Poor compositing performance on 965Q chipset with intel 2.2.1 driver

[Barry Scott]

Keith Packard wrote:
> On Thu, 2008-05-08 at 17:30 +0100, Simon Farnsworth wrote:
>   
>> Simon Farnsworth wrote:
>>     
>>> We're seeing poor compositing performance on 965Q chipset systems, but 
>>> not Q35 or 945 based systems; until we got the Q35, we'd assumed that 
>>> this was caused by the synchronous compositing bugs that Carl Worth 
>>> diagnosed, as the symptom is Xorg taking 100% CPU.
>>>
>>> Oprofile shows that on the 965Q (but not the Q35 or the 945), Xorg 
>>> spends most of its time in malloc.
>>>
>>>       
>> Sorry, brain-fade - memcpy, not malloc.
>>     
>
> An updated libpciaccess may help here - the frame buffer was getting
> mapped UC instead of WC.
>
>   
Using latest libpciaccess makes no difference MTRR status is the same:

# cat /proc/mtrr
reg00: base=0x00000000 (   0MB), size= 256MB: write-back, count=1
reg01: base=0x10000000 ( 256MB), size= 128MB: write-back, count=1
reg02: base=0x18000000 ( 384MB), size=  64MB: write-back, count=1
reg03: base=0x1c000000 ( 448MB), size=  32MB: write-back, count=1
reg04: base=0x1e000000 ( 480MB), size=  16MB: write-back, count=1
reg05: base=0x1f000000 ( 496MB), size=   8MB: write-back, count=1
reg06: base=0xe0000000 (3584MB), size= 256MB: write-combining, count=1

Another suggestions we can check out?

Barry