Walt Brainerd <[log in to unmask]>wrote:
> [code reordering transformations]
>Keith says, however, that you can usually tell that such a
>transformation
>has been performed. It seems to me that the only "legal" test should be
>performed by looking at the output of the program, not by looking at
>code,
>for example.
I agree. I think that any program which has the same semantic meaning
as the original should be acceptable (and the only difference the user
should be interested in should be the efficiency - both speed and
space - that the implementor's choice manages to attain).
A point was raised earlier about the order, number, and type of IEEE
exceptions generated by the optimized code vs. the unoptimized version.
This would require some interpretation of the part of the committee.
Is a program which generates an exception in violation of the standard
(and the implementation can do anything it likes) or are IEEE exceptions
part of the semantics of the standard? I don't know.
--
J. Giles
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|