00:00 lizmat joined 01:23 vendethiel joined 03:20 vendethiel joined 03:40 TimToady joined 04:24 vendethiel joined 05:37 lizmat joined 06:21 vendethiel joined 07:18 lizmat joined 07:51 FROGGS joined 08:28 travis-ci joined
travis-ci MoarVM build passed. Tobias Leich 'remove extra call to uv_run when closing a pipe 08:28
travis-ci.org/paultcochrane/MoarVM/...s/59010291 github.com/paultcochrane/MoarVM/co...00558d5c9e
08:28 travis-ci left 09:01 vendethiel joined, FROGGS joined 09:15 travis-ci joined
travis-ci MoarVM build passed. Paul Cochrane 'Free resources for **pieces before returning' 09:15
travis-ci.org/paultcochrane/MoarVM/...s/59012819 github.com/paultcochrane/MoarVM/co..._resources
09:15 travis-ci left 13:03 vendethiel joined 13:09 dalek joined
timotimo god damn it, jnthn 13:51
i was debugging why my stuff won't work 13:52
then i found out the text in the template is "PROFIELR_OUTPUT"
tadzik :D 13:54
p6profiler-qt now has progressbars all over the place 13:55
timotimo i'm working on a patch that'll let you provide a --profile-filename=test.json and it won't output the html around the json blob 14:00
the code works, but isn't that pretty to look at ATM
jnthn So typo! 14:27
Sorry, timotimo!
It was probably 2am or something when I was doing that part, and I probably just copy-psated the thing :)
timotimo :)) 14:31
yes, and i re-typed it into my "new" code
hooray, we're writing out the json blob 14:48
jnthn :) 14:59
timotimo 200 MB of profile data written ...
jnthn o.O 15:00
timotimo best thing: with the "write directly" approach, the cpu usage is above 90%
hm. nearly 90%
jnthn: it's the core setting compilation (minus optimize)
you'd expect that to be huge
i forgot to put a "time" in front, i think 15:04
:S
15:06 brrt joined
timotimo hmm. i still see many, many brk calls with ever growing addresses 15:07
i wonder what keeps growing the heap 15:21
500 megabytes of profile dump 15:22
between 200 and 250 kilobytes per seconds being written 15:24
huh, the frequency of brk calls has increased significantly 15:55
one gig of profiler data written 16:07
nwc10 the only way I was able to profile Test::Harness for the perl core tests was to add compression to Devel::NYTProf 16:08
this might be what you need here
timotimo we do want a sort of compression for the profiler 16:20
but that's more a semantic compression, not really a regular data compression
16:20 flussence joined 16:33 colomon joined 16:48 flussence joined 16:59 FROGGS joined 17:00 lizmat joined 17:05 lizmat_ joined 17:10 zakharyas joined 17:32 lizmat_ joined 17:48 brrt joined 18:16 flussence joined
nwc10 current bytecode version is "4", but we seem to have code supporting back to 2 18:20
can 2 and 3 go?
jnthn I think so. 18:22
We already updated serialization version
So there's no way you can run any really old anyway
nwc10 that was my assumption - anything that old will have too old a serialization blob 18:23
19:31 brrt joined 21:31 zakharyas joined