This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug c/56083] Vectorizer uses xor/movlps/movhps rather than movups
- From: "ubizjak at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Wed, 23 Jan 2013 16:23:47 +0000
- Subject: [Bug c/56083] Vectorizer uses xor/movlps/movhps rather than movups
- Auto-submitted: auto-generated
- References: <bug-56083-4@http.gcc.gnu.org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56083
--- Comment #3 from Uros Bizjak <ubizjak at gmail dot com> 2013-01-23 16:23:47 UTC ---
(In reply to comment #2)
> Uros, I completely agree, but note that the PR also mentions this line:
>
> xorps %xmm0, %xmm0
>
> which appears rather useless since the low and high parts are overwritten on
> the next 2 lines. Is this done on purpose to kill dependencies or something? Or
> is it a missed optimization? (simplify-rtx (or another place) is indeed missing
> a few things in that area)
IIRC, this insn is emitted to break dependency chain for %xmm0 register.
Processor is not _that_ smart ;)