|
00:13
MasterDuke joined
|
|||
| MasterDuke | ugexe: there are still some TODOs and XXXs in src/Raku. were those deliberately left out of github.com/rakudo/rakudo/pull/6423 ? | 00:16 | |
| ugexe | are they still relevant? some of them were | 00:55 | |
| MasterDuke | src/Raku/ast/code.rakumod:2524 seems like it might be | 00:58 | |
| ugexe | github.com/rakudo/rakudo/pull/6448/changes | 00:59 | |
| MasterDuke | ah | 01:00 | |
| src/Raku/ast/expressions.rakumod:857 perhaps | 01:01 | ||
| ugexe | irclogs.raku.org/raku-dev/2026-07-12.html#23:29 | 01:02 | |
| MasterDuke | 0 for 2 so far... | 01:03 | |
| src/Raku/ast/package.rakumod:600 | 01:04 | ||
| src/Raku/ast/statements.rakumod:176 | 01:05 | ||
| ugexe | lizmat has been working on POPULATE so i left it alone | ||
| MasterDuke | other than those, i would say (re)implementing more of src/Perl6/Optimizer.nqp | 01:06 | |
| ugexe | gist.github.com/ugexe/e30434b71cf5...6728c83b2d | ||
| MasterDuke | ah, statements.rakumod:176 is already there | 01:08 | |
| ugexe | yeah i actually did do that one | ||
| so that comment is actually stale now | |||
| or at least less stale? github.com/rakudo/rakudo/pull/6417 | 01:09 | ||
| MasterDuke | i saw the loop optimization PR, wasn't sure if that comment was truly stale | ||
| or if there were more optimizations not-yet-implemented | 01:10 | ||
| ugexe | we have the same optimizations as legacy with that PR | 01:11 | |
| MasterDuke | nice | 01:12 | |
| i assume you mean just for loops? | 01:13 | ||
| ugexe | yeah | 01:14 | |
| gist.github.com/ugexe/9b10e419de88...3cfdc4464f i found this report from when i worked on some other optimizations like a month ago. there was significant push back on the viability of porting src/Perl6/Optimizer.nqp to rakuast | 01:22 | ||
| MasterDuke | yeah, i've looked through src/Perl6/Optimizer.nqp a bunch thinking about how to do that stuff in rakuast and didn't have any great ideas | 01:26 | |
| i wonder how many papers on optimizing asts would be useful starting points | 01:29 | ||
| can you feed your system profiles (i.e., --profile *and* --profile-compile output) and ask it to use those as the starting point for implementing optimizations from scratch? | 01:40 | ||
|
01:42
MasterDuke left
|
|||
| ugexe | i mean i dont even have to feed it system profiles, it will do stuff like that on its own. it doesn't really follow the way a human would since it optimizes what it does for token consumption | 01:44 | |
| but yeah you could just feed it a profile and it would probably be able to derive useful stuff from it, just not as useful if you let it run/modify the code generating the stuff | 01:46 | ||
|
01:53
MasterDuke joined
|
|||
| MasterDuke | not sure what you mean. the profiles would be additional input/context to inform the modifying the code | 01:55 | |
| ugexe | Right but it can generate its own profiles while modifying the code as well | 02:01 | |
| MasterDuke | btw, if you're looking for a *huge* task. rebase nqp's `truffle` branch and finish it. would be amazing for the jvm backend | 02:06 | |
|
02:36
MasterDuke left
|
|||
| ugexe | i would be inclined to remove jvm support altogether | 02:45 | |
| it isn't accounted for in rakuast at all, and given none of the other backends materialied close to something production level readys of useful i'm not convinced they are worth the tech debt they put on the code base | 02:46 | ||
| coleman | I agree with removing JVM. | 03:27 | |
|
05:15
hurufu joined
|
|||
| patrickb | I agree to the tech dept assertion. I'd love to have a Inline::Java at some point though. | 06:55 | |
|
09:04
Pixi` joined
09:08
Pixi left
|
|||
| lizmat | ugexe MasterDuke I paused working on POPULATE to not be in the way of ugexe's work | 10:56 | |
| patrickb | Could we somehow turn the huge amount RakuAST commits to get the ecosystem working into a test suite? | 12:22 | |
| lizmat | I think most, if not all RakuAST commits came with tests distilled from the issue in the ecosystem | 13:53 | |
| once RakuAST becomes default, it can be moved to roast | |||
| ugexe | quite a few todos passing in roast under rakuast as well | 14:00 | |
| dev.azure.com/Rakudo/rakudo/_build...amp;l=1808 | |||
| patrickb | lizmat: Oh! I didn't know we were building a bigger test suite in parallel to the RakuAST work. This is very nice! ugexe++! | 15:09 | |
| This means the discrepancy of roast vs ecosystem will get a lot smaller, once those test move to roast. Or am I to naive here? | 15:10 | ||
| (I wonder because I recall a statement of ugexe, that roast is in no good shape to inform a new implementation, i.e. raku++) | 15:11 | ||
| ugexe | yeah you can see a bunch of new test files in t/02-rakudo for instance | 15:13 | |
| or i guess you can also look at github.com/rakudo/rakudo/commits/main/t where there are like 8 pages of my changes | 16:36 | ||
| i guess the next thing i might take a crack at is triaging roast todos / fudges | 18:16 | ||
| [Coke] | want another blinast run? | 18:37 | |
| anyone tackle the moarvm issue you raised? | |||
|
19:28
Pixi` left
19:29
Pixi joined
|
|||
| ugexe | yeah i suppose it would be useful to know where we are at. i was hoping github.com/raku-community-modules/...lob/pull/1 could be merged and released before the next run but ideally someone who actually uses it would check to make sure it works right still. jonathanstowe said they might get to it tomorrow here github.com/rakudo/rakudo/issues/63...4984232906 | 21:12 | |
| significant changes in the roast will actually be a giant pain while both have to pass it still, so i'm not sure how much i'll actually try to do there | 21:14 | ||
| both = both frontends | |||
|
21:50
hurufu left
|
|||