On Thu, 2002-03-07 at 16:43, Leif Delgass wrote: > On Thu, 7 Mar 2002, José Fonseca wrote: > > > Ok. I'm gonna look into this tomorrow. Both the r128 and radeon drivers > > have PCI support, so it shouldn'be difficult to bring the PCI support to > > the same state as the AGP (i.e., just initialize it). > > The quickest fix to make things easier for testers would probably be to > just disable the AGP/PCI init in the DRM for now, since it's not really > necessary yet. If we get PCI support added or start working on DMA, then > we could enable it again to test it out. >
So far I added an is_pci var (as is done is r128/radeon) and added code in the DDX driver to fill in that var, but I seemed to broke some shared structure as the X server is behaving rather badly. I hope to get to this when I get home. > > BTW, I just commited initial support for the tex coord premultiplication > > in the mach64-0-0-3-branch. This doesn't help with the lockups in multiarb > > though. > > I checked it out and it looks like the texcoords would get multiplied by > the rhw _and_ multiplied by w. Also, the utah-glx code and the previous > branch multiplied the w coord by the fourth texture coord, if there is > one. I haven't tested this yet, but I'll inline my diff here since it's You're probably right. Although I didn't mentioned in the reply, I pointed out in the CVS log that PTEX format isn't yet covered by the tex coords permultiplication changes. The problem isn't really in the _emit but rather on _interp where the normal case is rather complex to begin with, so I decided to take a little more of time to understand the meaning of the PTEX format and do it properly. > small. I'm going to try to find out where multiarb is choking, since it > shouldn't lock the card even if the texcoords are wrong. > Yes. I also plan to step instruction-by-instruction to see where it really fails, and also check if the multitexturing state is being setup properly, since I have the feeling this may be the problem. > > -- > Leif Delgass > http://www.retinalburn.net > > I've been a little busy with work lately since it had fallen behind when I dedicated myself to bring up the mach64-0-0-3-branch last week. Regards, Jose Fonseca _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
