See X.org bug #9614: http://bugs.freedesktop.org/show_bug.cgi?id=9614 The copied Xorg fb code in liboil doesn't correctly detect the presence of MMX and MMX extensions on AMD and AMD Geode processors. Patch in that xorg bug fixes the issue.
I don't think that patch is correct. As I understand it, CPUs that have the vendor string "Geode by NSC" do not have the AMD cpuid extensions, nor do they have mmxext. There are, however, processors labelled "Geode" that do not use the vendor string "Geode by NSC".
(In reply to comment #1) > I don't think that patch is correct. As I understand it, CPUs that have the > vendor string "Geode by NSC" do not have the AMD cpuid extensions, nor do they > have mmxext. There are, however, processors labelled "Geode" that do not use > the vendor string "Geode by NSC". Not correct; the patch actually came from AMD's Jordan Crouse. The Geode GX2 and Geode LX (both used on OLPC XO laptops for example) are reported as "Geode by NSC" yet they certainly do have MMX. For example, on a Geode GX2, /proc/cpuinfo says: vendor_id: Geode by NSC flags: fpu de pse tsc msr cx8 pge cmov mmx mmxext 3dnowext 3dnow This is also definitely the case for the LX.
Liboil on the XO detects "cmov mmx mmxext 3dnow 3dnowext". This appears to be correct according to /proc/cpuinfo.
Tagging patch; will triage later. ...As this isn't an Xorg-related bug, maybe the people pinged by this comment could handle it?
As mentioned in an earlier comment, this bug is unreproducable.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.