wgpu-mm VS godot-proposals

This is very exciting! (I had suspected it would slip to 114)

WebGPU implementations are still pretty immature, but certainly enough to get started with. I've been implementing a Rust + WebGPU ML runtime for the past few months and have enjoyed writing WGSL.

I recently got a 250M parameter LLM running in the browser without much optimisation and it performs pretty well! (https://twitter.com/fleetwood___/status/1638469392794091520)

That said, matmuls are still pretty handicapped in the browser (especially considering the bounds checking enforced in the browser). From my benchmarking I've struggled to hit 50% of theoretical FLOPS, which is cut down to 30% when the bounds checking comes in. (Benchmarks here: https://github.com/FL33TW00D/wgpu-mm)

I look forward to accessing shader cores as they mentioned in the post.

Godot was fine, for the most part, it's a good engine for small games. The main issue was around working together and resolving conflicts with git (Godot isn't really working and constantly kept changing IDs in resource files), and it kept losing data in scenes when the code changed so we had to rebuild things several times. On previous test games we also ran into huge performance issues in the editor when we wanted to do too much work in scenes, so we had to switch to doing most of the work in code.

The worst part was when I had to rip out a bunch of testing code that used @tool for release as it seemed to cause issues when doing a release build. The engine is generally not great when you want to follow clean software engineering (e.g. dependency injection is not possible, I opened a proposal to that end here[1]).

[1]: https://github.com/godotengine/godot-proposals/issues/8850

If we had a Mask2D, this would be super easy and straight forward.

So if that's what OP meant, clip_children property won't work. Instead OP needs to use a CanvasGroup parent and the Line2D (with CanvasItem Material in Subtract mode) and TextureRect as siblings, or any of the other methods explained here until we have a Mask2D node.

There is an open proposal to have a compile option for a LibGodot, as a Library https://github.com/godotengine/godot-proposals/issues/6267

Yes, there could be the option to open at least 2 tabs at the same time, when working with scripts. This has probably already been suggested in: https://github.com/godotengine/godot-proposals

I cannot tell exactly where the jitter is from the video, but if it is what I'm thinking of, it is a known issue, see here. There is a proposal to fix this, with tons of discussion here. Try some of the workarounds mentioned in these two links. Game looks great by the way!

It's in the release notes. Here is a link to the pull request, based on this proposal.

Currently not possible: https://github.com/godotengine/godot-proposals/issues/7213