> > movsl", on the assumption that the interlock problem is between the
> > shift/movsl instruction (but that's unlikely: that particular
> > combination shows up even in normal code). That would be:
>
> Actually Linus take a look at the pentium errata for some steppings. Im
> sure one of the early 100MHz parts had a bug for something like rep movs*
> instructions taking a path that was out of spec under a few conditions.
> I'll look it up tonight.
>
> So maybe the cyrix has a similar oops ;)
>
> Alan
It sounds hard for the victims, but it seems to be wise to get a new
CPU. The clib also uses assembler, and gcc creates such patterns,
too. BTW does anybody know of a similar problem in an AMD 486 DX/4
100MHz?
Ulrich