|
00:30
acidsys_ left
00:32
acidsys joined
01:56
kylese left
01:57
kylese joined
02:15
kylese left,
kylese joined
02:30
chenggong7788 joined
05:08
leppard left,
wayland76 left
05:41
leppard joined
06:46
abraxxa joined
06:47
abraxxa left
06:48
chenggong7788 left
06:49
abraxxa joined
07:08
belluzj joined
07:56
rnddim joined
07:57
ShimmerFairy left,
rnddim left
07:58
rnddim joined
08:02
rnddim left,
rnddim joined
08:15
lizmat joined
08:18
lizmat left
08:19
lizmat joined
08:29
belluzj left
09:13
lizmat left,
abraxxa left,
rnddim left
09:14
rnddim joined,
rnddim is now known as ShimmerFairy
|
|||
| Voldenet | >i wonder why rakupp benchmarks don't use arm64 rakudo | 10:05 | |
| the obvious reason could be that the author just doesn't have arm64 box | |||
| …another obvious reason is that rakupp interpreter has slower runtime, but native is quite fast | 10:26 | ||
| those are x86_64 results - i've deduced startup time from all benchmarks, because that's non-relevant for anything that actually needs performance i.tanomi.eu/lzshu.txt | 10:29 | ||
| or rather, if your script takes 500ms to load, you don't really care about perf :P | |||
| s/load/execute/ | |||
| it's puzzling why streq and fib are that much slower on rakupp | 10:32 | ||
| the machine is old skylake i5, benchmark is run-bench.raku | 10:36 | ||
| native exe is _always_ faster due to actually not including compile time | 10:37 | ||
| (well, almost always, but I was pointing out that the methodology is debatable if we include startup times) | 10:40 | ||
| Anyway, the benchmark I did is a very crude measurement on non-idle system, but it's something | 10:43 | ||
|
11:12
Sgeo left
11:26
leppard left
12:07
vasko45355861 left
12:41
chenggong7788 joined
12:52
chenggong7788 left
12:53
chenggong7788 joined
|
|||
| ugexe | Their benchmarks are on a m4 or m5 and they use arm builds of rakupp | 13:12 | |
|
13:27
kelt0m left,
kelt0m joined
|
|||
| ugexe | I mean they benchmark arm builds of rakupp even if I understand correctly | 13:36 | |
| Voldenet | well it is what it is, arm versions can be compared too | 13:50 | |
| just very niche architecture still | |||
| ugexe | Let me put it another way | 13:58 | |
| You basically have to go out of your way to compare arm64 builds of eg rakupp but x86 versions of rakudo | 13:59 | ||
| Instead of both being arm64 | |||
| Voldenet | oh wow | 14:00 | |
| ugexe | I’ll give them the benefit of the doubt that it’s because their ai hasn’t tried to find the arm64 rakudo builds and not that it’s trying to pad its benchmarks | ||
| Voldenet | maybe the assumption is that jitted version is faster | ||
| so emulated x86 is still faster | 14:01 | ||
| ugexe | I doubt it | ||
| Voldenet | the benchmarks are padded visibly tho - creating scripts that take 200ms to execute and comparing rakudo start time against them | 14:03 | |
|
14:06
chenggong7788 left,
chenggong7788 joined
15:04
abraxxa joined
15:06
kelt0m left,
kelt0m joined
15:17
kelt0m is now known as Guest7446,
Guest7446 left,
kelt0m joined
|
|||
| disbot | <antononcube> In my data science packages CSV parsing is only for the resource data (used in tests and examples.) So, in principle, I do not need a proper CSV parser. I can use ad hoc resource ingestion and processing. | 15:35 | |
|
15:58
abraxxa left
16:24
abraxxa joined
|
|||
| ab5tract | antononcube: that sounds like a plus for Raku in the "it doesn't even pay to modularize the text processing" column :) | 17:37 | |
| disbot | <antononcube> 💯 | 18:30 | |
|
18:51
Sgeo joined
|
|||
| apogee_ntv | Getting very close to Selkie 0.99, which is basically 1.0-rc1 | 20:16 | |
| I don't think zef has rc does it? | |||
| Just need to sweep all of the POD and docs and write a migration guide. | 20:17 | ||
| Still aiming for 1.0 in October, maybe early November now. | 20:18 | ||
| Order is docs -> 0.99 -> perf -> bugfixes (plz help by using) -> api stability gates -> site & supporting materials -> 1.0 | 20:27 | ||
|
20:37
silug9 joined
20:38
andinus left
20:39
notable6 left,
silug left,
silug9 is now known as silug
20:40
andinus joined
20:41
notable6 joined
20:48
abraxxa left
|
|||
| apogee_ntv | Ariza will be my next 1.0 after Selkie. Creating fully contained AppImage-style bundle per platform for Raku apps with recipe syntax. I think improving the distribution story to non-devs is a must. | 21:05 | |
|
23:09
hulk joined,
kylese left
|
|||