mogwai VS lineiform

Thanks for taking the time to read this! Please let me know here or at the repo (or on any of the other channels you may find me) if you have any concerns, bugs, feature requests, etc.

I’ve been working on the 0.6 release of mogwai, which is a library for building GUI views. This release is all about making the library domain agnostic, meaning it can be used to define views in any GUI domain (ie browser, native, gtk, wgpu game). I’m also trying to trim down the API and improve the RSX macro.

Here's a link to Mogwai: https://github.com/schell/mogwai - its Github readme also has other benchmark numbers (also from TodoMVC but with newer versions of Mogwai) that shows it faster than Preact and Mithril.

I solved the SSR problem in my library mogwai using conditional compilation to include an encoding of the DOM element when compiled for non-wasm targets, which you can see here). The extra field ServerNode allows me to provide an html_string function on non-wasm targets. On wasm targets the html_string function is backed by the DOM element's outerHtml method.

Sounds similar to lineiform. Which isn't all that stable or actively developed, but it is a cute approach to writing a meta-jit in rust. It's a weird approach, but IMO it's worth more experimentation.

Working on Lineiform, my meta-JIT library, some more.

In response to Cranelift, switching to my own Tangle IR won't be using Cranelift at all (it uses raw dynasm-rs for emitting instructions). I go into a bit in https://github.com/chc4/lineiform/issues/19, but Cranelift specifically has some rules about iflags, the type they use to conceptualize processor flags effects (e.g. add's carryout or overflow). You can only have one iflags value live at a time, and it can't overlap with any other math operation. This is a problem because the x86 we're lifting doesn't always follow that rule, so if we just emit Cranelift as we go it will panic and say we built an invalid function.

I go into a bit in https://github.com/chc4/lineiform/issues/19, but it's less a problem with its optimizer and more a problem with its IR constraints. Cranelift specifically has some rules about `iflags`, the type they use to conceptualize processor flags effects (e.g. add's carryout or overflow). You can only have one `iflags` value live at a time, and it can't overlap with any other math operation. This is a problem because the x86 we're lifting doesn't always follow that rule, so if we just emit Cranelift as we go it will panic and say we built an invalid function.

The iflags design in general is kinda awkward too, and was being rethought a few months ago when I was first getting this working; I think they're planning on redesigning the add carryout interface and things to be slightly more streamlined. I suspect that any redesigned interface will have similar problems with mismatch between what I want from Cranelift and what 90% of other uses of Cranelfit want, though, and so I decided to just make my own IR instead.

Working on the codegen backend for Lineiform again. I sketched out a plan on how to implement register allocation in a way that hopefully doesn't have horrible behavior in the majority of cases, and implemented ~half of it last week, and hopefully I'll implement the other half and instruction scheduling this week.

Yup, https://github.com/chc4/lineiform. It's not usable at all yet - I was building it on top of Cranelift, which turned out to be a fairly bad idea, so I'm going to have to essentially rewrite all of it with my own codegen backend I think. I've been hacking on it on and off but it's been much slower progress due to work (and writing a codegen backend is hard...)

I got struck by either a very dumb or very good idea a few days ago, and finally have a working (minimal) proof-of-concept for it: Lineiform is a meta-JIT library to nearly automatically get an optimizing method JIT from a Rust interpreter. It does dynamic recompilation on closures by lifting from x86 to Cranelift IR for runtime function inlining and constant propagation.