github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
MasterDuke timotimo: did you do anything with malloc_trim()? 00:21
timotimo ah, sorry, i haven't yet 00:52
what did i even do today ...
MasterDuke ah. if you or anybody else hasn't looked into it more in the next couple days i might get a chance to take a look again 01:05
but too tired right now to do anything constructive, off to watch some of the second season of the punisher 01:06
timotimo i was super tired all day today and right now my left hand is unhappy 01:09
timotimo perhaps tomorrow will be a "more talks and presentations" day then 01:10
brrt ohai #moarvm 08:04
yoleaux 18 Mar 2019 22:51Z <lizmat> brrt: news.ycombinator.com/item?id=19425485
brrt thank you lizmat, I responded :-)
nwc10 good *, brrt 08:05
brrt good * nwc10 08:11
what if the expression JIT IR is really too small
lizmat stackoverflow.com/questions/552093...-collector # seems to be overlooked ? 09:43
Guest27580 jnthn: perhaps you missed it but here's a gist showing a test which fails on the moar-pea branch, it's almost certainly the same as the one uncovered by nwc10: gist.github.com/dogbert17/e6b0d003...9513d780cd 14:33
MasterDuke new warnings when building moarvm, MIN and MAX redefined gist.github.com/MasterDuke17/d4b72...21513e2b3b 15:32
brrt that sucks... 16:21
MasterDuke jnthn, timotimo, et at.: i've tried putting a malloc_trim(0) in a couple different places in finish_gc() github.com/MoarVM/MoarVM/blob/mast...ate.c#L132 , but keep getting a segv when profiling. any suggestions? 17:33
discord6 <Rogue> Testing the bridge bot 17:42
timotimo yo 17:43
discord6 <timotimo> yo
brrt ohai discord6 18:00
MasterDuke jnthn: did you happen to catch my earlier question about malloc_trim() and the GC and profiling? 23:34
jnthn MasterDuke: Saw it, but no idea; didn't even glance what malloc_trim is for yet 23:39
MasterDuke jnthn: no hurry or anything. the experimentation and subsequent question was inspired by www.joyfulbikeshedding.com/blog/20...k-trimming 23:41
jnthn Hmm...what's the goal of calling this?
jnthn wonders how effective that'd be... 23:42
Our FSA and GC pools keep hold of everything they ever asked for :) 23:43
I guess if one had big amounts of stuff in malloc'd memory like buffers it maybe helps, but it sounds vulnerable to fragmentation stopping it doing much still?
What does the SEGV look like? 23:44
MasterDuke well, that example that samcv was talking about a little while ago, of run()ing a cat of a file in a loop uses a bunch less ram
according to /usr/bin/time 23:45
jnthn Hm, interesting
man7.org/linux/man-pages/man3/malloc_trim.3.html claims it's thread safe, which was gonna be my first guess
MasterDuke jnthn: some info here colabti.org/irclogger/irclogger_lo...-03-15#l15 23:46
afk for a little bit 23:47
jnthn Hmm, is profiling broken without this? :)
Just to sanity check 23:48
Also, does it blow anything up (tests, spectest, etc.) without profiling?
It feels odd that profiling would cause this
And nothing else
MasterDuke jnthn: haven't run a spectest yet. but nqp and rakudo do compile ok
oh, that example may break under profiling anyway... 23:49
jnthn OK, where'd you put it? inside the `if (is_coordinator)` is the sensible spot, I think
Sleep time :) o/ 23:54

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!