This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Conditionally turn off cfg cleanup for PHI-OPT if nothing is done
- From: Andrew Pinski <pinskia at physics dot uc dot edu>
- To: devang dot patel at gmail dot com (Devang Patel)
- Cc: dnovillo at redhat dot com (Diego Novillo), gcc-patches at gcc dot gnu dot org
- Date: Fri, 3 Mar 2006 14:30:53 -0500 (EST)
- Subject: Re: [PATCH] Conditionally turn off cfg cleanup for PHI-OPT if nothing is done
>
> On 3/3/06, Diego Novillo <dnovillo@redhat.com> wrote:
> > On 03/03/06 14:17, Devang Patel wrote:
> >
> > > If cleanup_cfg() is smart and if it does not start walking blocks
> > > unless cfg was manipulated then there is no need to worry about
> > > unnecessary TODO_cleanup_cfg flags. They cost more or less as much as
> > > empty call at compile time.
> > >
> > OK, if you have a scheme by which cleanup_cfg can determine whether it
> > needs to run, without context information and without walking the CFG,
> > I'm all ears.
>
> CFG is manipulated using hooks only. It is possible for these hooks to
> set cfg-is-dirty bit. Right now it doesn't do this. It may require
> more work then I'm imagining but that seems right approach. Unless you
> know some gotcha that I am completely missing :) And it sounds like, I
> am missing something here.
What happens when you propagate 0 into an if, CFG is not manupulated at
all in that case as far as I know.
-- Pinski