02:11 benchable6 joined 02:17 benchable6__ left
releasable6 Next release in ≈6 days and ≈15 hours. There are no known blockers. Please log your changes in the ChangeLog: github.com/rakudo/rakudo/wiki/ChangeLog-Draft 03:00
07:47 melezhik joined 09:57 melezhik left 11:05 librasteve_ joined
timo found something that was making the compilation of BOOTSTRAP/v6c.moarvm take longer than it needed to 12:11
almost 10% of its time spent in the linefileof method of HLL::Compiler :D 12:14
we never anticipated a very large number of "#line" directives in nqp code, so the just over 3k line directives in bootstrap/v6c.nqp caused very poor performance on our linear scan through the directives
not helpful for the performance of pretty much any other compilation i'm afraid 12:18
ab5tract m: Q| class A { has @.a is readonly = <a b c> }; my $a = A.new; $a.a = <c d e>; dd $a.a |.AST.EVAL 12:34
camelia ===SORRY!===
Unknown compilation input 'qast'
ab5tract lizmat: maybe this is an indication that things are slightly better for RakuAST? 12:35
timo do we need the experimental use for something like that?
ab5tract Nope. It’s a handy patch for debugging that probably should not survive the release 12:37
m: Q| say “hallo!” |.AST.EVAL
camelia hallo!
timo mhm 12:39
12:45 melezhik joined
timo m: say 2553181 * 100 / 2571245 12:49
camelia 99.297461
13:12 camelia left, camelia joined
nine lizmat: camelia should be in #raku-doc now 13:12
timo m: say 2523212 * 100 / 2571245 13:30
camelia 98.1319166
timo m: say 2521040 * 100 / 2571245 13:37
camelia 98.0474439
timo m: say 2518832 * 100 / 2571245 13:44
camelia 97.9615711
timo haha, oh yeah!
.o( i might have b0rked it in the middle and gotten wrong results? ) 14:01
m: say 2516506 * 100 / 2571245 # or maybe it's been actually correct 14:02
camelia 97.8711091
timo but the code is ugly and i don't like it 14:04
m: say 2493135 * 100 / 2571245 # or maybe it's been actually correct 14:06
camelia 96.962172
timo huh. i made the 12-rakuast/origins.rakutest break, it now includes the lines of "method throw" and "sub die" in the output, which the test is not expecting 14:15
i think i found the mistake (in something else i did before) 14:34
14:55 melezhik left
timo why is it so easy to mess up a "simple" binary search :D 15:33
ab5tract What is this equation? 15:50
timo you mean the numbers i've been calculating? 15:51
ab5tract Yeah. Are you saying Raku is giving the wrong result? 16:27
timo no 16:28
i've been decreasing the size of the v6c bootstrap nqp source file 16:31
Geth nqp/faster_comp_line_directives: ddfc4348b1 | (Timo Paulssen)++ | 3 files
Transform comp_line_directives from an AoS to a SoA

Instead of having an object array with many three-element obj-arrays in it, have two int-arrays and a str-array.
This removes one indirection from line number lookup when searching through the array of comp line directives. Also, this should use less memory at ... (15 more lines)		 17:37
nqp: timo++ created pull request #844:
Transform comp_line_directives from an AoS to a SoA
timo behold!
m: say 2065410 * 100 / 2142295 # difference in size of gen/moar/ast.nqp with my changes in an upcoming rakudo branch 17:49
camelia 96.4110918
timo i think the code there is actually giving wrong results now that i "improved" it 19:33
20:57 Guest66 joined 21:00 Guest66 left
[Coke] jdv: am I still on for release in a week? 22:35
If there are any bumps left to do for moar/nqp into nqp/rakudo, can someone press the button on those? 22:49
releasable6 Next release in ≈5 days and ≈19 hours. There are no known blockers. Please log your changes in the ChangeLog: github.com/rakudo/rakudo/wiki/ChangeLog-Draft 23:00
lizmat nine++ thanks! 23:07

Powered by Cro and the Raku® Programming Language.

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

Thank you!