Re: set of patches from the last 3 weeks

Eric Anholt wrote:

>
> Hopefully Monday I'll have real internet access here through the 
> school.  I just hope that the Great Firewall that's rumored to exist 
> won't stop ssh.  Once I get that, I'll commit my trident driver.  I 
> could get you CVS access to work on Trident Blade stuff yourself, too, 
> at that point.
>
Good news.

> I'll also commit rendercheck to xapps then, though it still suffers 
> from issues.  It's good enough to point out problems in the Composite 
> implementations I've played with (fb, R128 Blend, Trident 9525 Blend 
> in the new and old trident drivers).

Excellent news.

>
> My main project recently has been overhauling the ATI driver.  I've 
> added R128 xvideo support through the front-end scaler (R100 would be 
> easy to do, too) and hardware cursor support.  I also made it always 
> construct DMA packets and then submit them either through MMIO, 
> psuedo-DMA (which would allow Radeon Composite without the DRM), DRM 
> DMA, and hopefully soon DMA using AGP or ringbuffer in framebuffer 
> without the DRM.  I ought to check what the penalty is for using 
> pseudo-DMA instead of direct MMIO, but I feel the cleanup in the 
> driver is worth it anyway.

Sounds great.
A lot of paths to explore !
I have got a r200 ql but I didn't give a try on this (huge) driver 
(yet)! I'll try to follow it.

I successfully merge the cyberblade code into the xserver-trident-0.5 by 
replacing/adapting blade_draw.c and some minor changes in draw.(c/h) 
,reg.h and uncomment cyberblade/i1 in stub.c (full tar.gz join).
I have only done few tests against epia (cyberblade/i1).
I also plan to add xvideo support to trident driver and investigate some 
other way to accel with 3d funcs (hazardous area).