Geth rakudo/rakuast: 301 commits pushed by (Jonathan Worthington)++, (Moritz Lenz)++, (Elizabeth Mattijsen)++, (Timo Paulssen)++, (Brad Gilbert)++
review: github.com/rakudo/rakudo/compare/4...f4669071dc
00:01
rakudo/rakuast: 857f3ce562 | (Jonathan Worthington)++ | 3 files
Update RakuAST nodes for Parameter changes
00:02
rakudo/rakuast: 89b254ff76 | (Jonathan Worthington)++ | tools/templates/6.e/core_sources
Remove a moved file from 6.e core sources
Geth rakudo: 3748884da2 | (Patrick Böker)++ | 5 files
Adapt filenames in binary release scripts

In #3996 the decision was made to rename the binary release archives to also include the toolchain used. This enables us to offer release files with multiple different toolchains for the same OS.
Fixes #3996.
06:22
linkable6 RAKUDO#3996 [closed]: github.com/rakudo/rakudo/issues/3996 [Windows][build] Make the offered Rakudo (Windows) versions consistent
rakudo: 734631b678 | (Patrick Böker)++ (committed using GitHub Web editor) | 5 files
Merge pull request #4071 from patrickbkr/bin-rel-renames

Adapt filenames in binary release scripts
lizmat Files=1346, Tests=117141, 233 wallclock secs (29.91 usr 8.03 sys + 3147.92 cusr 285.13 csys = 3470.99 CPU) 10:05
MasterDuke .tell patrickb github.com/pypa/pip/blob/master/.a...AMDisk.ps1 might be interesting to steal from 10:15
tellable6 MasterDuke, I'll pass your message to patrickb
Geth rakudo: c080e59404 | (Daniel Green)++ | src/Perl6/Actions.nqp
Gen faster code for some kinds of `when` cases

If it's a type or literal, we don't need to do a full smartmatch. Unfortunately this isn't dramatically faster, since the primary cost is in generating a Backtrace for the Failure created when a case doesn't match.
... (17 more lines)
11:56
rakudo: 0006475ad0 | (Daniel Green)++ | 3 files
Don't create an expensive Backtrace if failing...

to coerce a Str to a Numeric for comparsion against a Numeric literal
  `when` case. We only need to know it didn't succeed, so add a flag to
  `val()` and `Str.Numeric()`.
The example code from the previous commit now drops from ~1.4s to ~0.35s
lizmat ah, I should probably should have squashed that :(
Geth rakudo: 260f015888 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 3 files
Merge pull request #4026 from MasterDuke17/gen_faster_code_for_some_given_when_cases

Gen faster code for some kinds of `when` cases
nqp: 69708d23eb | (Elizabeth Mattijsen)++ | docs/ops.markdown
Add documentation of Parametric Extension Opcodes

Initial version, transmogrified from
   MoarVM/docs.6model-parametric-extensions.markdown
Hopefully to be expanded upon after I grok the functionality more thoroughly.
13:04
Geth ¦ rakudo: vrurg self-assigned Blin 2020.12, round 1 github.com/rakudo/rakudo/issues/4078 14:22
Geth nqp: 829f1d42f9 | (Elizabeth Mattijsen)++ | src/vm/moar/QAST/QASTRegexCompilerMAST.nqp
Don't use an iterator if we don't have to
15:57
[Tux] Rakudo v2020.11-65-g260f01588 (v6.d) on MoarVM 2020.11-52-g31899fd37
csv-ip5xs0.841 - 0.962
csv-ip5xs-208.070 - 8.854
csv-parser26.040 - 27.397
csv-test-xs-200.391 - 0.402
test8.169 - 8.510
test-t1.913 - 1.969
test-t --race0.841 - 0.852
test-t-2035.503 - 35.616
test-t-20 --race9.694 - 10.977
16:24
Geth rakudo: dod38fr++ created pull request #4080:
fix missing =over =back in docs/running.pod
16:46
Geth ¦ rakudo: lizmat self-assigned shell output pollutes subsequent command output in REPL github.com/rakudo/rakudo/issues/4079 17:16
Geth rakudo/master: 8 commits pushed by (Elizabeth Mattijsen)++ 19:27
lizmat afk for a few hours& 19:28
timotimo what's this about, correctness or performance? 20:34
nine performance
dogbert17 nine: does t/spec/integration/advent2012-day21.t hang for you? 20:36
timotimo that seems counter-intuitive 20:41
dogbert17 offending commit seems to be 260f015888c7 20:48
linkable6 (2020-12-01) github.com/rakudo/rakudo/commit/260f015888 Merge pull request #4026 from MasterDuke17/gen_faster_code_for_some_given_when_cases
nine timotimo: as we discussed in #moarvm, MVMIter is surprisingly slow, as it doesn't JIT as well as manual atpos. atpos gets devirtualized as does the shift on the MVMIter, but the latter then does an MVM_repr_at_pos 21:06
nine And the result of nqp::shift has to be booleanized for the loop 21:07
timotimo ah, mhm 21:19
should be possible to devirt the shift on the iterator at the very least
vrurg sena_kun: are you around by occasion? 21:29
sena_kun vrurg, yes 21:34
not for long, though
nine timotimo: possible, yes. OTOH simply not using iterators gives the same performance benefit :) 21:39
vrurg sena_kun: can you have a look at github.com/Raku/roast/issues/707 when have type? Looks like rakudo archive is bundled with outdated roast. 21:41
timotimo mhm
sena_kun vrurg, during release roast is checked against master, 6.c-errata and 6.d-errata for number of times and it passes. 21:44
vrurg sena_kun: it passes, but what we have in the archive is passing with TODOs. I thought these were un-todoed by lizmat before the release. 21:45
sena_kun: I mean, nothing terrible, but a little bit not ideal. :)
sena_kun Sorry, don't really have time to think about it now. :S 21:46
I'll try to look at it tomorrow.
vrurg sena_kun: no problem whatsoever. Just a note. 21:47