jump/alignment considerations

Mark Hemment (markhe@sco.COM)
Mon, 27 Jan 1997 10:07:59 +0000 (GMT)


> on the other hand, doing an
> jg over
> jmp __down_failed
> over:

Be careful. The Pentium (and, I guess Pro), branch prediction is
not clean as it should be.
I don't have the references at hand, but I remember reading about
the above jump construct. Something about consecutive jumps sharing
the same BTB entry, and causing mis-applying BTB entries.