On 05/13/2013 01:33 PM, Max Filippov wrote: > memory access at points (1) and (2) can abort the instruction (it did so > b/o the bug, but it may do so legitimately when you add MMU support), > but it has modified REG(1) at those points, which will not be restored. > It's probably worth carrying register modifications in some temporary > until after the point (2).
... Which you'll recall that I pointed out during initial review, AG. r~
