Hi all,
I've recently been experimenting with Brotli for compression and decompression inside various kernels, and the results so far have been interesting [1], but one thing I've been contemplating is trying to avoid having to use the FPU, since dropping the need for kernel_fpu_begin/end would be nice in Linux, for example.
Quickly searching, I can't find any obvious indication anyone else has gone down this rabbit hole.
Before I consider diving in further (since this is just for fun), is there some reason that I'm not seeing that a fixed-point implementation absolutely couldn't work here, or anyone who's experimented with this?
Thanks!
- Rich