[Bug middle-end/46761] New: [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[Bug middle-end/46761] New: [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

           Summary: [4.6 Regression] -fgraphite-identity produces wrong
                    code for array initialization arr[i] = i
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz
                CC: spop@gcc.gnu.org
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu


Created attachment 22599
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22599
reduced testcase

This might be the common cause of several testsuite failures that appear with
-fgraphite-identity in testflags.

Output:
$ gcc -O -fgraphite-identity pr46761.c
$ valgrind -q ./a.out 
==20462== Conditional jump or move depends on uninitialised value(s)
==20462==    at 0x40053F: main (pr46761.c:12)
==20462== 
Aborted

Tested revisions:
r167356 - fail
r165699 - fail
r161659 - OK
4.5 r166509 - OK

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

--- Comment #1 from Zdenek Sojka <zsojka at seznam dot cz> 2010-12-02 02:32:58 UTC ---
C failures I am currently (trunk r167356) seeing at x86_64-linux with
-fgraphite-identity:

FAIL: gcc.c-torture/execute/20100416-1.c execution,  -O1 
FAIL: gcc.dg/vect/pr37539.c execution test
FAIL: gcc.dg/vect/vect-25.c execution test
FAIL: gcc.dg/vect/vect-26.c execution test
FAIL: gcc.dg/vect/vect-27.c execution test
FAIL: gcc.dg/vect/vect-28.c execution test
FAIL: gcc.dg/vect/vect-29.c execution test
FAIL: gcc.dg/vect/vect-54.c execution test
FAIL: gcc.dg/vect/vect-56.c execution test
FAIL: gcc.dg/vect/vect-7.c execution test
FAIL: gcc.dg/vect/vect-72.c execution test
FAIL: gcc.dg/vect/vect-peel-1.c execution test
FAIL: gcc.dg/vect/vect-peel-3.c execution test
FAIL: gcc.dg/vect/vect-strided-store-u16-i4.c execution test
FAIL: gcc.dg/vect/slp-1.c execution test
FAIL: gcc.dg/vect/slp-2.c execution test
FAIL: gcc.dg/vect/slp-20.c execution test
FAIL: gcc.dg/vect/slp-21.c execution test
FAIL: gcc.dg/vect/slp-22.c execution test
FAIL: gcc.dg/vect/slp-multitypes-1.c execution test
FAIL: gcc.dg/vect/slp-multitypes-12.c execution test
FAIL: gcc.dg/vect/slp-multitypes-2.c execution test
FAIL: gcc.dg/vect/slp-reduc-5.c execution test
FAIL: gcc.dg/vect/slp-reduc-6.c execution test
FAIL: gcc.dg/vect/no-vfa-vect-57.c execution test
FAIL: gcc.dg/vect/wrapv-vect-7.c execution test
FAIL: gcc.dg/vect/no-section-anchors-vect-outer-4h.c execution test
FAIL: gcc.dg/vect/no-scevccp-slp-30.c execution test
FAIL: gcc.dg/vect/no-scevccp-slp-31.c execution test

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.12.02 03:03:05
   Target Milestone|---                         |4.6.0
     Ever Confirmed|0                           |1

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> 2010-12-02 03:03:05 UTC ---
It is caused by revision 164772:

http://gcc.gnu.org/ml/gcc-cvs/2010-09/msg01071.html

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

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

--- Comment #3 from Alexander Monakov <amonakov at gcc dot gnu.org> 2010-12-03 16:54:26 UTC ---
Sometimes graphite generates wrong guards for regions.  The following patch
fixes the attached testcase for me, but I have not tested it any further.

diff --git a/gcc/graphite-clast-to-gimple.c b/gcc/graphite-clast-to-gimple.c
index 9a90ef7..8a80033 100644
--- a/gcc/graphite-clast-to-gimple.c
+++ b/gcc/graphite-clast-to-gimple.c
@@ -986,7 +986,7 @@ graphite_create_new_loop_guard (sese region, edge
entry_edge,
                  : PLUS_EXPR, type, ub, one);

   /* When ub + 1 wraps around, use lb <= ub.  */
-  if (integer_zerop (ub_one))
+  if (TREE_OVERFLOW_P (ub_one))
     cond_expr = fold_build2 (LE_EXPR, boolean_type_node, lb, ub);
   else
     cond_expr = fold_build2 (LT_EXPR, boolean_type_node, lb, ub_one);

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

--- Comment #4 from Zdenek Sojka <zsojka at seznam dot cz> 2010-12-06 21:23:00 UTC ---
(In reply to comment #3)
> Sometimes graphite generates wrong guards for regions.  The following patch
> fixes the attached testcase for me, but I have not tested it any further.

This patch fixes all failures except two:
FAIL: gcc.c-torture/execute/20100416-1.c execution,  -O1 (PR46758)
FAIL: gfortran.dg/array_constructor_12.f90  -Os  execution test (probably the
same as PR46758 ; I didn't run fortran testsuite for comment #1)

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |amonakov at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #5 from Alexander Monakov <amonakov at gcc dot gnu.org> 2010-12-08 18:05:23 UTC ---
Thanks.  I'll submit this patch

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

--- Comment #6 from Alexander Monakov <amonakov at gcc dot gnu.org> 2010-12-17 12:55:03 UTC ---
Author: amonakov
Date: Fri Dec 17 12:54:59 2010
New Revision: 167980

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=167980
Log:
    PR middle-end/46761
    * graphite-clast-to-gimple.c (graphite_create_new_loop_guard): Prefer
    to use unadjusted UB.

testsuite:
    * gcc.dg/graphite/pr46761.c: New.


Added:
    trunk/gcc/testsuite/gcc.dg/graphite/pr46761.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/graphite-clast-to-gimple.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/46761] [4.6 Regression] -fgraphite-identity produces wrong code for array initialization arr[i] = i

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46761

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

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

--- Comment #7 from Alexander Monakov <amonakov at gcc dot gnu.org> 2010-12-17 12:57:27 UTC ---
A different patch was committed after the review.  Closing as fixed.