This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
AVX types: __m256d versus __v4sd
- From: Marc Glisse <marc dot glisse at inria dot fr>
- To: gcc-help at gcc dot gnu dot org
- Date: Sun, 18 Mar 2012 20:29:30 +0100 (CET)
- Subject: AVX types: __m256d versus __v4sd
- Reply-to: gcc-help at gcc dot gnu dot org
Hello,
I noticed that the AVX vector of 4 double has 2 names, __v4sd and __m256d
(actually we have the same duality for SSE), and I'd like to know when I
should use which name.
The only difference is the may_alias attribute. I expect that casting a
__v4sd* to a double* doesn't require the may_alias attribute, and that
problems only start if I try casting a __v4sd* to a __v4di*?
On the other hand, Intel only documents __m256d (the may_alias version),
so portability says to use that one.
--
Marc Glisse