Re: [patch] Fix PR tree-optimization/46167

[Ira Rosen <IRAR at il dot ibm dot com>]

Richard Guenther <richard.guenther@gmail.com> wrote on 26/10/2010 09:12:50
PM:

> On Tue, Oct 26, 2010 at 11:53 AM, Ira Rosen <IRAR@il.ibm.com> wrote:
> >
> > Hi,
> >
> > This patch adds a check that a basic block exists before calling
> > flow_bb_inside_loop_p.
> >
> > Bootstrapped and tested on x86_64-suse-linux.
> > Committed to trunk.
> >
> > Now testing the patch for 4.5 on x86_64-suse-linux. OK for 4.5 once the
> > testing completes?
>
> Why is the stmt not in any BB?

It's a function argument.

Ira

>
> Richard.
>
> > Thanks,
> > Ira
> >
> > ChangeLog:
> >
> > ? ? ? ?PR tree-optimization/46167
> > ? ? ? ?* tree-vect-patterns.c (vect_recog_dot_prod_pattern): Check that
> > there
> > ? ? ? ?is basic block for the statement.
> >
> > testsuite/ChangeLog:
> >
> > ? ? ? ?PR tree-optimization/46167
> > ? ? ? ?* gcc.dg/vect/O-pr46167.c: New.
> > ? ? ? ?* gcc.dg/vect/vect.exp: Run tests with -O.
> >
> >
> > 4.6 patch:
> >
> > Index: testsuite/gcc.dg/vect/O-pr46167.c
> > ===================================================================
> > --- testsuite/gcc.dg/vect/O-pr46167.c ? (revision 0)
> > +++ testsuite/gcc.dg/vect/O-pr46167.c ? (revision 0)
> > @@ -0,0 +1,11 @@
> > +/* { dg-do compile } */
> > +
> > +int foo (char c, int i)
> > +{
> > + ?int s = 0;
> > + ?while (i--)
> > + ? ?s += c;
> > + ?return s;
> > +}
> > +
> > +/* { dg-final { cleanup-tree-dump "vect" } } */
> > Index: testsuite/gcc.dg/vect/vect.exp
> > ===================================================================
> > --- testsuite/gcc.dg/vect/vect.exp ? ? ?(revision 165969)
> > +++ testsuite/gcc.dg/vect/vect.exp ? ? ?(working copy)
> > @@ -126,6 +126,11 @@ set O1_VECTCFLAGS $DEFAULT_VECTCFLAGS
> > ?lappend O1_VECTCFLAGS "-O1"
> > ?lappend O1_VECTCFLAGS "-fdump-tree-vect-details"
> >
> > +global O_VECTCFLAGS
> > +set O_VECTCFLAGS $DEFAULT_VECTCFLAGS
> > +lappend O_VECTCFLAGS "-O"
> > +lappend O_VECTCFLAGS "-fdump-tree-vect-details"
> > +
> > ?lappend DEFAULT_VECTCFLAGS "-O2"
> >
> > ?# Tests that should be run without generating dump info
> > @@ -287,6 +292,10 @@ dg-runtest [lsort [glob -nocomplain $src
> > ?dg-runtest [lsort [glob -nocomplain $srcdir/$subdir/O1-*.\[cS\]]] ?\
> > ? ? ? ? "" $O1_VECTCFLAGS
> >
> > +# With -O
> > +dg-runtest [lsort [glob -nocomplain $srcdir/$subdir/O-*.\[cS\]]] ?\
> > + ? ? ? ?"" $O_VECTCFLAGS
> > +
> > ?# -fno-tree-reassoc
> > ?set VECT_SLP_CFLAGS $SAVED_VECT_SLP_CFLAGS
> > ?lappend VECT_SLP_CFLAGS "-fno-tree-reassoc"
> > Index: tree-vect-patterns.c
> > ===================================================================
> > --- tree-vect-patterns.c ? ? ? ?(revision 165969)
