This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFA:] Fix PR 34773 (4.3 regression)
- From: Eric Botcazou <ebotcazou at adacore dot com>
- To: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 1 Feb 2008 18:17:53 +0100
- Subject: Re: [RFA:] Fix PR 34773 (4.3 regression)
- References: <200801312302.m0VN2gGh014120@ignucius.se.axis.com>
> An insn for cris-elf, that is the last of the open-coded negdf2
> sequence, is transformed at the end of fwprop1 (before
> fwprop_done) to the following. (NB: whether such sequences
> without actual call insns should be wrapped in REG_LIBCALL and
> REG_RETVAL is IMHO doubtful, but that's what
> emit_no_conflict_block does, so here we are)
Yes, libcall sequences historically don't always contain a call.
> gcc/testsuite:
> PR rtl-optimization
> * gcc.target/cris/torture/pr34773.c: New test.
>
> gcc:
> PR rtl-optimization
> * reg-notes.def (EQUAL): Mention significance of combination of
> REG_EQUAL and REG_RETVAL.
> * fwprop.c (try_fwprop_subst): Don't add REG_EQUAL to an
> insn that has a REG_RETVAL.
This is OK, thanks.
:REVIEWMAIL:
--
Eric Botcazou