This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
RE: reload bug in 3.4.6
- From: "Dave Korn" <dave dot korn at artimi dot com>
- To: "'Momchil Velikov'" <velco at fadata dot bg>, <gcc at gcc dot gnu dot org>
- Date: Fri, 26 May 2006 16:40:22 +0100
- Subject: RE: reload bug in 3.4.6
On 26 May 2006 16:23, Momchil Velikov wrote:
> Now r3 is stored at [r14 + 160], which is incorrect. The reason is
> that when substituting the reload register into insn 2780, the insn
> 22942 is modified also, because they share the following rtx: (plus:SI
> (reg/f:SI 14 r14) (const_int 124 [0x7c])). Since insn 2780 modifies
> r3, r3 is no longer equal to ``r14 + 124'' at insn 22942, which
> results in incorrect store.
>
> I'd appreciate any suggestions for fixing this.
Unsharing the rtx sounds like the right thing to do to me; the internals
docs "Structure sharing assumptions" says things like
"* No RTL object appears in more than one place in the RTL structure
except as described above. Many passes of the compiler rely on
this by assuming that they can modify RTL objects in place without
unwanted side-effects on other insns."
and by the time we're in reload that should definitely be the case for a
complex binary operator.
How did they get to be shared in the first place? That's probably the
underlying cause of the bug.
cheers,
DaveK
--
Can't think of a witty .sigline today....