This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Post Dominator II
- To: gcc at gcc dot gnu dot org
- Subject: Re: Post Dominator II
- From: Marcio de Oliveira Buss <ra990898 at ic dot unicamp dot br>
- Date: Wed, 5 Jul 2000 16:28:25 -0300 (EST)
On Wed, 5 Jul 2000, Richard Henderson wrote:
> On Tue, Jul 04, 2000 at 03:07:10PM -0300, Marcio de Oliveira Buss wrote:
> >
> > compute_preds_succs(S_Preds, S_Succs, Num_Preds, Num_Succs);
>
> FYI, this has been removed from the mainline compiler since 2.95.
But I have actually the version 2.95.2 and this function is essential
to the computation, I think. Without it, the information about
predecessors/succesors will not be computed at the time which
compute_dominators is called. Am I right?
>
> > compute_dominators(Dominators, PostDominators, S_Preds, S_Succs);
>
> You needn't provide storage for Dominators if you don't want;
> leave the argument NULL and avoid the work of computing it.
Yes, it is right. Thanks.
>
> > compute_immediate_dominators(Ipdom, PostDominators);
> >
> > Meaning that, after the computation, Ipdom[bb] contains the index of
> > the basic block that would be the unique parent of 'bb' in the
> > post-dominator tree. Of course, this parent could have many children.
>
> Yes.
>
>
> r~
>