Kudos and thanks to Michel Dänzer for one or both of these:
36c19067 Fix vline range calculations.
bb0e1531 Check for empty vline ranges after clamping.
At first, I thought I was imagining things -- everything seemed
faster after building a new xf86-video-ati. My previous build had been
13b3aed4 EXA/r6xx+: fix rop setting for overlapping copies
which is the commit just before the latest ones, so my new build only
involved the two commits mentioned above. (I also left the other parts
of the X stack unchanged: libdrm, xorg-server, and mesa. My kernel is
3.2.4 + many cherry-picks from drm-fixes and drm-next, but when I built
it today there was no immediate effect; the speed-up occurred only after
the rebuild of the radeon driver.)
The 'torcs' game frame rate always varies over a wide range while
playing -- normally 20 to 45 fps on my HD 5750 card, with occasional
spikes into the 50' -- but now mostly stays well above 40 fps, usually
in the upper 50's. Since my monitor's refresh rate is only 60 Hz, the
performance is now close to a level such that further improvements would
make no visible difference (for me).
The 'prboom-plus' game has no fps indicator, but anyone could perceive
the difference between the old and new radeon driver builds. This is
especially true if I maximize the view (hiding the dashboard) and play
at fullscreen 1920x1200. It's like I bought a new video card! This
particular program is now as responsive as it was with my NVidia GeForce
7950 using the proprietary driver. (That card died in the summer of
2009, and I have been using only ATI/AMD cards since... and those only
with the open source radeon driver.)
Thanks again,
Dave W.
_______________________________________________
xorg-driver-ati mailing list
[email protected]
http://lists.x.org/mailman/listinfo/xorg-driver-ati
