Perl 6 language and compiler development 2.0 | Logs at irclog.perlgeek.de/perl6-dev/today | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by moderator on 6 May 2018.
00:05 BenGoldberg joined 00:17 statisfiable6 joined
AlexDaniel weekly: reportable (weekly) gist.github.com/ea59754327f31447c2...8be869f03b 01:19
notable6 AlexDaniel, Noted!
AlexDaniel weekly: reportable (monthly) gist.github.com/e750c5101ea4a6d8d3...7e1851d899
notable6 AlexDaniel, Noted!
01:57 ilbot3 joined
moderator Perl 6 language and compiler development 2.0 | Logs at irclog.perlgeek.de/perl6-dev/today | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
03:25 nativecallable6 joined 06:33 quotable6 joined
lizmat Files=1239, Tests=76314, 314 wallclock secs (15.43 usr 5.44 sys + 2156.33 cusr 217.72 csys = 2394.92 CPU) 06:40
06:55 ufobat joined 07:03 robertle joined 07:57 jsimonet joined
lizmat weekly 08:46
notable6: weekly
notable6 lizmat, 14 notes: gist.github.com/ad880ec69f78c34a86...878fca1e20
Geth star: fcd64f2c13 | (Steve Mynott)++ | docs/announce/2018.04.md
add 2018.04.1.md
 09:34
|Tux| Rakudo version 2018.04.1-42-gf14dc6bdd - MoarVM version 2018.04.1-59-g8521118b1
csv-ip5xs0.846 - 0.930
csv-ip5xs-208.344 - 8.759
csv-parser36.572 - 37.169
csv-test-xs-200.463 - 0.486
test8.908 - 9.025
test-t2.347 - 2.387
test-t --race0.989 - 1.000
test-t-2042.053 - 43.318
test-t-20 --race14.541 - 15.041
 10:20
2018-05-02 14:17:19 test-t 2.372
2018-03-29 14:36:32 test-t 2.372
2018-05-05 13:00:58 test-t 2.362
2018-05-02 14:24:37 test-t 2.360
2018-05-07 12:11:27 test-t 2.347
:) 10:21
samcv i'm ready to make string search faster again.
changes landing soon for 5-15x faster
for indexing a flat 32 bit haystack 10:22
wonder if that will make test-csv faster too
|Tux| too
samcv is curious if you saw any changes from github.com/MoarVM/MoarVM/commit/6d69c6b08 10:25
that log you have, does that mean the last times you have is between 5-02 and 5-07? 10:27
timotimo samcv: i think that's the very best times ever 10:28
samcv 15% speedup for all hash access or setting i'm guessing may account for it though 10:29
|Tux| what timotimo said
Full log of all recorded timings: tux.nl/Talks/CSV6/speed.log 10:30
|Tux| will return one entry that ****s up tyhe graphs ... 10:31
fixed 10:32
samcv, tux.nl/Talks/CSV6/speed4.html scroll to the end of the page
samcv [Tux]: what am i looking for? 10:34
|Tux| you wanted to know the meaning of the speeds 10:35
samcv oh ok
|Tux| no gaps
If I see a new low, I post the top-5 overall
2.347 is a new low
samcv ah ok 10:44
Geth nqp: c5d1c0d7f1 | (Samantha McVey)++ | tools/build/MOAR_REVISION
Bump MoarVM for string indexing speedups

Changes: 2018.04.1-52-gdf91ba7..2018.04.1-64-g6e2b54c7d e71ba7381 Implement a optimized memmem 32 bit search 4-15x faster 37d5b9850 Factor out the memmem wrapper for searching Grapheme32 blobs a91b3d465 insert prof_allocated in the exact right place a51902837 commit a missing break statement ... (7 more lines)		 10:49
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...g6e2b54c7d
rakudo: 6ae0526f22 | (Samantha McVey)++ | tools/build/NQP_REVISION
Bump NQP/Moar + jvm/js fixes

