Evergreen accel works great on my Radeon HD 5750!

[Dave Witbrodt]

My humble thanks to Alex Deucher, Dave Airlie, and the others working on 
open source Evergreen support for X Windows.

I purchased a Radeon HD 5750 in February, when the first announcement of 
Evergreen KMS support appeared.  My purpose was at least to help with 
testing, and possibly begin browsing the ATI docs and radeon source code 
with the idea of contributing code, or at least code fixes.

Other commitments have prevented me from serious diving into coding so 
far -- my only code contribution was to point out a firmware typo back 
in April -- and even my testing was infrequent because a bug that caused 
a pink line to appear on my HDMI monitor when using KMS (which made me 
paranoid about the possibility of damage) was only fixed in the second 
half of 2010.

Since August, I have been distracted by the prospect of buying the house 
I've been renting for the past decade.  I had been keeping up to date 
with kernels, xf86-video-ati, mesa, xorg-server, etc., up through about 
Aug. 20.  On Oct. 24, I finally set aside a day to play around with all 
of the software changes I've missed over the past 2 months.  A great 
deal of Evergreen work happened during that time.

I used git to build various versions of RC 2.6.36 kernels (esp. 
drm-radeon-testin), mesa libs, and xf86-video-ati drivers between late 
Aug. and late Oct.  I saw segfaults, screen corruption, and refusals of 
the X server to even start.  I was getting depressed, and was expecting 
to have to resort to my older HD 4850 card, as I approached the latest 
available versions of the software.

It was the very last combination,

     kernel 2.6.36 (mainline)
     xf86-video-ati 6.13.99 (git commit fc7aa63 of Oct. 24)
     mesa 7.10-devel (git commit 70f60c9 of Oct. 24), r600c

that finally gave me a perfectly working system -- no segfaults, no 
corruption, no errors or warnings in the log files, etc.  In fact, HD 
5750 performance seems as good or better in every respect compared to 
support for my HD 4850 (which I considered to be perfectly fine).

Believing that it was too good to be true, I decided to stress test the 
system for a week before deciding that HD 5750 support was working this 
perfectly.  I am happy to report that it has been a week, and all of the 
software I use daily, and all of the 3D software I have thrown at it to 
try to make it stumble, has worked without corruption or segfault or any 
other negative consequence.

Nice work, folks!


Thanks again,
Dave Witbrodt