More info Re: dual head radeon -- again

[Mike A. Harris]

On Wed, 17 Mar 2004 hansel at hansel.mnstate.edu wrote:

>Experimental support for radeon, indeed! Nevertheless I have some progress
>and predictable error messages. The config file is attached in its
>entirety. Some failed experiments are commented out (visible warts).
>
>I think the most significant error, if accurate is the failure to setup 
>the address space on the card.
>
>-- X server captured output
>  --snip--
>  (WW) RADEON(0): Failed to set up write-combining range (0xf0000000,0x2000000)
>  (WW) RADEON(1): Failed to set up write-combining range (0xf2000000,0x2000000)
>  --snip--

No, the above are Warning messages, indicating that X was unable 
to set up MTRRs on your CPU.  This means either:

- Your kernel does not have MTRR support compiled in or available

or

- The kernel MTRR support does not support your specific hardware

or

- The kernel has blacklisted your hardware due to known bugs in 
  the MTRR implementation (ie: some serverworks chipsets like OSB4)

or

- All MTRRs are already set, or a conflicting MTRR is set.  
  /proc/mtrr lists the MTRRs

or

- The MTRR code mishandles a particular MTRR implementation.


Note that this is /only/ a performance thing.  When X is able to
use MTRR support, there is a performance gain anywhere up to 2.5
times.  Lack of working MTRRs only reduces performance, it does
not at all cause lockups/hangs/crashes or otherwise affect
stability, and does not cause any other type of problems.  You
can think of MTRR support as being "turbo video".



-- 
Mike A. Harris     ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat