Re: Off by one error in drawing code for Radeon VE

On Mon, 2002-11-25 at 02:57, Kevin E Martin wrote: 
> On Mon, Nov 25, 2002 at 12:02:32AM +0100, Michel Dänzer wrote:
> > On Son, 2002-11-24 at 22:53, Kevin E Martin wrote:
> > > On Sun, Nov 24, 2002 at 09:57:36PM +0100, Michel Dänzer wrote:
> > > > On Don, 2002-11-21 at 06:44, Andrew P. Lentvorski wrote: 
> > > > > Okay, you wanted isolation; you got isolation. ;)
> > > > > 
> > > > > The option which makes the problem go away is:
> > > > > 
> > > > >         Option  "XaaNoDashedTwoPointLine"
> > > > 
> > > > While this has been much improved in current CVS, Mark's linetest
> > > > program still produces some artifacts. Kevin, have you gotten around to
> > > > look into that yet?
> > > 
> > > Yes.  I rebuilt from the latest CVS tree, and ran xtest and Mark's
> > > linetest program.  It all worked fine for me -- i.e., no artifacts.
> > > 
> > > The latest I've heard is that the person who had the problems and
> > > rebuilt from CVS but didn't install the new XAA.  They were going to try
> > > installing the new XAA to see if that helps (or they could just build a
> > > non-loadable server for testing).
> > > 
> > > However, if Mark's linetest program produces artifacts for you, please
> > > let me know the exact coordinates, the dash pattern used, and if end
> > > caps are on/off, since I have not been able to reproduce this problem.
> > > This info should be trivial to gather in the debugger.  It might simply
> > > be that there are specific dash patterns that cause problems that I have
> > > not tested.  Also, please let me know if this happens on big or little
> > > endian machines, or both.
> > 
> > I told you two months ago that I get artifacts both on my TiBook with a
> > 7500 Mobility as well as on an Athlon box with a 7500. You said you
> > thought you knew what the problem is.
> 
> Yes, I remember that e-mail conversation, but I believe the problem you
> pointed out two months ago is a different problem and is specific to
> Radeon 7500 (RV200) and later cards.  FYI, I think the solution to the
> drawing errors on the 7500 and later cards is to properly set the
> bres_cntl field in the dst_line_patcount register.

Does that mean we could implement full-blown Bresenham lines with the
newer chips?


> However, the problem from the original e-mail in this thread was said to
> be with a Radeon VE (see the subject line), which is older than the 7500
> and does not appear to have the bres_cntl field.  As I mentioned above,
> I've rebuilt from the latest CVS tree, and ran xtest and Mark's linetest
> program.  It all worked fine on my VE card.  At this point, I'm not sure
> what the problem with the VE is.

I see, sorry for the confusion.


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast