Michael Bebenita
Michael Bebenita
We have SIMD code for 8-bit block copies but not for 16-bit. See related patch that added 16-bit block copies for full precision references. https://review.xiph.org/1049/
When working on #133 I noticed that sometimes strides are not a multiple of the cache line. This feels wrong, since it would lead us to read across cache line...
See comments in: https://review.xiph.org/986/ @jmvalin mentioned this is the path we should take, @tterribe do you see any issues with this?
https://review.xiph.org/794/ introduces a `~2x` slowdown, assembly code should be written to speed this up.
Our PVQ code is difficult to understand and can benefit from some refactoring work.
The intro README for this project doesn't say much. It would be great if we highlighted some of the performance and flexibility features of this parser, and why people should...
It would be nice to be able to zoom in in certain parts of the image and compare the quality of compressed images at different bit rates: 