[Bug middle-end/50731] New: [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[Bug middle-end/50731] New: [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

             Bug #: 50731
           Summary: [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hjl.tools@gmail.com


On Linux/x86, revision 179959 gave

FAIL: gcc.dg/torture/vector-shift2.c  -O0  execution test

Revision 179912 is OK.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-14 16:42:37 UTC ---
Revision 179952 is OK.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-10-15
   Target Milestone|---                         |4.7.0
     Ever Confirmed|0                           |1

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-15 08:58:20 UTC ---
Confirmed.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

Uros Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rth at gcc dot gnu.org

--- Comment #3 from Uros Bizjak <ubizjak at gmail dot com> 2011-10-17 17:25:53 UTC ---
CC'd author of suspected commit.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

--- Comment #4 from Uros Bizjak <ubizjak at gmail dot com> 2011-10-24 17:39:09 UTC ---
Revision 179958 FAILs.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-27 11:00:49 UTC ---
I suppose it's just completely optimized away at -O1+ and thus broken
unconditionally.  Shift lowering gives us

  vuchar_1 = { 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4 };
  vchar0_2 = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 };
  vchar1_3 = { -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
};
  D.2720_7 = BIT_FIELD_REF <vchar1_3, 8, 0>;
  D.2721_8 = D.2720_7 << 1;
  D.2722_9 = BIT_FIELD_REF <vchar1_3, 8, 8>;
  D.2723_10 = 1;
  D.2724_11 = BIT_FIELD_REF <D.2723_10, 8, 8>;

and that last two lines already look completely bogus...

We are lowering from

  vuchar_1 = { 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4 };
  vchar0_2 = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 };
  vchar1_3 = { -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
};
  i1.0_4 = vchar1_3 << 1;
  i1 = i1.0_4;
  D.2714_5 = BIT_FIELD_REF <i1, 8, 8>;

but the generic binary lowering and do_binop do not handle a scalar
RHS separately but do

static tree
do_binop (gimple_stmt_iterator *gsi, tree inner_type, tree a, tree b,
          tree bitpos, tree bitsize, enum tree_code code)
{
  a = tree_vec_extract (gsi, inner_type, a, bitsize, bitpos);
  b = tree_vec_extract (gsi, inner_type, b, bitsize, bitpos);
  return gimplify_build2 (gsi, code, inner_type, a, b);

that of course only works for element 0.

We came here via

  gcc_assert (code != VEC_LSHIFT_EXPR && code != VEC_RSHIFT_EXPR);
  new_rhs = expand_vector_operation (gsi, type, compute_type, stmt, code);

(gdb) p code
$6 = LSHIFT_EXPR

Fixing it in do_binop like

static tree
do_binop (gimple_stmt_iterator *gsi, tree inner_type, tree a, tree b,
          tree bitpos, tree bitsize, enum tree_code code)
{
  if (TREE_CODE (TREE_TYPE (a)) == VECTOR_TYPE)
    a = tree_vec_extract (gsi, inner_type, a, bitsize, bitpos);
  if (TREE_CODE (TREE_TYPE (b)) == VECTOR_TYPE)
    b = tree_vec_extract (gsi, inner_type, b, bitsize, bitpos);
  return gimplify_build2 (gsi, code, inner_type, a, b);

is easiest.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dnovillo at gcc dot gnu.org

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-27 11:38:48 UTC ---
*** Bug 50843 has been marked as a duplicate of this bug. ***

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-27 12:11:57 UTC ---
Fixed.

[Bug middle-end/50731] [4.7 Regression] FAIL: gcc.dg/torture/vector-shift2.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50731

--- Comment #7 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-27 12:11:04 UTC ---
Author: rguenth
Date: Thu Oct 27 12:10:57 2011
New Revision: 180564

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=180564
Log:
2010-10-27  Richard Guenther  <rguenther@suse.de>

    PR middle-end/50731
    * tree-vect-generic.c (do_binop): Handle scalar operands.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-vect-generic.c