bacground stuff
<hkaiser>
heller: even inspect?
<heller>
inspect, good point!
<heller>
only ported hpx
<hkaiser>
jbjnr: stop believeing that the sleeping threads do any harm, pls
<heller>
well "ported"
<jbjnr>
hkaiser: I have to eliminate all the possibilities before I can be certain.
<hkaiser>
heller: hpx itself does not have too many spots using regex
<heller>
no, it was rather straight forward
<jbjnr>
when we use hpx:threads=less, then the timing stuff looks sensible, this indicates that there might be contention and disabling all other threads is one way of being sure
<heller>
the question is if it is worth to switch inspect over ... but it probably makes sense
<hkaiser>
if we want to get rid of boost.regex we need to switch inspect as well
<hkaiser>
what's threads=less?
<jbjnr>
less than more
<hkaiser>
no idea what you're talking about
<jbjnr>
on a 36 core machine, if we use 30 threads, timings looks reasonable, when using 36, things go pants up.
<jbjnr>
when I disabled the timer threads, I saw an improvment, now I want to disable the io threads too
<hkaiser>
k
<jbjnr>
but am worrried about stray calls to io thread executor coming from unexepected places
<hkaiser>
the io threads are not used by hpx itself at all
<jbjnr>
I want to put the timer threads and io threads into the resource manager framework so we can treat them the same
<hkaiser>
sure, that's where they belong
<jbjnr>
does hpx::cout use them?
<hkaiser>
hmmm, could be
<hkaiser>
need to look
<jbjnr>
is there a special executor call to use the io threads? if so I can put a breakpoitnt in it. didn't look yet
<hkaiser>
if it's used then through a lower level interface than executors
<jbjnr>
there's a lot of code in executors/execution_information+friends that troubles me
<hkaiser>
jbjnr: I don't think you need to worry about the executor information stuff
<jbjnr>
eek bbiab
denis_blank has joined #ste||ar
<jbjnr>
hkaiser: I was concerned by the information stuff because there are calls to "get_pu_mask" and things in there, which struck me as odd. This code has been poked around quite a bit by shoshana and I am now cleaning more up. things like
<jbjnr>
// customization point for interface get_pu_mask()
<jbjnr>
make me wonder - why do we need that at all?
<hkaiser>
jbjnr: this will all change in the near future, don't pay too much attention
<jbjnr>
when I change the pu mask stuff, it breaks this, so I kinda nee to pay attention a bit
<hkaiser>
this is meant to expose some necessary execution resource properties for executors
<hkaiser>
why should changing the pu mask stuf break anything
<jbjnr>
because I am moving stuff around.
<hkaiser>
it shouldn't matter how the resource manager splits things as long as the correct pu mask is reported
<jbjnr>
from runtime/etc to resource manager and thread manager etc etc
<jbjnr>
a lot of duplicated APIs etc
<hkaiser>
well, yah - that could use a clean up
<hkaiser>
jbjnr: I'd be happy to help with that
<jbjnr>
I've don a lot already, but I was secretly hoping that executor info stuff was obsolete or could otherwise be removed etc.
<hkaiser>
jbjnr: why don't you do your cleanup and leave the executor layer for me to update once you're done?
<jbjnr>
ok.
<jbjnr>
(I'll just comment out anything I break)
<jbjnr>
:)
<hkaiser>
k
<hkaiser>
I'd like to be part of that whole refactoring, though - at least from afar
<jbjnr>
you will be. when you reject my PR!
<jbjnr>
and then do it yourself anyway :)
<hkaiser>
ok
<hkaiser>
heller: #2756: can I go ahead now?
<hkaiser>
or #2761?
<hkaiser>
or any of the others?
<jbjnr>
I'll be pushin my latest stuff to github soon, I'll send you the branch info, I'm still working off the one on shoshana's repo at the moment, but keep merging in the recent master and other feature fixes.
<hkaiser>
jbjnr: ok
<jbjnr>
not made any real progress for a few days, just using apex etc to investigate what's going on
<hkaiser>
jbjnr: btw, can we close the mindmap PR?
<hkaiser>
I don't think this will get merged without any automation
<jbjnr>
close, or merge and close. I don't like the 300+ files in it myself.
<jbjnr>
ah. close without merge
<jbjnr>
then we should close it and just put the mindmap itself on the publications page as a link to interesting stuff
<jbjnr>
"If you want to fully enjoy the features of the map (like hyperlinks internal and external to the map, collapsing and unrolling branches) as well as edit it, please open the .xmind file with [XMind](http://www.xmind.net/), an open-source except for pro features, mind-mapping software.
<hkaiser>
jbjnr: so she created that manually?
<jbjnr>
yes. all of it!
