Patch to mouse accel code (integer version)

Jan Brunner brunner at intergga.ch
Sun Nov 14 16:46:10 PST 2004


>> That said, code which can easily avoid FP is always nicer.
> 
> 
> I'll think about it tomorrow. I estimate that the code is going to get 
> messy without FP but maybe I find something nice.

I implemented the patch without floats and I think it's not messy at all 
and probably quite a bit faster even on systems with FPUs. When first 
thinking about an integer version without writing the code, I didn't see 
that there's still going to be only one division per axis because the 
original division gets "delayed".

The Bugzilla entry:
http://freedesktop.org/bugzilla/show_bug.cgi?id=1688

The patch:
http://freedesktop.org/bugzilla/attachment.cgi?id=1308

I don't know whether the 2 new variables are named in a good way. Feel 
free to change.

PS: Thanks Jim Gettis for making me research a bit about why casts from 
float to int are so expensive. (It's because the CPUs are normally set 
to round to the nearest integer but the C standard (and the old code) 
requires rounding down. Changing the mode and resetting it requires a 
lot of cycles.)

-- 
Best Regards,
Jan Brunner <brunner at intergga.ch>



More information about the xorg mailing list