[autovect] [patch] Make store motion use alias oracle

Revital1 Eres ERES@il.ibm.com
Wed Mar 14 11:23:00 GMT 2007



Dorit Nuzman/Haifa/IBM wrote on 14/03/2007 11:24:12:

> Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz> wrote on 14/03/2007
11:18:58:
>
> > Hello,
> >
> > > Attached is the patch that fix the problem exposed in tree-if-conv.c
as
> > > was discussed previously.  It was bootstrap and tested on ppc
together
> > > with the lim patch.
> > >
> > > OK for autovect?
> >
> > assuming that you do the required testing, this is OK for mainline as
> > well.
>
> in that case, just put it in mainline, and lets synch up the branch with
> mainline right after that.

If I understand correctly Zdenek's approval for mainline refers only to
the ifcvt fix.  Is the lim patch is OK for autovect too?

Thanks,
Revital
>
> thanks,
> dorit
>
> >
> > Zdenek
> >
> > > Thanks,
> > > Revital
> > >
> > > (See attached file: ifcvt_fix_patch_14_3.txt)
> > >
> > > 2007-03-14  Revital Eres  <eres@il.ibm.com>
> > >
> > >         * tree-if-conv.c (may_have_side_effect_p): New
> > >           function to check if stmt is if-convertible.
> > >           (if_convertible_gimple_modify_stmt_p): Call it.
> >
> > > Index: tree-if-conv.c
> > > ===================================================================
> > > --- tree-if-conv.c   (revision 122880)
> > > +++ tree-if-conv.c   (working copy)
> > > @@ -334,9 +334,38 @@
> > >    return true;
> > >  }
> > >
> > > +/* Return true if STMT may have certain side effects
> > > +   and thus ifcvt can not be applied.  Otherwise return false.  */
> > > +static bool
> > > +may_have_side_effect_p (tree stmt)
> > > +{
> > > +  tree lhs, rhs;
> > > +
> > > +  if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
> > > +    return true;
> > > +
> > > +  if (stmt_ends_bb_p (stmt))
> > > +    return true;
> > > +
> > > +  if (stmt_ann (stmt)->has_volatile_ops)
> > > +    return true;
> > > +
> > > +  lhs = GIMPLE_STMT_OPERAND (stmt, 0);
> > > +  if (TREE_CODE (lhs) == SSA_NAME
> > > +      && SSA_NAME_OCCURS_IN_ABNORMAL_PHI (lhs))
> > > +    return true;
> > > +
> > > +  rhs = GIMPLE_STMT_OPERAND (stmt, 1);
> > > +
> > > +  if (TREE_SIDE_EFFECTS (rhs))
> > > +    return true;
> > > +
> > > +  return false;
> > > +}
> > > +
> > >  /* Return true, if M_EXPR is if-convertible.
> > >     GIMPLE_MODIFY_STMT is not if-convertible if,
> > > -   - It is not movable.
> > > +   - It may have certain side effects.
> > >     - It could trap.
> > >     - LHS is not var decl.
> > >    GIMPLE_MODIFY_STMT is part of block BB, which is inside loop LOOP.
> > > @@ -352,11 +381,12 @@
> > >        print_generic_stmt (dump_file, m_expr, TDF_SLIM);
> > >      }
> > >
> > > -  /* Be conservative and do not handle immovable expressions.  */
> > > -  if (movement_possibility (m_expr) == MOVE_IMPOSSIBLE)
> > > +  /* Be conservative and do not handle expressions with
> > > +     certain side effects.  */
> > > +  if (may_have_side_effect_p (m_expr))
> > >      {
> > >        if (dump_file && (dump_flags & TDF_DETAILS))
> > > -   fprintf (dump_file, "stmt is movable. Don't take risk\n");
> > > +   fprintf (dump_file, "stmt with side effects. Don't take risk\n");
> > >        return false;
> > >      }
> > >
> >



More information about the Gcc-patches mailing list