This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix canonicalization of addresses
On Tue, Jan 6, 2009 at 11:19 PM, Eric Botcazou <ebotcazou@adacore.com> wrote:
>> To repeat: it's a canonicalization. And it helps for canonicalizing a
>> * (-2) - 8 to (a + 4)*(-2)
>> and allowing CSE with another a + 4. I don't think that in this case
>> we want to have (-4 - a)*2.
>
> The - - should be canonicalized elsewhere.
How? By extracting the common multiplication by -1? ...
> Changing the sign without further
> analysis is just plain wrong and appears to be an unexpected fallout of the
> change, according to the comment:
>
> /* No identical multiplicands; see if we can find a common
> power-of-two factor in non-power-of-two multiplies. This
> can help in multi-dimensional array access. */
Well, the change was to merge different similar transformations if I remember
correctly. Also the change just introduced a more complete canonicalization.
So the comment is out-of-date.
And yes, we miss a (1 - i) * -2 to (i - 1) * 2 canonicalization, but I
think I said that, no?
Why do you keep arguing? (or rather, why do I?)
Richard.