> On Tue, 2003-01-28 at 14:48, Mike Uhler wrote:
>
> > If the patch assumes that one can look backward by one instruction
> > in the STATIC code to determine if the instruction is in a
> > delay slot, one can not have code that jumps directly to the
> > instruction following another branch, as this would cause the
> > code to assume that it was in the delay slot of the branch.
>
> A while back, when working on a different architecture that also had
> branch delay slots, it took me a while to get my head around the
> branch-in-a-delay-slot case, e.g.
>
>
> 10: b 100
> 20: b 30
> 30: foo
> ...
> 100: bar
>
> where the actual program flow would be
>
> 10
> 20
> 100
> 30
>
> and instruction 100 would be considered to be in the delay slot of 20.
>
> I was *very* happy when I first looked at MIPS to see that this was
> specified as unpredictable, even if it was pretty cool to be able to
> make the CPU execute a single instruction in the middle of nowhere.
> Pointless, but cool. :)
I presume that you're talking about Sparc, where such a construct is
used to execute a single instruction out of a table. This is, in
fact, very, very unpredictable on a MIPS implementation, ranging from
reserved instruction, to branching to one of the two branch targets,
to wandering off into hyperspace. So please do not assume that because
a particular implementation does something that all implementations
do the same thing. In this particular case, I can guarantee you that
you won't like the answer you get.
/gmu
--
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
Michael Uhler, VP, Systems, Architecture, and Software Products
MIPS Technologies, Inc. Email: uhler@mips.com Pager: uhler_p@mips.com
1225 Charleston Road Voice: (650)567-5025 FAX: (650)567-5225
Mountain View, CA 94043 Mobile: (650)868-6870 Admin: (650)567-5085
|