<hkaiser>
uhh
<jbjnr>
she spent two months just going through the code, seeing how it all fit together and all the time was typing notes into the mind map thingy
<hkaiser>
ahh, now I understand
<hkaiser>
so this mind map is her brain dump in some sense
<jbjnr>
the mind map file itself is self contained. it was the "export to html" that produced all the images etc and bloated it to silly proprtions
<hkaiser>
then we really should preserve it
<jbjnr>
yes. That's what led to the PR. but I'll have a look and see if we can't just extract the main map file and keep that in the repo
<hkaiser>
jbjnr: or somewhere else...
<jbjnr>
like the publications links page
<hkaiser>
yes
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
<hkaiser>
jbjnr: I will need to work on integration of hpx with legacy mpi codes and that would require to have something like the resource manager
<hkaiser>
jbjnr: I'd appreciate it if you could gave me access (pointed me to) the latest version
EverYoun_ has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
<github>
[hpx] hkaiser created papi_counters (+1 new commit): https://git.io/v7Jt2
<github>
hpx/papi_counters 29095fd Hartmut Kaiser: Adding documentation for PAPI counters...
<Guest8314>
[hpx] hkaiser opened pull request #2772: Adding documentation for PAPI counters (master...papi_counters) https://git.io/v7JtK
bikineev has quit [Remote host closed the connection]
diehlpk has joined #ste||ar
hkaiser has quit [Quit: bye]
bikineev has joined #ste||ar
diehlpk_ has joined #ste||ar
diehlpk has quit [Ping timeout: 248 seconds]
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 258 seconds]
zbyerly_ has quit [Remote host closed the connection]
zbyerly_ has joined #ste||ar
zbyerly_ has quit [Ping timeout: 258 seconds]
bikineev has quit [Read error: Connection reset by peer]
eschnett has quit [Quit: eschnett]
hkaiser has joined #ste||ar
EverYoun_ has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
diehlpk_ has quit [Ping timeout: 240 seconds]
mut894674 has joined #ste||ar
mut894674 has quit [Client Quit]
bikineev has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
bikineev has quit [Remote host closed the connection]
bikineev has joined #ste||ar
mars0000 has quit [Quit: mars0000]
mbremer has quit [Quit: Page closed]
<github>
[hpx] atrantan opened pull request #2773: Add sync_images() method to spmd_block class (master...spmd_block_with_sync_images) https://git.io/v7J2K
parsa has quit [Quit: Zzzzzzzzzzzz]
Matombo has joined #ste||ar
pree has quit [Quit: AaBbCc]
parsa has joined #ste||ar
Matombo has quit [Remote host closed the connection]
EverYoun_ has joined #ste||ar
EverYoung has quit [Ping timeout: 246 seconds]
jbjnr has quit [Remote host closed the connection]
aserio has quit [Quit: aserio]
diehlpk_ has joined #ste||ar
bikineev has quit [Remote host closed the connection]
EverYoun_ has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
mars0000 has joined #ste||ar
<github>
[hpx] hkaiser created boost_date_time (+1 new commit): https://git.io/v7J6b
<github>
hpx/boost_date_time c044856 Hartmut Kaiser: Removing dependency on Boost.Date_Time
<github>
[hpx] hkaiser opened pull request #2774: Removing dependency on Boost.Date_Time (master...boost_date_time) https://git.io/v7J6h
EverYoun_ has joined #ste||ar
<K-ballo>
wow
EverYoung has quit [Ping timeout: 258 seconds]
EverYoun_ has quit [Ping timeout: 246 seconds]
EverYoung has joined #ste||ar
EverYoung has quit [Remote host closed the connection]
EverYoung has joined #ste||ar
<hkaiser>
K-ballo: I'm fighting with a weird issue related to the error handling you introduced recently
<hkaiser>
not sure how to describe it
<hkaiser>
things blow up because some exception_info gets overwritten
<hkaiser>
looks like a bad codegen :/
<K-ballo>
uh oh
<K-ballo>
were/how?
<hkaiser>
just run hello_world with a wrong command line option (Debug mode)
<hkaiser>
it will blow up while entering the function diagnostic_information(exception_info const&)
<K-ballo>
ok
<K-ballo>
building...
<hkaiser>
K-ballo: also I think that the exception_info members file, line, and function are never initialized
<hkaiser>
the code always used the default contructor of exception_info
<hkaiser>
while throwing an exception, that is
jgoncal has quit [Quit: jgoncal]
<K-ballo>
can't build fast enough
diehlpk_ has quit [Ping timeout: 240 seconds]
* K-ballo
adds a sprinkle of /bigobj
<K-ballo>
entry point must be defined.. ugh?
jgoncal has joined #ste||ar
<K-ballo>
hkaiser: what did it take to use hpx_main?
<hkaiser>
bigobj should be the default already
<hkaiser>
K-ballo: hpx_main: what do you mean?
<K-ballo>
I'm building on an external project
<K-ballo>
what were the requirements for using hpx_main.hpp
<hkaiser>
none
<hkaiser>
it #defines main() user_main and #includes hpx_init.hpp