This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Aliasing fun
- From: Daniel Berlin <dan at dberlin dot org>
- To: Robert Dewar <dewar at gnat dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 25 Jan 2002 12:21:11 -0500 (EST)
- Subject: Re: Aliasing fun
On Fri, 25 Jan 2002, Robert Dewar wrote:
> <<It's not so much they don't understand, as they keep ignoring the
> substantive issue in favor of arguing about undecidability.
> Guess it's just more fun.;
> >>
>
> Actually Dan, the way the mail shows up in my mail reader, it is you who
> have been aggressively arguing that undecidability was a significant issue.
I said, actually, that "Aliasing is hard to reason about formally, because
you run into undecidability issues".
Then richard replied that alias sets were somehow different to reason
about [formally], which they aren't.
I have not, nor ever, aggresively argued that from the non-formal
persrpective, it makes any difference.
It obviously doesn't.
> And now it seems that all you meant was that it's difficult to decide, which
> of course is something else entirely.
No, I mean what I said. It's hard to reason about formally, because in
reasoning abuot it formally, undecidability is a significant issue.
>From a practical perspective, it's difficult to decide, but undecidability
plays no role.
Please don't put words in my mouth.
I qutie clearly said that it's an issue from a formal perspective, and in
formal proofs, because mark said it would be easy if the algorithms are
easy.
You have been trying to argue that in a non-formal reality, like a
compiler, it makes no difference, which while true (and i have not
disagreed), but pointless to that particular issue.
--DAn