Re: Linux in a binary world... a doomsday scenario

[Nix <nix@esperi.org.uk>]

On Tue, 13 Dec 2005, Helge Hafting mused:
> Nix wrote:
> [...]
> 
>>FWIW when glxgears is running, X consumes about 2% more CPU time
>>than normal: it's almost impossible to detect.
>>
> Sure - the load is low - and so is performance.  As if the machine
> isn't really trying - perhaps the driver is waiting when there is
> no need to wait.

If it gets it wrong X deadlocks, but yes, it does seem... odd.

>>It sounds to me almost like direct rendering is disabled, which will of
>>course have catastropic effects on performance. What does glxinfo say?
>>
> No.  While it is bad, it is not as bad as sw rendering. Tuxracer
> with sw rendering is unbearable - 2 seconds per frame or so!

Ah, yes, tuxracer does have fairly high demands, doesn't it.

[delay]

I just installed the latest ppracer alpha (the closest out there to
tuxracer)

OK, I see 25-40fps with that, with DRI. If I turn on stencil buffering
it gets unbearable, but that's just asking too much of the card I think.

>>The X startup log can also be useful here. I've noticed that if you
>>get the AGPMode wrong in either direction, the results can be
>>catastrophic: if it's too low the card is terribly slow and if it's
>>too high you soon get the X server hanging as it waits forever for
>>the card to respond to something it hasn't had time to receive (or
>>something like that, anyway).
>>
> Well, there is no AGPmode for a PCI card, is there?

Ah, this is a pure-PCI 9250, is it? (I wasn't aware you could get hold
of those anymore... I think X supports them, but textured stuff is
necessarily going to be slower.)

-- 
`I must caution that dipping fingers into molten lead
 presents several serious dangers.' --- Jearl Walker
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/