Changes: 2018.04-19-g19e936b6f..2018.04.1-44-gc5d1c0d7f c5d1c0d7f Bump MoarVM for string indexing speedups e77f3306f Fix a minor typo in a test description d3958afdd [JVM] Ignore hint for multiple inheritance 707eb7713 Dump annotations in predictable order ... (27 more lines)		 10:51
¦ rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....gc5d1c0d7f
samcv [Tux]: any chance you can run the csv test again with the bump? 10:52
|Tux| running ... 10:53
samcv :)
timotimo oooh, i'm looking forward to seeing that! 10:54
|Tux| Side note: 10:56
+ cd /pro/3gl/CPAN/rakudo/install
+ zef update
===SORRY!===
This type cannot unbox to a native string: P6opaque, Failure
timotimo oh, huh 10:57
|Tux| I don't see a 4-15% speedup :P 10:58
samcv well. of 32 bit string indexing. maybe you don't have that
(flat 32 bit string)
not sure what yours are
[Tux]: is that a new error?
timotimo how do 32bit and 8bit compare if they have the same contents? 10:59
|Tux| I think it is from p6doc, which is not required for the CSV tests
as said, a side note
samcv timotimo: not sure. 8bit should generally win, but this latest change makes them much more competitve
timotimo that's very cool 11:00
samcv it's similar to the memmem we use for 8 bit nstrings, and the jump table uses modulus 256 so we can still keep a small enough array (256 elements)
so it's better in all cases, and we can always optimize how we determine the modulus or change it to some other type function, or do a higher modulus and larger array maybe 11:01
timotimo could also have a sparse array for this, for example with a two-layer approach or something 11:02
samcv that sounds slower? 11:03
i'd think it'd be faster just to not jump forward quite as much than adding more complexity
timotimo mhm
samcv but we could experiment
if the needle is one grapheme it does a fast path of just going through the blob which optimizes pretty well. i may also decide to write one that indexes faster for length 2 needles of 32 bits 11:04
timotimo ah, damn, the travis-ci failure for moarvm is because of the bump and just restarting the jobs doesn't help because it wants to use latest nqp 11:05
samcv :( 11:06
timotimo we could probably put some logic into the CI script to get that right
but i CBA right now :) 11:07
|Tux| Rakudo version 2018.04.1-43-g6ae0526f2 - MoarVM version 2018.04.1-64-g6e2b54c7d
csv-ip5xs0.843 - 0.868
csv-ip5xs-208.479 - 8.571
csv-parser36.122 - 36.556
csv-test-xs-200.451 - 0.490
test8.819 - 9.070
test-t2.298 - 2.319
test-t --race0.960 - 0.979
test-t-2041.265 - 41.967
test-t-20 --race14.463 - 14.746
 11:10
timotimo wow
|Tux| so, yes, definitely faster
samcv++
samcv yay! 11:11
|Tux| tux.nl/Talks/CSV6/images/test-t-d180.png
Geth roast: 856b60a41b | (Elizabeth Mattijsen)++ | S05-mass/rx.t
Unfudge now passing test for RT #124898
 11:19
synopsebot RT#124898 [new]: rt.perl.org/Ticket/Display.html?id=124898 [REGEX] S05-mass/rx.t line:503 reason: 'anchors and after'
Geth star: 742f5f59e0 | (Steve Mynott)++ | tools/star/windows-msi.pod
Inline is now broken on windows
roast: 959cc3f4ae | (Elizabeth Mattijsen)++ | S05-metasyntax/lookaround.t
Unfudge now passing tests for RT #131964
 11:20
synopsebot RT#131964 [new]: rt.perl.org/Ticket/Display.html?id=131964 [REGEX] Anchors ^, ^^, $, $$, «, » confused in <?after>
timotimo yay 11:22
samcv [Tux]: you only split on single codepoints yes? 11:33
aside from \\r\\n?
|Tux| I do, but it is not forbidden to set the sep_char to "\\x{1223}\\x{3425}"
|Tux| checks ... 11:34
there are tests that check if setting sep to longer works, but there are no tests (yet) that check if that actually works :( 11:36
samcv working on another speed improvement. it doesn't benefit anything if the first grapheme of the needle is not in the string. but if it is in the string in many places, then it makes it 45% faster (on top of the previous improvement) 11:37
|Tux| I had one case in the past where the sep was "--" (two minus-signs)
samcv such as ("abcdefghij" x 10000).index("ac") for example (if haystack was flat 32 bit string) 11:38
|Tux| FYI: I do not really care if a multi-codepoint-sep is (much) slower than a single-codepoint-sep (or quo, or esc, or eol) 11:39
98% of the time (probably more) the sep is either ',' or ';'
