Perl 6 language and compiler development | Logs at colabti.org/irclogger/irclogger_logs/perl6-dev | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by AlexDaniel on 12 June 2018.
01:22 Kaiepi left 01:56 PufferBot joined, Kaiepi joined 04:04 Kaiepi left 04:05 PufferBot left 04:44 Kaiepi joined 05:43 x[LGWs4x4i]uG2N0 joined 06:05 robertle joined 06:28 Kaiepi left 06:29 AlexDaniel left, AlexDaniel joined 06:30 Kaiepi joined 06:33 fake_space_whale left
lizmat Files=1246, Tests=76393, 381 wallclock secs (16.26 usr 5.56 sys + 2668.57 cusr 260.40 csys = 2950.79 CPU) 07:24
benchable6: 2017.01,2018.01,2018.04, 2018.06 say (^Inf).hyper.grep( *.is-prime ).skip(999).head 07:37
benchable6 lizmat, starting to benchmark the 4 given commits
lizmat, ¦2017.01: ««run failed, exit code = 1, exit signal = 0»» ¦2018.01: «0.7017» ¦2018.04: «0.7192» ¦2018.06: «0.6407» 07:38
lizmat, benchmarked the given commits and found a performance difference > 10%, now trying to bisect
lizmat, gist.github.com/ee98308dcca3561d46...128bc4de20
lizmat benchable6: 2017.09,2018.01,2018.04, 2018.06 say (^Inf).hyper.grep( *.is-prime ).skip(999).head 07:39
benchable6 lizmat, starting to benchmark the 4 given commits
lizmat, ¦2017.09: «10.0265» ¦2018.01: «0.6959» ¦2018.04: «0.6106» ¦2018.06: «0.6597» 07:40
lizmat, benchmarked the given commits and found a performance difference > 10%, now trying to bisect
lizmat, gist.github.com/879bf1a494fb43e1e3...3e79127896 07:42
09:33 skids joined 10:01 lizmat left 10:02 skids left 10:04 Ven`` joined 10:11 ilogger2 joined 10:49 reportable6 left, reportable6 joined, ChanServ sets mode: +v reportable6
AlexDaniel retoasts 11:09
jnthn . 12:12
yoleaux 03:09Z <Zoffix> jnthn: does spesh do anything interesting with type constraints? What's the answer to the question: "Is there any potential performance benefit to specifying `Int` type constraint instead of just leaving it at Mu/Any?"
jnthn .tell Zoffix In various places, Mu is treated as "don't even type check this", which is why we can use it to pass around non-Perl 6 types (e.g. NQP types, VM-provided objects). So that sounds like a win. But then if we have `sub foo(Int $i) { }` then it spots Int !~~ Iterable and Iterable !~~ Int and then emits code that just deconts any incoming value but does *not* wrap it up in a readonly Scalar. For 12:16
yoleaux jnthn: I'll pass your message to Zoffix.
jnthn Any it can't assume that, so conservatively emits code to create the container to make sure it stays as an item.
.tell Zoffix So far as spesh goes, it will try to eliminate whatever type checks it can, usually replacing them with much cheaper deopt guards. In some cases, it do one guard of something and then avoid re-checking later, so removing type checks from whole chains of calls. 12:18
yoleaux jnthn: I'll pass your message to Zoffix.
12:20 Ven`` joined
jnthn .tell Zoffix In summary, you stand to get better memory behavior with declaring the types, and in places where the default would be Any then it's all the same to spesh, so Any has no performance benefit over Int. 12:21
yoleaux jnthn: I'll pass your message to Zoffix.
[Tux] Rakudo version 2018.06-234-gcb2ea9328 - MoarVM version 2018.06-309-g428ed0e2a
csv-ip5xs1.209 - 1.268
csv-ip5xs-208.241 - 8.538
csv-parser24.545 - 24.581
csv-test-xs-200.444 - 0.452
test10.355 - 10.552
test-t2.386 - 2.389
test-t --race0.984 - 1.031
test-t-2041.290 - 42.179
test-t-20 --race13.435 - 13.779
 13:06
Geth nqp/truffle: 5 commits pushed by (Paweł Murias)++ 13:08
13:08 pmurias joined 13:16 Ven`` left
Geth nqp/truffle: 4361c5c07d | (Paweł Murias)++ | src/vm/jvm/bin/run_tests.pl
[truffle] Run passing tests
 13:28
13:47 Ven`` joined 14:12 thundergnat joined
thundergnat Some more information about the module build failure I reported yesterday. colabti.org/irclogger/irclogger_log...07-21#l216 14:13
They are still happening for me, with completely fresh install of Rakudo HEAD and fresh install of zef into completely emptied directories. 14:14
See this gist for exact error information: gist.github.com/thundergnat/057aef...553cc3323c
I still haven't filed it as an issue because I'm still not convinced I am not causing it somehow... but I can't figure out how. :-( 14:15
Anyway, I can't stay online continuously, but I read the logs pretty regularly, if you need/want more info, ping me and I'll get back to you. 14:17
14:18 thundergnat left 14:23 pmurias left 14:25 pmurias joined
dogbert17 I see the same problems as thundergnat 15:01
perhaps it has something to do with the fact that anonymous access seems to be disallowed for the host used in the tests 15:10
Geth rakudo: AlexDaniel assigned to lizmat Issue `please 'use experimental :pack'` even if code does exactly that github.com/rakudo/rakudo/issues/2111
AlexDaniel assigned to jnthn Issue `please 'use experimental :pack'` even if code does exactly that github.com/rakudo/rakudo/issues/2111

