Re: [Dri-devel] Fw: Re: Savage driver in Mesa and DRI trunk

[Alex D. <ag...@ya...>]

--- Felix Kühling <fx...@gm...> wrote:
> On Fri, 5 Mar 2004 07:34:30 -0800 (PST)
> Alex Deucher <ag...@ya...> wrote:
> 
> > 
> > --- Felix K_hling <fx...@gm...> wrote:
> > > Mark,
> > > 
> > > I'm forwarding your message to the dri-devel mailing list. I
> can't
> > > test
> > > randr here. Alex how much do you know about randr? Can you
> reproduce
> > > and
> > > problems?
> > 
> > Unfortunately, I haven't really messed with randr much.  I don't
> > personally find much use for it.  However I do have the same
> notebook
> > Mark has so I may take a look at it.  Does the dri cvs tree have
> xrandr
> > in it? unfortunately that notebook has a pretty old distro on it.
> 
> I just realized that I do have xrandr installed. I wasn't aware that
> it
> was part of XFree86 4.2.1 which I still have installed on my box. Or
> is
> it a Debian add-on? Either way, I'm going to look into it sometime.
> 

I did some testing with xrandr, and it indeed causes corruption.  It
seems to work with the stock xfree86 savage driver.  I suspect the
problem is that the stride in the GBD and streams registers are set up
differently for the DRI vs. what Tim did and they do not get updated
properly on an xrandr "mode" switch so the drawing engine draws with
the wrong stride.  Tim uses the vga regs for the stride and crtc
offset, while the DRI driver uses the stream regs.  I'm going to have
to dig into xrandr more to figure out how to to alert the driver of the
new stride.  OTOH, I think xrandr uses switchmode(), so the GBD and
stream regs should be updated, however it might not work depending on
which parts of the ScrnInfoRec xrandr changes.  I could be wrong this
is just a guess for now.  I need to do more research and testing.

Alex

> Felix


__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you’re looking for faster
http://search.yahoo.com