| 15 Aug 2020 |
 yorickpeterse | https://gitlab.com/inko-lang/inko/-/blob/master/vm/src/scheduler/process_worker.rs is the meat of the work-stealing logic | 21:50:20 |
| yorickpeterse | Threads used for running processes have two operating modes: normal, and exclusive. In normal mode they can run any process. In exclusive mode they only run the process that pinned itself to the thread | 21:50:56 |
| yorickpeterse | Happy to share more if you need anything :) | 21:53:05 |
| 16 Aug 2020 |
 hernytan | In reply to @yorickpeterse:matrix.org
Inko uses a similar approach. Currently we only reduce for method calls, but this probably will change at some point hmm, i can think of that being a potential issue in long for loops, which was partially why I wanted to look at Inko's implementation. But i'm still very interested in taking a look; though it will take some time to parse all that code :) | 15:26:20 |
| hernytan | thanks for the help! I will look at that code then | 15:26:56 |
| hernytan | Erlang's trick is to disallow loops, so that reductions are easier to count using only method calls. But i don't think this is a good approach | 15:28:45 |
| yorickpeterse | hernytan: Loops are implemented using methods, and method calls in loops would also be pretty frequent. With that said, when we start inlining code we'll probably need to add more bits and pieces | 21:34:15 |
| yorickpeterse | e.g. we can just add a "reduce" instruction, then have the compiler emit that when it optimised methods away | 21:34:41 |
| 18 Aug 2020 |
|  cheer changed their profile picture. | 00:37:30 |
| 1 Sep 2020 |
 kiti_nomad | 几乎所有的开源项目的每月捐款名单里都没有中国人 | 16:03:01 |
| kiti_nomad | Almost all open source projects have no Chinese in their monthly donations😂 | 16:09:32 |
| yorickpeterse | kiti_nomad: most open source projects sadly don't have any donations to begin with :) | 21:39:33 |
| yorickpeterse | Unrelated to that, Inko's self-hosting type-checker/compiler can now type-check (generic) method calls. It's been taking a while, but I'm quite happy with how it's turning out so far | 21:40:20 |
| yorickpeterse | definitely a lot of cleaning up to do, and coming up with better names for things here and there | 21:40:31 |
| yorickpeterse | Overall I am quite satisfied with the code thus far; definitely much better than the mess that is the Ruby compiler | 21:41:00 |
| yorickpeterse | Which means one of two things: 1) I did a decent job 2) I messed up/overlooked something big :D | 21:41:17 |
| 6 Sep 2020 |
| kiti_nomad | I guess it’s not the reason, because you did not provide a payment method that is convenient for Chinese people to donate | 15:51:30 |
| yorickpeterse | kiti_nomad: I'm not aware of the requirements/etc, but there's an Open Collective and GitHub Sponsors; both access conventional credit cards | 19:07:30 |
| yorickpeterse | errr s/acces/allow use of | 19:08:09 |
| 9 Sep 2020 |
| yorickpeterse | I believe I finally managed to handle all AST nodes in the type-checker | 20:20:48 |
| yorickpeterse | The error handling AST nodes were a bit of a challenge, but I'm quite happy with the setup | 20:21:02 |
| yorickpeterse | It's a bit stricter than the Ruby compiler though. For example, you can do this when using the Ruby compiler: try foo.bar.baz. If foo or foo.bar throws, the Ruby compiler will allow this | 20:21:53 |
| yorickpeterse | In the self-hosting compiler you'd have to write try { try { foo }.bar }.baz. This looks ugly, but fortunately it's very rare (e.g. no code in Inko's standard library does this that I remember) | 20:22:35 |
| yorickpeterse | Similarly, this is not valid either: try foo(bar) where bar would raise | 20:23:08 |
| yorickpeterse | In the self-hosting compiler, you must try try foo(try bar) instead | 20:24:53 |
| yorickpeterse | This additional strictness makes it easier for the compiler to verify that all throwing code is handled properly, something the Ruby compiler is not very good at | 20:25:20 |
| yorickpeterse | Ha spoke too soon: I forgot about raw VM instructions that need to be type-checked properly | 21:15:52 |
| 15 Sep 2020 |
|  kirkkaf joined the room. | 10:14:40 |
| 16 Sep 2020 |
| yorickpeterse | Got all VM instructions supported. Two things left to do I believe (for the type-checker):
- Closure argument type inference
- Finding/fixing any bugs by running the type-checker on the standard library
| 20:05:35 |
| yorickpeterse | Inferring closure arguments will be a challenge. In the Ruby compiler this is implemented super hacky, but I don't want to repeat that | 20:06:09 |