  Kaiepi++ created pull request #2116: Implement native-descriptor attribute for IO::Socket::Async		 15:19
AlexDaniel and yes:
c: a847abe523^,a847abe523 package Foo { use experimental :pack; sub go() is export { "".encode.unpack("*") }; }; import Foo; go()
committable6 AlexDaniel, ¦a847abe523^: «» ¦a847abe: «Use of the 'unpack' method is experimental; please 'use experimental :pack'␤ in sub go at /tmp/juXla5Uwul line 1␤ in block <unit> at /tmp/juXla5Uwul line 1␤␤ «exit code = 1»»
AlexDaniel (2018-07-18) github.com/rakudo/rakudo/commit/a8...9c3738c291 15:20
that was a bit surprising to me
jnthn: ↑ :) 15:21
.tell thundergnat thanks GH#2111
yoleaux AlexDaniel: I'll pass your message to thundergnat.
synopsebot GH#2111 [open]: github.com/rakudo/rakudo/issues/2111 [regression][⚠ blocker ⚠] `please 'use experimental :pack'` even if code does exactly that
jnthn AlexDaniel: Hm, don't think I did anything much with pack/unpack? 15:28
Not that I remember anyway :P
AlexDaniel jnthn: which is why it's so interesting, yes!
jnthn Agree the error looks odd, though
AlexDaniel I mean, it's bisected to github.com/rakudo/rakudo/commit/a8...9c3738c291 15:29
which is directly these changes: github.com/perl6/nqp/commit/192523...11979313e3
well, I mean these: github.com/MoarVM/MoarVM/compare/2...gfae91a6eb
jnthn is a bit surprised about a lexical pragma affecting a method... 15:30
AlexDaniel I know! I'm doubting committable a bit :) but committable never ever gave a wrong result so far…
ah, you mean that… 15:32
jnthn Ah
AlexDaniel 🤷
jnthn github.com/rakudo/rakudo/blob/2fa0...f.pm6#L285
That makes it look more related
AlexDaniel fresh toasts: toast.6lang.org/ 15:52
several modules unhappy because of pack
jnthn That looks much better than the previous run 15:55
15:55 Ven`` left
jnthn Ah, I still should put in that Proxy compat thing, though that's easy 15:56
AlexDaniel jnthn: ↑ a ticket for convenience
jnthn Thanks
I'll see if I can look at the tomorrow. Kinda need a bit of time off it today.
AlexDaniel that's OK, I hope it's clear that nobody should be holding their breath for this release :) 16:07
although I hope we'll get somewhere by the end of the following week 16:08
jnthn I sure hope so
fwiw, I'm on vacation from Friday
For a week
Can almost certainly deal with the Proxy thing tomorrow, and probably the other issue too. 16:09
AlexDaniel well yeah, once the issue with proxy and another one with pack are resolved I'm hoping to see a much cleaner result from toaster 16:10
other blockers are hopefully not busfactored on you :) 16:11
jnthn: btw it was amazing to see all the blockers disappear this week, very nice job 👍
16:12 fake_space_whale joined
jnthn Thanks. It was quite a bit of work, though I'm glad to have them resolved. :) 16:14
Means I'll be able, especially after vacation, to get back to the more fun task of putting in new optimizations. :-)
AlexDaniel heh, it's a lot of fun doing release prep these days! :) 16:15
jnthn I'll aim to make the next one less "fun" for us all. :-) 16:17
AlexDaniel I do find it somewhat enjoyable though. Bisecting stuff, trying to make sense of all the failures 16:19
otherwise it's just changelogging and hitting the button to cut the release… boring :) 16:20
16:20 fake_space_whale left 16:34 fake_space_whale joined 16:51 robertle joined 17:40 sivoais left 18:11 sivoais joined 18:14 pmurias left 18:20 pmurias joined
19:29 Ven`` joined 19:35 Kaiepi joined 20:08 dct joined 20:11 dct left
Geth nqp: Kaiepi++ created pull request #488:
Implement stubs for getting file descriptors from async sockets		 20:23
20:31 Kaiepi left, Kaiepi joined 20:46 cognominal joined 20:47 Ven`` left 20:48 Ven`` joined 20:54 pmurias left 21:00 brrt joined 21:07 MasterDuke joined 21:15 cognominal left 21:16 cognominal joined 21:23 timo joined 21:41 Ven`` left 21:57 lizmat joined 22:18 robertle left 22:22 brrt left
Geth rakudo: jmaslak++ created pull request #2117:
Improve err for obs positionals w/ quantifiers		 23:02
23:06 timo left 23:45 fake_space_whale left

Powered by Cro and the Raku® Programming Language.

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

Thank you!