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.
samcv AlexDaniel: great 00:30
yoleaux 19 Jun 2018 22:27Z <AlexDaniel> samcv: everything looks good, let's release :)
02:08 lizmat joined 03:50 tbrowder_ left 03:51 zostay left, chansen_ left, SmokeMachine left 05:11 ufobat_ joined 05:24 zostay joined 05:33 chansen_ joined 05:36 SmokeMachine joined
nine unicode.org/consortium/adopted-cha...tml#b1F42A 05:48
05:57 chansen_ left, zostay left 05:58 SmokeMachine left 06:19 zostay joined 06:21 tbrowder_ joined, chansen_ joined, SmokeMachine joined, chansen_ left, chansen_ joined 06:22 SmokeMachine left, SmokeMachine joined 06:23 nine left 06:24 nine joined 06:33 skids left 06:50 robertle joined 07:58 brrt joined 08:23 yoleaux joined
[TuxCM] Rakudo version 2018.05-121-g304239d7a - MoarVM version 2018.05-136-geb2f5a72b
csv-ip5xs0.886 - 0.895
csv-ip5xs-207.421 - 7.445
csv-parser23.654 - 23.778
csv-test-xs-200.456 - 0.480
test8.489 - 8.573
test-t2.302 - 2.666
test-t --race0.996 - 0.997
test-t-20 --race14.122 - 15.133
 08:30
08:35 [TuxCM] left 09:44 brrt left 09:46 brrt joined 09:54 brrt left
Geth rakudo: a167e6cca9 | (Jonathan Worthington)++ | src/vm/moar/spesh-plugins.nqp
Make `.?` spesh plugin bug compatible with `.`

The spesh plugin for `.?` fixed a long-standing issue with method augmented into parent classes not being visible to subclasses. While that's in theory good, in practice there was code in the wild that used `.?` in a filtering step and then `.` to call, and inconsistent behavior between the two (`.?` being fixed and `.` not yet) caused breakage. This is discussed in issue #1923. Therefore, re-introduce the bug to `.?`, until we can fix `.` too.		 09:57
synopsebot_ RAKUDO#1923 [open]: github.com/rakudo/rakudo/issues/1923 [regression] Regression related to MONKEY-TYPING and spesh-plugins merge
rakudo/rescalar: 33 commits pushed by (Jonathan Worthington)++
review: github.com/rakudo/rakudo/compare/f...56cf527acf		 10:01
jnthn Just a rebase :)
10:48 releasable6 left 10:49 releasable6 joined 11:06 brrt joined 11:30 AlexDaniel joined 12:33 raschipi joined
AlexDaniel maybe rakudo changelog should say that it comes with newer MoarVM, or something along the lines 13:06
because cool moarvm improvements are typically not seen in rakudo changelog, and it's non-obvious that there's a moarvm changelog also 13:07
Geth rakudo/rescalar: f1f3832e11 | (Jonathan Worthington)++ | 3 files
First small step for the assignment spesh plugin

So far, it just specializes the case where we have a simple assign to do (no whence, no Nil, typecheck passes). It sets guards and then turns this all into a bindattr, which spesh can then further lower to a pointer write guarded with the GC write barrier.
No worthwhile measurements available yet, due to missing guard JIT and the fact that the generated code still has a good bit of cruft that we could do with eliminating.		 13:24
brrt i'm reading 'missing guard JIT' and hear a call-to-action 13:36
timotimo jnthn: what does it mean that almost all spesh guard indices in the spesh log are 5, but a few are 1? no other values at all? 13:37
13:52 skids joined 14:31 brrt left
jnthn timotimo: That's just the index of the result entry 14:56
So it means most things are monomorphic and have 4 guards :)
timotimo is the cruft you were refering to that still needs removed the wval for the Scalar type that's left over at the end? 15:01
jnthn That, but also the guards are inserted always based on what the plugin says 15:03
But in some cases we already have a guard for that
timotimo ah, we're not eliminating them based on facts we already know? 15:04
jnthn Right
timotimo because we can't know that the guard we're looking at isn't the one that established the fact in the first place
i.e. a naive implementation would just throw out all guards all the time
jnthn Well, but it's a bit easier than that 15:05
We can look if the property we're about to insert a guard for already holds
Due to a previous guard
And then not insert
timotimo ah, that'd just be a check of the facts, then
jnthn Right
15:17 robertle left 15:54 robertle joined 15:55 perlpilot joined 16:01 brrt joined
Geth rakudo/rescalar: 48a4873ccf | (Jonathan Worthington)++ | src/vm/moar/spesh-plugins.nqp
Handle non-nominal assignment type check in plugin
 16:05
16:40 brrt left
Geth rakudo/rescalar: 802c8d873f | (Jonathan Worthington)++ | src/vm/moar/spesh-plugins.nqp
Spesh plugin handling of bindpos assignment

Which recognizes then we are doing a Scalar assignment that should bind the Scalar into an array on the first assignment, and provides a spesh result that does the exact set of things needed. For type check and no type check case.		 17:02
17:20 PufferBot joined 17:23 PufferBot left 17:25 PufferBot joined 17:27 PufferBot left 17:29 PufferBot joined 17:32 [TuxCM] joined 17:48 [TuxCM] left, ggoebel left 17:52 [TuxCM] joined 18:05 ggoebel joined
[Coke] . 18:09
18:09 [TuxCM] left 18:19 ggoebel left 18:35 ggoebel joined 18:48 brrt joined, brrt left 19:14 lizmat left 19:23 lizmat joined 19:55 [TuxCM] joined 19:59 ufobat_ left 20:37 robertle left 20:39 [TuxCM] left 21:00 lizmat left 21:04 skids left 21:16 raschipi left 21:23 victors joined 21:29 lizmat joined 21:56 victors left 22:09 MasterDuke joined 22:22 Kaiepi joined 22:24 skids joined 22:26 MasterDuke left 22:52 perlpilot left
Geth rakudo/dyn-opt1: dbc7c34829 | (Zoffix Znet)++ | src/core/stubs.pm6
Make dynamic var lookup ~15% faster | W4anD0eR96++ | timotimo++

W4anD0eR96++ for asking the question that lead to this
  timotimo++ for helping with spesh stuff
  -----
... (7 more lines)		 23:06
23:07 MasterDuke joined 23:52 lizmat left 23:56 TimToady joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!