CPU usage xorg 7.2 vs current

[Andy Furniss]

Roland Scheidegger wrote:

> Since the only real difference seems to be that
> shadowUpdatePacked uses way more cpu, I've got some doubts. Maybe MTRR
> setting failed or something like that?

Ahh, thanks there does seem to be a difference. 7.2 /proc/mtrr is

reg00: base=0x00000000 (   0MB), size=1024MB: write-back, count=1
reg01: base=0xc0000000 (3072MB), size= 256MB: write-combining, count=2
reg02: base=0xd0000000 (3328MB), size=  16MB: write-combining, count=1

while git does not have the last line.

I was running the same kernel for the tests, though 7.2 would have been 
built with an older kernel - one that has a matching /usr/include/linux.

This is LFS and it explicitly says not to link /usr/include/linux to 
anything, saying those should be the headers that glibc was built 
against, not running kernel. Maybe this is why?

Looking at the autogen.sh output I see -

checking asm/mtrr.h usability... yes
checking asm/mtrr.h presence... yes
checking for asm/mtrr.h... yes
checking machine/mtrr.h usability... no
checking machine/mtrr.h presence... no
checking for machine/mtrr.h... no

I had a look at configure --help, but couldn't spot any way to specify 
where my running kernel headers are - not that I know that this would 
fix it anyway, just guessing.

Any ideas how to get mtrr working?