Michel Dänzer <[EMAIL PROTECTED]> writes:
> On Wed, 2005-01-19 at 10:06 +0100, Dominique Dumont wrote:
>>
>> With a Radeon 9200, kdm and a dual session setup, DRIscreenInit fails:
>> (EE) RADEON(0): [dri] DRIScreenInit failed. Disabling DRI.
>
> Is this the case for both sessions? It's a limitation of the DRI that it
> can only be enabled on one session at a time.
No. You're right. DRI is disabled for screen :0 and enabled for screen
:1.
Note that this behavior is somewhat counter-intuitive:
- the first session you get on screen has no DRI and you must switch
seession to get a display with DRI.
- KDE has an option ("reserve" keyword in /etc/kde3/kdm/Xservers) to
dynamically create new sessions. When started, kde session :0 has
DRI. A session created afterwards (with KDE menu) has no DRI. (which
makes more sense, IMO, than Xfree default's behavior)
I agree that this bug is minor but it's a PITA until you figure out
the behavior ;-)
>> As a consequence, a lot of games are unplayable. For instance
>> gravitywars is very slow and the load of my machine goes up to 20.
> That sounds weird; for one, the DRI mostly makes a difference for
> OpenGL, does gravitywars use that?
No. only SDL.
> Also, a load of 20 corresponds to 20 processes using as much CPU as
> they get, which would surprise me with gravitywars under any
> circumstances.
You may be right. I cannot reproduce the load of 20 problem. The best
I could do was a 2.50 load. There must have been another factor that I
have not found yet (including a plain mistake on my side ...). I'll
better check what's going on when my son complains again.
I've verified that gravitywars works fine on screen :1 (with
DRI). gravitywars ship is fast.
On screen :0 , gravitywars is slower than on screen :1 but
playable.
Here are some numbers on these 2 situations:
-----------------------------------------------------------
| screen | DRI | CPU load | %CPU Xfree | %CPU gravitywars |
-----------------------------------------------------------
| :0 | N | 1.7 to 2 | 70% | 30% |
-----------------------------------------------------------
| :1 | Y | 1 to 1.8 | 43% | 53% |
-----------------------------------------------------------
So, missing DRI does make a difference, but not very
dramatic.
Sorry for the false alarm.
Cheers
