This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug rtl-optimization/23837] [4.0/4.1/4.2 regression] Wrong code with REG_NO_CONFLICT notes (caused by combine)
- From: "amylaar at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 16 Dec 2005 14:37:29 -0000
- Subject: [Bug rtl-optimization/23837] [4.0/4.1/4.2 regression] Wrong code with REG_NO_CONFLICT notes (caused by combine)
- References: <bug-23837-5724@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #28 from amylaar at gcc dot gnu dot org 2005-12-16 14:37 -------
(In reply to comment #25)
> Smarter folks than me (iant ;-) suggest that "a multi-word rotate will normally
> need all the input bits when setting any of the output bits", so the entire
> no-conflict thing doesn't make sense here.
This is not really relevant, since the the insns that are using other words
than
the result word are computing their (intermediate) results into pseudos, and
thus get emitted in front of (i.e. outside) the REG_NO_CONFLICT block.
combine is really to blame for combining insns into a REG_NO_CONFLICT block
in a way to invalidate the REG_NO_CONFLICT note without removing the note.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23837