This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: RFC -- More problems with disappearing labels
- To: Mark Mitchell <mark at codesourcery dot com>
- Subject: Re: RFC -- More problems with disappearing labels
- From: law at redhat dot com
- Date: Mon, 14 May 2001 13:03:26 -0700
- cc: gcc at gcc dot gnu dot org
- Reply-To: law at redhat dot com
In message <20010514113034T.mitchell@codesourcery.com>you write:
> >>>>> "Jeffrey" == Jeffrey A Law <law@cygnus.com> writes:
>
> Jeffrey> If we deferred cleaup until life analysis, then we could
> Jeffrey> actually verify that when we find a dangling REG_LABEL
> Jeffrey> note that the destination of set in the insn with the
> Jeffrey> REG_LABEL note is never used.
>
> I like this one.
>
> If the block containing the REG_LABEL is reachable, and the register
> set isn't dead, then we're in trouble, and we should abort. That
> seems like a good consistency check.
OK. I'll go with this one. It's actually not any harder than the others.
We basically twiddle the code in life analysis to not die when presented
with a deleted label in one place and do a scan over the insns after we've
finished life analysis/dead code elimination searching for dangling
REG_LABEL notes.
jeff