jpeg-decoder VS rav1e

… but that's a problem surprisingly limited to C or 'find symbols by name'-land. Inside a Rust executable, there is no problem applying multiversion or manual alternatives to relevant methods that are actually in the hotpath, and thus executing them with a recent instruction set while being compatible at the same time. That's infinitely harder if you can't simply compile the function multiple times because it would lead to (undetected, dynamic) linker conflicts en-masse.

Given your qualifications, optimizing the Huffman decoding in jpeg-decoder crate sounds like a good opportunity. 50% of the entire decoding time is spent there; any improvements in it would translate to significant performance gains for decoding JPEG images. The tracking issue is here: https://github.com/image-rs/jpeg-decoder/issues/155

Thanks. Issue filed [here under jpeg-decoder module](https://github.com/image-rs/jpeg-decoder/issues/185)

Sure. You'll see it very often in codec implementations. From rav1e, a fast AV1 encoder mostly written in Rust: https://github.com/xiph/rav1e/tree/master/src/x86

Large portions of the algorithm have been translated into assembly for ARM and x86. Shaving even a couple percent off something like motion compensation search will add up to meaningful gains.

Or the current reference implementation of JPEG: https://github.com/libjpeg-turbo/libjpeg-turbo/tree/main/sim...

Again, if anything AOM would be the one restricting licenses to AV1 (if they chose to) except AOM has stated and also published AV1 in a way to allow license free access to development (which allows people to make forks of the official build like it's open source) and usage. (1)(2) I don't see why they would suddenly change this.

You can speed up the sampling process with --vmaf n_subsample=5, which in my experience works more accurately than either 2 or 4, possibly due to this bug/feature present in multiple encoders. You might also need to manually set the number of threads used for VMAF calculation with --vmaf n_threads=16, but YMMV.

That remember me of https://github.com/xiph/rav1e which is an AV1 encoder

I put in the request for ffmpeg passthrough mastering-display data a few years ago and haven't heard of any support yet.

It's getting there.. Initial steps for FFmepg: https://patchwork.ffmpeg.org/project/ffmpeg/list/?series=8444 rav1e: https://github.com/xiph/rav1e/pull/3000

It's entirely possible to re-use bits of other HW encoders for the first pass (motion estimation, etc).