K 10
svn:author
V 10
nwhitehorn
K 8
svn:date
V 27
2009-09-14T03:53:48.631527Z
K 7
svn:log
V 573
Fix some register sizes, and fix the cacheline size determination HID5
for G5s. This helps not mire us in an infinite nest of decrementer
interrupts, which is always good, and gets the machine mounting root over
NFS before dying.

Remaining mysteries:
1. The 64-bit kernel sees the CPU tick almost exactly 3 times faster than
the 32-bit kernel. This makes no sense.
2. There is some memory corruption associated with thread switching. Once
the kernel is multithreaded, I get interesting errors like "corrupt
spinlock" and miscellaneous assertion failures in random places.

END