On Fri, 11 Nov 2011 21:18:31 +0100, Daniel Vetter <dan...@ffwll.ch> wrote: > On Fri, Nov 11, 2011 at 20:58, Eric Anholt <e...@anholt.net> wrote: > > Oh, yeah, swap. Good point, sounds like a plan. If you manage to get > > reliable swapping of tiled data on 945g, I'll owe you all the beers. > > If you mean i945G as in desktop variant, patch 6/9 should fix the > swizzle detection on that one - it can do bit17 swizzling, too. I'd be > interested in whether this fixes all the swap issues, if you have that > machine still around ... > > But as I've said there's also the gm45 bug which looks like a part of > the main memory in not swizzled. On that topic: Do you still remember > details about that machine where parts of the memory can be > unswizzled?
The management engine/mismatched dimms issue making "L-shaped" memory (part dual channel, part single channel) in the 945-gm45 era was never handled (somewhere in g45-ilk era, there was a reg the BIOS could set that would make L-shaped memory have consistent swizzling anyway). I never found precise docs on how to figure out what the shape of the L was -- I suspect you'd get to find out by poking at things like TOLUD and experimenting with binding all the pages you can grab to find which are swizzled.
pgpwmgMDTEhpp.pgp
Description: PGP signature
_______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