Geth star: 1920a6380e | (Steve Mynott)++ | tools/star/release-guide.pod
Server moved. Just one now
|Tux| over 75% of the time, quo is '"', with the other 25% '\\' or undefined 11:40
over 75% of the time, quo is '"', with the other 25% "'" or undefined
over 75% of the time, esc is '"', with the other 25% '\\' or undefined
samcv yeah that's what i expected
Geth ¦ nqp: MasterDuke17 assigned to timo Issue Profiles missing some GCs github.com/perl6/nqp/issues/440 11:54
¦ nqp: MasterDuke17 self-assigned Profiles missing some GCs github.com/perl6/nqp/issues/440
star: 040ead6ca3 | (Steve Mynott)++ | tools/star/release-guide.pod
tweak automated annouce instructions and suggest use of git over ssh
 12:02
stmuk .tell Zoffix rakudo.org/post/announce-rakudo-st...se-2018-04 is in place and mostly looks great. There is a minor issue with a scroll bar being present maybe due to fixed width being too small issue 12:04
yoleaux stmuk: I'll pass your message to Zoffix.
Geth star: 652bfc9c1d | (Steve Mynott)++ | tools/star/release-guide.pod
mention perl6 WWW dep needed for automated announce
 12:07
stmuk zoffix++ # slaying the wordpress monster 12:18
stmuk runs away to enjoy the sun 12:19
virtualsue++ # macOS build
AlexDaniel weekly: rakudo star 2018.04 released 12:37
notable6 AlexDaniel, Noted!
samcv yay!
timotimo nice. 12:41
AlexDaniel stmuk++ 12:43
13:23 ilbot3 joined
moderator Perl 6 language and compiler development 2.0 | Logs at irclog.perlgeek.de/perl6-dev/today | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
14:11 committable6 joined, quotable6 joined, undersightable6 joined
Geth nqp: 615d4643a3 | (Elizabeth Mattijsen)++ | tools/build/MOAR_REVISION
Bump nqp to get moar jitted opcodes
 16:11
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...5-g1f92140
rakudo: 852464a603 | (Elizabeth Mattijsen)++ | tools/build/NQP_REVISION
Bump NQP to get the latest bunch of jit fu
 16:26
¦ rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....5-g615d464
lizmat [Tux]: on my machine, test-t got faster again after ^^^ 16:29
|Tux| <zucht>Oké, ik draai de test wel weer …</zucht> 16:30
Luister ik ondertussen NOG een keer naar www.youtube.com/watch?v=UCT9VpLoN84
lizmat wait until you see the result :-)
|Tux| 😱 16:35
not spectecular though
16:36 lizmat joined
lizmat :-( 16:37
too bad
|Tux| it is still running 16:38
Rakudo version 2018.04.1-44-g852464a60 - MoarVM version 2018.04.1-76-gb38df02a8
csv-ip5xs0.833 - 0.854
csv-ip5xs-208.300 - 8.416
csv-parser36.170 - 36.476
csv-test-xs-200.460 - 0.477
test8.484 - 8.689
test-t2.227 - 2.313
test-t --race0.937 - 1.028
test-t-2041.727 - 42.304
test-t-20 --race14.683 - 15.226
 16:43
8.689
8.689
(Bouncing mouse)--
16:47 dogbert17 joined
lizmat [Tux]: looks to me another ~3% or so 16:50
16:50 statisfiable6 joined
|Tux| indeed, but you made me expect more 16:50
though below 2 is now in sight 16:51
timotimo hah. 16:53
just another ~10%
lizmat I would say 5% but yeah :-) 16:55
fwiw, this last round is by brrt's JIT template additions
timotimo to get from 2.227 to 2.0 looks like about 10% to me? 16:56
m: say 2.227 * 0.9; say 2.227 * 0.95
camelia 2.0043
2.11565
lizmat timotimo: ah, you mean it like that... yes, you're right :) 16:58
17:36 lizmat joined 17:37 tailgate joined
Geth geth: 0fb3bb8552 | (Zoffix Znet)++ (committed using GitHub Web editor) | lib/Geth/Plugin/GitHub.pm6
Shorten list of authors if there are many
 18:26
18:27 Geth joined
Geth rakudo: df4de34ab5 | (Elizabeth Mattijsen)++ | CREDITS
Proper spelling of Paweł Murias
 18:33
18:47 unicodable6 joined 19:00 Kaiepi joined 19:01 BeastieBot joined 19:20 brrt joined 19:54 Kaypie joined 20:07 |Tux| joined 20:29 robertle_ joined 20:48 jnthn1 joined 20:57 ilmari[m] joined
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/05/07/...-solution/ 21:35
21:38 ilmari[m] joined 21:48 dct joined 21:53 nativecallable6 joined, unicodable6 joined 22:17 skids joined 23:43 BenGoldberg joined
MasterDuke .ask [Tux] could you generate some graphics for test-t-20 and test-t-20 --race ? i feel like those haven't gotten much faster, which is odd, cause i'd expect it to be more visible with them 23:45
yoleaux MasterDuke: I'll pass your message to [Tux].

Powered by Cro and the Raku® Programming Language.

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

Thank you!