|
01:05
MasterDuke joined
|
|||||||||||||||||||||||||||||||||||||||
| MasterDuke | . | 01:07 | |||||||||||||||||||||||||||||||||||||
| tellable6 | 2026-10-17T19:48:22Z #moarvm <lizmat> MasterDuke did you see github.com/Raku/nqp/commit/e5e5a905f1 ? | ||||||||||||||||||||||||||||||||||||||
| MasterDuke | lizmat: yeah, i've pulled it and am working from HEAD. interestingly, i figured out how to get the `MVM_frame_lexical_lookup_using_frame_walker` calls for `$?CLASS` down. gist.github.com/MasterDuke17/1dbea...277f897050 has the patch. with that it drops to ~2m in my example, which is close to what i originally expected. | 01:10 | |||||||||||||||||||||||||||||||||||||
| *however*...it's ~10% slower. ~6s for my longer example with that patch instead of ~5.5s | |||||||||||||||||||||||||||||||||||||||
| ah ha. need to use getattr_i for the int attributes...that brings the time down to about the same | 01:39 | ||||||||||||||||||||||||||||||||||||||
|
01:57
MasterDuke left
06:21
djinni` left
06:33
djinni` joined
|
|||||||||||||||||||||||||||||||||||||||
| [Tux] |
|
07:55 | |||||||||||||||||||||||||||||||||||||
| csv-test-xs 0.014 - 0.015 | |||||||||||||||||||||||||||||||||||||||
| tux.nl/Talks/CSV6/speed4-20.html / tux.nl/Talks/CSV6/speed4.html tux.nl/Talks/CSV6/speed.log | |||||||||||||||||||||||||||||||||||||||
| lizmat has marked github.com/rakudo/rakudo/pull/6871 for review | 12:31 | ||||||||||||||||||||||||||||||||||||||
|
12:45
japhb left
12:50
japhb joined
|
|||||||||||||||||||||||||||||||||||||||
| [Coke] | c: e98da4e062 e | 14:33 | |||||||||||||||||||||||||||||||||||||
| committable6 | [Coke], ¦e98da4e: «» | ||||||||||||||||||||||||||||||||||||||
|
17:38
lizmat left
17:50
lizmat_ left
17:51
lizmat joined
18:16
iwuvmika left
18:18
iwuvmika joined
18:22
iwuvmika left
18:31
hankache left
18:32
hankache joined
|
|||||||||||||||||||||||||||||||||||||||
| Geth | rakudo/lizmat-91: 4bc2281c87 | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/Sorting.rakumod Some further optimizations: 1. Create ultra-fast path for 0,1,2 element lists 2. Remove special handling of 0,1,2 element lists from slow path 3. Get rid of type checking boundary altogether In benchmarks all number of elements are either within noise, or positive (up to 2.15x as fast). Except 3 element (-20..-10%), 4 element (-14..-7%) and 5 element lists(-9..-5% for Str). |
18:36 | |||||||||||||||||||||||||||||||||||||
|
19:27
iwuvmika joined
19:58
finanalyst joined
21:07
finanalyst left
21:17
iwuvmika left
|
|||||||||||||||||||||||||||||||||||||||
| [Coke] | blin clear through e98da4e062 | 23:57 | |||||||||||||||||||||||||||||||||||||
| linkable6 | (2026-10-22) github.com/rakudo/rakudo/commit/e98da4e062 Remove return statement from "parse_win" in val processing (#6862) | ||||||||||||||||||||||||||||||||||||||