00:26 rmv left 00:45 johnjay left 02:23 rmv joined, rmv left, rmv joined 02:28 rmv left 03:15 apogee_ntv left, apogee_ntv joined 03:47 lichtkind__ joined 03:50 lichtkind_ left
disbot <frostcod> interesting github.com/ash/rakupp 04:16
04:18 johnjay joined
ugexe the unfortunate thing is the roast is hardly a measure of what to implement 04:22
compared to the ecosystem its almost insignificant even
although they could download the ecosystem and be able to infer a lot without needing to run it all 04:28
04:50 kylese joined 05:09 wayland76 left 05:16 wayland76 joined
Voldenet This is actually nice, hot code in theory could be optimized into cpp libs with no rewrites 05:25
in practice, I've tried and almost none of my code works except trivial examples 05:35
05:39 hurufu joined
Voldenet but when it does work it's certainly fast 05:40
AH, it's actually slower – compile step is blazingly fast so it appears faster 05:57
06:20 rmv joined, rmv left, rmv joined 06:22 hwj joined 06:23 hwj left, hwj joined, hwj left 06:24 hwj joined 06:25 rmv left 06:39 wayland76 left, wayland76 joined 07:09 Sgeo left
wayland76 Sorry to hear that. But maybe someday it'll pick up speed. 07:17
07:52 dakkar joined 08:01 rnddim joined, ShimmerFairy left 08:02 rnddim left, rnddim joined, rnddim is now known as ShimmerFairy 08:04 ShimmerFairy left, ShimmerFairy joined
Voldenet well, raku needs plenty of runtime optimizations, otherwise code will stay very dynamic 08:16
e.g. rakupp thinks treats <123> as Int, but <123 abc> gets two Str 08:18
s/thinks//
<123 456> is two strings too
though markdowns say what in rakupp is faster, but it requires specific code style – why would I bother writing it in raku if I wanted to use C syntax 08:22
and of course, it's only faster on m1 chip which is kinda rare 08:27
09:21 rmv joined, rmv left, rmv joined 09:26 rmv left
hwj Voldenet, do you mean m1 specifically, or Apple silicon in general? 11:22
Voldenet What I meant was that moarvm raku has no jit on apple silicon, so it's rakupp can be faster 12:42
s/it's//
and on intel-x64 I can't reproduce piece of code actually being faster 12:43
13:07 wayland76 left 13:08 wayland joined 13:25 rmv joined 13:30 rmv left 14:07 hwj left
timo having more than one compiler again would be good for rakudo as well. i haven't looked at more than one or two of the .md files in that repo yet 15:13
will be interesting to see how raku++ handles serialization of the results of "constant" and BEGIN, and especially anything where repossession has to happen 15:15
El_Che lo 15:23
tbrowder: I'll try to spin a container
tbrowder: debian 12?
15:27 rmv joined 15:31 rmv left
El_Che curious about the raku++ 15:38
specifically native binaries for compiled code
16:54 dakkar left 17:22 clemtre joined
Voldenet compiling raku code to binaries directly is very ambitious goal 17:24
ugexe doing so with rather naive raku code isn't too difficult 17:26
./rakuapp-full -e 'say [+] 1..100'
5050
like i have a branch that can do ^. but beyond naive raku code it gets much harder very quick 17:27
17:29 rmv joined, rmv left, rmv joined 17:33 rmv left
leont Voldenet: well, rakudo's compiletime is its weak spot, so it being faster there makes sense. 17:37
To be honest, I suspect it's much more interesting to write a re-implementation of nqp. 17:38
17:58 clemtre left 18:39 hurufu left 18:40 hurufu joined
Voldenet I'm not going to lie, if things I normally use worked ootb in rakupp, I'd use it 19:03
tbrowder El_Che: please try debian 13 (trixie) first if you can 19:32
20:00 hwj joined 20:28 hurufu left 20:32 rmv joined, rmv left, rmv joined 20:43 avuserow left 21:05 ShimmerFairy left 21:06 ShimmerFairy joined 21:10 hwj left 21:32 avuserow joined 21:34 wayland left, destroycomputers left, gcd left, bd3i left, acidsys left, leah2 left, abbe left, eseyman left, kelt0m left, atcroft left, committable6 left, bloatable6 left, peder left, pierrot left, jjatria left, committable6 joined, bloatable6 joined, abbe joined, eseyman joined, gcd joined, peder joined, jjatria joined, pierrot joined, bd3i joined, kelt0m joined, acidsys joined, leah2 joined 21:35 destroycomputers joined 21:38 wayland joined
[Coke] You could "just" add another backend to NQP 21:43
[Coke] stares at his very old todo list item about .net
ugexe at this point i'm in favor or just getting rid of the other backends 21:58
rakuast has ignored them entirely
22:55 Sgeo joined 23:11 lichtkind__ left