[Bug tree-optimization/48765] New: ICE in vect_transform_stmt

[Bug tree-optimization/48765] New: ICE in vect_transform_stmt

[pthaugen at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

           Summary: ICE in vect_transform_stmt
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: pthaugen@gcc.gnu.org
                CC: bergner@gcc.gnu.org, meissner@gcc.gnu.org
              Host: powerpc64-linux
            Target: powerpc64-linux
             Build: powerpc64-linux


Created attachment 24096
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24096
reduced testcase

Following showed up recently in build of cpu2000 benchmark 176.gcc.

> /home/pthaugen/install/gcc/trunk/bin/gcc -c -m64 -O3 -mcpu=power6 stupid.c
stupid.c: In function ‘stupid_life_analysis’:
stupid.c:33:1: internal compiler error: in vect_transform_stmt, at
tree-vect-stmts.c:4867
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Bug tree-optimization/48765] ICE in vect_transform_stmt

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

William J. Schmidt <wschmidt at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |richard.sandiford at linaro
                   |                            |dot org

--- Comment #1 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2011-04-26 20:11:53 UTC ---
The bug was introduced with r172817, committed by Richard Sandiford.  Richard,
could you please take a look?

tree-vect-loop.c:vectorizable_reduction() is returning false after failing the
"Worthwhile without SIMD support?" check for this statement:

D.2185_89 = D.2184_88 | prephitmp.53_140;

The vectorization factor for the loop is 1, but vect_min_worthwhile_factor
(BIT_IOR_EXPR) = 2, resulting in the return of false, which in turn causes an
assert failure.

Thanks,
Bill

[Bug tree-optimization/48765] ICE in vect_transform_stmt

[irar at il dot ibm.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

Ira Rosen <irar at il dot ibm.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.04.27 07:25:43
                 CC|                            |irar at il dot ibm.com
     Ever Confirmed|0                           |1

--- Comment #2 from Ira Rosen <irar at il dot ibm.com> 2011-04-27 07:25:43 UTC ---
I'll take a look.

Ira

[Bug tree-optimization/48765] ICE in vect_transform_stmt

[irar at il dot ibm.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

--- Comment #3 from Ira Rosen <irar at il dot ibm.com> 2011-04-27 10:25:16 UTC ---
As analyzed by Richard, the problem (btw, unrelated to Richard's patch) is that
the vectorization factor is changed after the analysis for this loop, and the
same checks that worked fine during the analysis, now fail during the
transformation.

I'll try to move the vectorization factor calculation before the statements
analysis to avoid this. 

Ira

[Bug tree-optimization/48765] ICE in vect_transform_stmt

[irar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

--- Comment #4 from irar at gcc dot gnu.org 2011-04-28 19:50:30 UTC ---
Author: irar
Date: Thu Apr 28 19:50:28 2011
New Revision: 173132

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=173132
Log:

        PR tree-optimization/48765
        * tree-vectorizer.h (vect_make_slp_decision): Return bool.
        * tree-vect-loop.c (vect_analyze_loop_operations): Add new
        argument to indicate if loop aware SLP is being used.  Scan
        the statements and update the vectorization factor
        according to the type of
        vectorization before statement analysis.
        (vect_analyze_loop_2): Get a return value from
        vect_make_slp_decision, pass it to 
        vect_analyze_loop_operations.
        (vectorizable_reduction): Set number of copies to 1 in case of
        pure SLP statement.
        * tree-vect-stmts.c (vectorizable_conversion,
        vectorizable_assignment, vectorizable_shift,
        vectorizable_operation, vectorizable_type_demotion,
        vectorizable_type_promotion, vectorizable_store,
        vectorizable_load): Likewise.
        (vectorizable_condition): Move the check that it is not SLP
        vectorization before the number of copies check.
        * tree-vect-slp.c (vect_make_slp_decision): Return TRUE if
        decided to vectorize the loop using SLP.


Added:
    trunk/gcc/testsuite/gcc.dg/vect/pr48765.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-vect-loop.c
    trunk/gcc/tree-vect-slp.c
    trunk/gcc/tree-vect-stmts.c
    trunk/gcc/tree-vectorizer.h

[Bug tree-optimization/48765] ICE in vect_transform_stmt

[irar at il dot ibm.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48765

Ira Rosen <irar at il dot ibm.com> changed:

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

--- Comment #5 from Ira Rosen <irar at il dot ibm.com> 2011-04-29 07:01:44 UTC ---
Fixed.