[Openchrome-users] XAA and fallback

[Lluís Batlle i Rossell]

On Fri, Apr 06, 2012 at 08:50:18PM +0200, Llu?s Batlle i Rossell wrote:
> On Wed, Apr 04, 2012 at 02:16:36PM -0300, Mario Rugiero wrote:
> > I'm not sure. To talk about the path I think I should know the algorithm
> > used. The thing is that flash uses your CPU only, so all the processing
> > consumes CPU cycles, mplayer, since it uses xv, uses the GPU for some
> > operations, thus using less CPU cycles than flash. The algorithm used can
> > actually be the same, mplayer could be even more resource consuming
> > overall, but it uses less CPU because its process is done elsewhere, that's
> > what I mean.
> 
> I've seen that the destination lines of my test go to the AGP addresses,
> 0xb0000000.
> 
> I've also tried x11perf -shmput500, which may be similar to what flashplayer
> uses. I get a maximum of 53/s; What do you get, in K8M800/K8N800?
> 
> If flash takes 50% of the cpu, may be at the limit for 25fps with the 50% left..
> 
> I've tested a slightly newer notebook, with an Intel controller, and 'x11perf
> -shmput500' gives 1150/s, which is about 20 times more... Shouldn't the
> unichrome chip do more than 53/s?
> 

Ah, Mario, remember that my system shows 50% of usage of flash, 50% of usage of
X. The usage of X is in that backtrace I pasted at the first post. I don't try
to fix the 50% of usage of flash, but that of X.