This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Bug in cse_insn() and emit_libcall_block()
- To: Stephane dot Carrez at worldnet dot fr (Stephane Carrez)
- Subject: Re: Bug in cse_insn() and emit_libcall_block()
- From: Joern Rennecke <amylaar at cambridge dot redhat dot com>
- Date: Wed, 28 Feb 2001 02:15:00 +0000 (GMT)
- Cc: gcc at gcc dot gnu dot org
> - In cse_insn(), cse.c:5074, there is a check for a REG_EQUIV note that could
> be present in the insn. As far as I understand, the purpose of this test
> is to detect the above 'set' instruction that copy the stack slot
> to a register. In that case, the 'set' is marked as src_volatile, which
> disables the cse replacement.
No, it is meant to detect the case where an argument that is passed on the
stack is copied into a register.
> I would like to know if the generation of REG_EQUIV note for this kind
> of problem is the good way to go, and if fixing in that direction would be
> acceptable.
No, it isn't. No, it can lead to incorrecft code generation.
Look for REG_EQUIV in rtl.texi to learn what that note means.