This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug middle-end/43251] [4.4 Regression] Erroneous code with -ftree-vectorize
- From: "rguenth at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 19 Mar 2010 09:56:33 -0000
- Subject: [Bug middle-end/43251] [4.4 Regression] Erroneous code with -ftree-vectorize
- References: <bug-43251-18868@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #14 from rguenth at gcc dot gnu dot org 2010-03-19 09:56 -------
(In reply to comment #13)
> Hi!
>
> I was wondering if this bug is likely to be fixed in the next GCC release; is
> this likely to be the case?
Unlikely. Backporting the one-line change will cause massive performance
problems elsewhere (with scalar times complex multiplication). There is
more to be backported.
The question is if the Fortran standard allows C*CX(IX) to be unconditionally
implemented as C * __real CX(IX) + _I * C * __imag CX(IX) or not.
It would be also interesting to know what vectorization makes for a difference
here, as it shouldn't change semantics. Thus, did anyone analyze why
only the vectorized variant exhibits the problem?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43251