This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] PR/29840, fwprop chokes on PA
- From: Paolo Bonzini <paolo dot bonzini at lu dot unisi dot ch>
- To: Eric Botcazou <ebotcazou at libertysurf dot fr>
- Cc: gcc-patches at gcc dot gnu dot org, Dave Anglin <dave dot anglin at nrc dot ca>, Steven Bosscher <stevenb dot gcc at gmail dot com>
- Date: Fri, 22 Dec 2006 13:22:34 +0100
- Subject: Re: [PATCH] PR/29840, fwprop chokes on PA
- References: <458802F8.email@example.com> <firstname.lastname@example.org>
Eric Botcazou wrote:
In some cases, fwprop would forward propagate call-clobbered hard
registers across function calls. This is fixed easily by passing
DF_HARD_REGS to df_init: then dataflow will create fake definitions for
these call clobbered registers, which is enough to fix the testcase in
the PR audit trail. (I don't know PA assembler, but the bad
propagations disappear from the fwprop1 dump).
What's the relationship between this and...
Another fix, merged from dataflow branch, is to skip artificial defs and
...that? (a comment would be nice on the new DF_REF_ARTIFICIAL guards).
To bootstrap on PA, you need both.