Replies: 1 comment 1 reply
|
Thank you for this crate. I put out 3 PRs. I am extending Asterinas and needed this to work in the kernel. Please note I used AI (Opus 5.5) for all of this. |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi @myers! We came across your fork and the bare-metal work on the
zarn-*branches: boxing the FSE sequence tables so a decode fits a 32 KiB stack, staging Raw/RLE block reads through the heap instead of a 128 KiB stack array, and gating the x86 SIMD paths and the encoder for targets withoutstd.That is exactly the direction we want the crate to go (no_std + alloc, small stacks, embedded and kernel users), and we'd love to have it upstream if you're up for it. A few notes in case they help:
mainnow has magicless decoding (FrameDecoder::set_magicless, plus the encoder side), so that part of your branch may no longer be needed. Happy to hear if it misses something your use case relies on.mainhas moved a fair bit since your base, mostly in the sequence decoder. If rebasing is a chore, tell us which changes matter most to you and we can help carry them over.no_stdgating. A test that pins the stack budget you need would be great, so it doesn't regress later.It would also be good to hear about your setup: what target and environment you run the decoder in, and which constraints bit you first. That helps us keep it working for you. No pressure at all, and thanks for building on it!
All reactions