Pointer acceleration improved again

[Simon Thum]

Hi List,

I finally managed to come up with an improved velocity approximation
algorithm, so the resulting pointer acceleration is even more predictable.

The good news: It runs fine in my tree for a few weeks now, and actually
it is a simplification. It doesn't change anything user-visible (except
for the advanced settings), all high-level concepts stay as they are now.

It performs much better than to old algorithm in terms of precision,
numerical stability and, to some extent, noise tolerance. Also, it has
less tweaks and they relate better to the problem domain.

To sum it up, I would have done it like this in the first place. It's
just that I had no cheap solution for identifying non-linear motion,
which is now handled by a heuristic.

Reviews welcome*.

Cheers,

Simon


 * If you apply it you need the config cleanup I posted earlier.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-dix-simplified-velocity-approximation-algorithm.patch
Type: text/x-patch
Size: 24905 bytes
Desc: not available
Url : http://lists.x.org/archives/xorg-devel/attachments/20090311/13f667ec/attachment-0002.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-xf86-sync-to-dix-changes-in-velocity-approximation.patch
Type: text/x-patch
Size: 3356 bytes
Desc: not available
Url : http://lists.x.org/archives/xorg-devel/attachments/20090311/13f667ec/attachment-0003.bin