> > +++ tree-vect-patterns.c ? ? ? ?(working copy)
> > @@ -257,7 +257,7 @@ vect_recog_dot_prod_pattern (gimple last
> > ? stmt = SSA_NAME_DEF_STMT (oprnd0);
> >
> > ? /* It could not be the dot_prod pattern if the stmt is outside the
loop.
> > */
> > - ?if (!flow_bb_inside_loop_p (loop, gimple_bb (stmt)))
> > + ?if (!gimple_bb (stmt) || !flow_bb_inside_loop_p (loop, gimple_bb
> > (stmt)))
> > ? ? return NULL;
> >
> > ? /* FORNOW. ?Can continue analyzing the def-use chain when this stmt
in a
> > phi
> >
> >
> > 4.5 patch:
> >
> > Index: testsuite/gcc.dg/vect/O-pr46167.c
> > ===================================================================
> > --- testsuite/gcc.dg/vect/O-pr46167.c ? (revision 0)
> > +++ testsuite/gcc.dg/vect/O-pr46167.c ? (revision 0)
> > @@ -0,0 +1,11 @@
> > +/* { dg-do compile } */
> > +
> > +int foo (char c, int i)
> > +{
> > + ?int s = 0;
> > + ?while (i--)
> > + ? ?s += c;
> > + ?return s;
> > +}
> > +
> > +/* { dg-final { cleanup-tree-dump "vect" } } */
> > Index: testsuite/gcc.dg/vect/vect.exp
> > ===================================================================
> > --- testsuite/gcc.dg/vect/vect.exp ? ? ?(revision 165269)
> > +++ testsuite/gcc.dg/vect/vect.exp ? ? ?(working copy)
> > @@ -122,6 +122,11 @@ set O1_VECTCFLAGS $DEFAULT_VECTCFLAGS
> > ?lappend O1_VECTCFLAGS "-O1"
> > ?lappend O1_VECTCFLAGS "-fdump-tree-vect-details"
> >
> > +global O_VECTCFLAGS
> > +set O_VECTCFLAGS $DEFAULT_VECTCFLAGS
> > +lappend O_VECTCFLAGS "-O"
> > +lappend O_VECTCFLAGS "-fdump-tree-vect-details"
> > +
> > ?lappend DEFAULT_VECTCFLAGS "-O2"
> >
> > ?# Tests that should be run without generating dump info
> > @@ -283,6 +288,10 @@ dg-runtest [lsort [glob -nocomplain $src
> > ?dg-runtest [lsort [glob -nocomplain $srcdir/$subdir/O1-*.\[cS\]]] ?\
> > ? ? ? ? "" $O1_VECTCFLAGS
> >
> > +# With -O
> > +dg-runtest [lsort [glob -nocomplain $srcdir/$subdir/O-*.\[cS\]]] ?\
> > + ? ? ? ?"" $O_VECTCFLAGS
> > +
> > ?# -fno-tree-reassoc
> > ?set VECT_SLP_CFLAGS $SAVED_VECT_SLP_CFLAGS
> > ?lappend VECT_SLP_CFLAGS "-fno-tree-reassoc"
> > Index: tree-vect-patterns.c
> > ===================================================================
> > --- tree-vect-patterns.c ? ? ? ?(revision 165269)
> > +++ tree-vect-patterns.c ? ? ? ?(working copy)
> > @@ -256,7 +256,7 @@ vect_recog_dot_prod_pattern (gimple last
> > ? stmt = SSA_NAME_DEF_STMT (oprnd0);
> >
> > ? /* It could not be the dot_prod pattern if the stmt is outside the
loop.
> > */
> > - ?if (!flow_bb_inside_loop_p (loop, gimple_bb (stmt)))
> > + ?if (!gimple_bb (stmt) || !flow_bb_inside_loop_p (loop, gimple_bb
> > (stmt)))
> > ? ? return NULL;
> >
> > ? /* FORNOW. ?Can continue analyzing the def-use chain when this stmt
in a
> > phi
> >
> >