This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH for reorg.c: Avoid side_effects_p when "opposite arithmetic".
- To: m dot hayes at elec dot canterbury dot ac dot nz
- Subject: Re: PATCH for reorg.c: Avoid side_effects_p when "opposite arithmetic".
- From: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- Date: Thu, 13 Apr 2000 04:43:39 +0200
- CC: hans-peter dot nilsson at axis dot com, gcc-patches at gcc dot gnu dot org
> Date: Thu, 13 Apr 2000 10:17:32 +1200 (NZST)
> From: Michael Hayes <m.hayes@elec.canterbury.ac.nz>
> Hans-Peter Nilsson writes:
> > For targets with delay-slots that can hold a "plus" or "minus" arithmetic
> > insn, and where an operand can have a side-effect (like a memory operand
> > with a post-increment or a volatile access), a bug is lurking in
> > reorg.c. This test-case might trig it; it does for CRIS:
>
> There are a number of bugs in reorg.c (especially for targets with
> multiple delay slots). Herman Ten Brugge has repeatedly posted
> patches to this list but they have never been approved.
I guess I'm lucky in that delay-slots in CRIS are simple; single
and non-annullable, conditioned in the .md with a straightforward
per-insn-alternative attribute.
> I'm attaching the current diffs to reorg.c that I have in my c4x
> development tree. I'd be interested to see if they fix your problem.
I can't say definitely no, but I did not spot a fix for the bug
at hand. Thanks anyway. BTW, the patch is now installed
following Jeff's approval; the test-case is committed as
gcc.c-torture/execute/20000412-6.c
brgds, H-P