tree-ssa/lno compute_may_aliases failure

[Dorit Naishlos]

> IIRC PTA cannot be applied once we are in SSA form.  That's why you see
> it sequenced before pass_build_ssa (Dan, has that changed?).

In that case, maybe I should instead postpone delete_pta pass to after the
second invocation of compute_may_aliases? (PTA is a prerequisite for
compute_may_alias pass).

> You seem to have run into a bug in pass_may_alias.  It almost looks like
> we are not marking some variable for renaming.  Could you file a PR with
> your patch and test case and assign it to me?  I'm currently busy with
> something else and I don't want to forget about this one.

thank you. it's PR 15555.

dorit



                                                                                                                                       
                      Diego Novillo                                                                                                    
                      <dnovillo@redhat.        To:       Dorit Naishlos/Haifa/IBM@IBMIL                                                
                      com>                     cc:       "gcc@gcc.gnu.org" <gcc@gcc.gnu.org>                                           
                                               Subject:  Re: tree-ssa/lno compute_may_aliases failure                                  
                      20/05/2004 16:00                                                                                                 
                                                                                                                                       




On Wed, 2004-05-19 at 17:29, Dorit Naishlos wrote:

> Is there a known reason why compute_may_aliases shouldn't be applied
where
> I applied it? should I have called an additional pass before it?
>
None.  In theory, you should be able to run alias anywhere.  However,
IIRC PTA cannot be applied once we are in SSA form.  That's why you see
it sequenced before pass_build_ssa (Dan, has that changed?).

You seem to have run into a bug in pass_may_alias.  It almost looks like
we are not marking some variable for renaming.  Could you file a PR with
your patch and test case and assign it to me?  I'm currently busy with
something else and I don't want to forget about this one.


Thanks.  Diego.