Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2016-01-07

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
02:33 colomon joined #moarvm
02:47 ilbot3 joined #moarvm
02:47 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
02:50 colomon joined #moarvm
03:12 vendethiel joined #moarvm
04:15 colomon joined #moarvm
06:53 colomon_ joined #moarvm
07:02 domidumont joined #moarvm
07:06 domidumont joined #moarvm
07:24 FROGGS joined #moarvm
07:54 zakharyas joined #moarvm
09:48 leont joined #moarvm
12:54 virtualsue joined #moarvm
13:06 brrt joined #moarvm
13:16 pochi joined #moarvm
13:18 stmuk__ joined #moarvm
13:20 Peter_R joined #moarvm
14:08 hoelzro o/ #moarvm
14:09 hoelzro has anyone run into a situation where turning tracing on only results in a handful of Op 159 (getcode) prints?
14:09 hoelzro I'm tracing "my $sum = 0; for ^10 { $sum += $_ }" and all I see are two Op 159
14:13 JimmyZ m: my $sum = 0; for ^10 { $sum += $_ }
14:13 camelia rakudo-moar 988a78: ( no output )
14:14 JimmyZ spesh log ? or trace macro in interp.c?
14:28 hoelzro the latter
14:28 hoelzro MVM_TRACE
14:31 brrt no idea...
14:32 JimmyZ hoelzro: I know
14:33 JimmyZ you should turn off CGOTO
14:33 hoelzro what's that?
14:35 JimmyZ how do you enble TRACE?
14:36 JimmyZ make tracing?
14:36 hoelzro by stupidingly editing interp.c by hand =/
14:36 JimmyZ you should type `make tracing`
14:37 * hoelzro tries
14:39 hoelzro there we go
14:39 hoelzro thanks JimmyZ!
15:25 brrt joined #moarvm
16:00 hoelzro diakopter: yesterday you mentioned a branch that had some deserialization fixes; which branch is it?
16:32 diakopter hoelzro: cache_sc_idx
16:32 hoelzro diakopter: thanks!
16:32 diakopter but it fails on windows last time we checked
16:32 hoelzro oh, that's fine with me =) I just want to see if it fixes a bug I'm battling
16:33 diakopter it just fixes caching during bootstrapping and makes spectesting loads faster
16:34 diakopter hoelzro: feel free to merge master into it
16:35 diakopter nm I did it
16:35 dalek Heuristic branch merge: pushed 140 commits to MoarVM/cache_sc_idx by diakopter
16:36 diakopter hoelzro: see updated https://github.com/MoarVM/MoarVM/compare/cache_sc_idx
16:36 hoelzro thanks =)
17:09 Peter_R joined #moarvm
17:19 Peter_R joined #moarvm
17:21 domidumont joined #moarvm
17:26 camelia joined #moarvm
17:36 FROGGS joined #moarvm
18:08 diakopter hoelzro: anything?
18:12 ilbot3 joined #moarvm
18:12 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
18:20 lizmat http://www.dagolden.com/index.php/2589/perl-5-and-perl-6-are-mortal-enemies/   # food for thought
18:51 vendethiel joined #moarvm
19:07 hoelzro diakopter: my bug sadly still occurs =(
19:11 hoelzro if anyone's curious; https://rt.perl.org/Ticket/Display.html?id=127089
19:53 nwc10 gosh. I'd forgotten how fast a Raspberry Pi v2 is
19:53 nwc10 it's about half the speed of x86_64 with ASAN
19:53 nwc10 (on one core)(ie the setting)
20:03 * diakopter should try it on the Zero
20:04 diakopter hoelzro: that's interesting
20:04 nwc10 "might take longer". And I think you'll need to configure (more) swap as I think it peaked at over .75G of RAM
20:05 hoelzro diakopter: what's more interesting is calling .map in item context fixes it
20:05 * hoelzro has to add that to the ticket
20:08 diakopter m: https://gist.github.com/diakopter/7981dc5b1d0ff7598e53
20:08 camelia rakudo-moar ed6ec7: OUTPUT«before␤after␤»
20:08 diakopter hoelzro: what am I doing wrong? ^
20:08 hoelzro diakopter: the sub has to be from another module
20:10 diakopter oh
20:11 hoelzro I'll add that to the ticket too
20:38 leont joined #moarvm
21:19 domidumont joined #moarvm
21:26 thzre joined #moarvm
21:51 hoelzro the plot thickens: I think it's just map
22:00 hoelzro hmm, note() does it too
22:32 diakopter hoelzro: sounds like it's sink context
22:33 diakopter hoelzro: probably TimToady could diagnose it
22:34 hoelzro diakopter: sink context and list
22:34 hoelzro it also depends on what's in the block
22:34 hoelzro <...>.map({ $_ }) is fine
22:34 hoelzro <...>.map({ $_ ~ 'suffix' }) is not
22:34 hoelzro neither is .map({ $_ + 1 })
22:34 diakopter o_O
22:35 hoelzro using note() and no map is also broken
22:35 hoelzro it's...a very interesting bug =)
22:36 diakopter it might be something else also, but I feel strongly that it's an overeager optimization from a missing want or WANT
22:47 hoelzro it might be bisect-o-clock when I get home
22:48 hoelzro see if this ever worked
22:48 * hoelzro relies too heavily on bisect for P6 work
23:03 * lizmat decides to stash the setup-repositories work
23:04 lizmat it only gave about 8 msecs improvement, at the expense of not so nice code
23:04 lizmat :-(
23:05 hoelzro lizmat++ # trying things
23:05 lizmat I thought the same code was also run at bare startup, but it wasn't
23:06 lizmat 8 msecs at a bare startup, I would quickly take
23:10 timotimo hmm
23:10 timotimo i would also take that, yeah
23:10 timotimo given a startup is currently like 100 msecs
23:11 timotimo hm, or 90
23:14 lizmat more like 110 for me, atm
23:14 timotimo mhm
23:15 diakopter time perl -e 'print 1' # 2 ms
23:16 lizmat diakopter: try that with -MMoose  :-)
23:17 diakopter yeah, I submitted a pull request to Moose that halves its startup time, but it was rejected for mostly bad reasons
23:18 timotimo i'm glad it got rejected, so perl6 looks better in comparison ... :P
23:18 diakopter well :P
23:20 timotimo i'm hopeful we'll at least halve the bare startup time over the next half year
23:20 diakopter I'm quite hopeful it's doable
23:21 timotimo and it's likely that our ram usage on bare startup can also go down a good chunk
23:21 timotimo i don't actually know how we'll achieve that, though
23:21 diakopter what's the ram usage now
23:27 timotimo timo@schmetterling /p/a/ibm> time perl6 -e ''
23:27 timotimo 0.08user 0.01system 0:00.09elapsed 100%CPU (0avgtext+0avgdata 65144maxresident)k
23:27 timotimo in other words "very bad" :)
23:29 pochi_ joined #moarvm
23:30 timotimo i meant to build a separate setting that has a few classes missing to see how the ram usage does down from that
23:30 timotimo to get a little measurement of how good lazy deserialization works

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary