REQUEST: Up to 256 channels

[lpn256]

This project has serious potential for the recording industry. In the past, I've used WavPack files to store raw stems because of how many channels it supports, various speaker configs et cetera. It supports up to 256 channels. I think this, at normal mode and maybe some multithreading, could potentially beat WV at "normal" compression levels in the same time as the max compression settings for WavPack on 256-channel, etc files. Note, I've only tried about 16 channels max with WavPack. Also, memory.

Notably, channel decorrelation. Something to the equivalent of mid-side except there is an "average channel" and a million little "difference" channels compressed extra. Not sure if there's an industry term for this.

[slmdev]

The predictor design is designed for stereo data, but no input transformations (as mid/side) are done.
Stereo "decorrelation" is directly applied in the OLS core.

One way to support > 2 channels is to directly use the current design and
split the input into joint-channels and mono-channels.

If you split the channels on the file level, code changes will be minimal,
as you are just encoding "multiple independent files" within a frame.

Feel free to add this to the current code base