This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug rtl-optimization/27883] [4.0/4.1 regression] in schedule_insns, at sched-rgn.c:3038 on mips
- 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: 31 Oct 2006 14:34:52 -0000
- Subject: [Bug rtl-optimization/27883] [4.0/4.1 regression] in schedule_insns, at sched-rgn.c:3038 on mips
- References: <bug-27883-12387@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #17 from amylaar at gcc dot gnu dot org 2006-10-31 14:34 -------
(In reply to comment #4)
> It is curious that life2 is running immediately before sched, instead of
> immediately after combine. If we ran it immediately after combine, we could
> get rid of the REG_DEAD/REG_UNUSED support in distribute_notes, which is
> probably the most complicated, and most buggy, part of combine. This would
> also speed up the compiler a little bit.
The combiner itself relies on REG_DEAD notes to find out what it can combine.
If you leave an invalid REG_DEAD note, it can lead to wrong code being emitted.
If you don't leave in valid REG_DEAD notes in their proper place, multi-step
instruction combinations will stop working - and with the three insn limit on
any single combination, this is likely to stop quite a lot of combinations.
Most combiner bridge patterns will become useless.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27883