[Bug tree-optimization/56778] New: ICE on several benchmarks after r196775.

[Bug tree-optimization/56778] New: ICE on several benchmarks after r196775.

[ysrumyan at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

             Bug #: 56778
           Summary: ICE on several benchmarks after r196775.
    Classification: Unclassified
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ysrumyan@gmail.com


We got ICE on several benchmarks that can be reproduced on the simple
test-case:
typedef struct {
  float a,b,c;
} S;

S * arr[100];

void bar (float *in[], int n)
{
  int i;
  for (i=0; i<n; i++)
    (*in)[i] = -arr[i]->b;
}

and compile on x86 with following options:

-march=core-avx2 -O3

[Bug tree-optimization/56778] ICE on several benchmarks after r196775.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2013-04-02
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-02 09:04:15 UTC ---
Confirmed.  I will have a look (there is a not properly removed stmt with
still has active immediate uses).

[Bug tree-optimization/56778] [4.9 Regression] ICE on several benchmarks after r196775.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.9.0
            Summary|ICE on several benchmarks   |[4.9 Regression] ICE on
                   |after r196775.              |several benchmarks after
                   |                            |r196775.

[Bug tree-optimization/56778] [4.9 Regression] ICE on several benchmarks after r196775.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-02 11:19:35 UTC ---
The issue is that gimple_seq_add_seq updates stmts added.  But we update SSA
form _before_ adding the sequence into the IL stream:

  /* End loop-exit-fixes after versioning.  */

  update_ssa (TODO_update_ssa);
  if (cond_expr_stmt_list)
    {
      cond_exp_gsi = gsi_last_bb (condition_bb);
      gsi_insert_seq_before (&cond_exp_gsi, cond_expr_stmt_list,
                             GSI_SAME_STMT);
    }

gimplify.c has a function that does gimple_seq_add_seq_without_update.

There are two ways to fix this, one is to insert the sequence before
updating SSA form, another is to consistently use
gimple_seq_add_seq_without_update.
Or to make gimple_seq_add_seq not update stmts (sequences are inserted
later anyway, which can do the update).  Not sure why it would make sense
at all to update stmt operands for sth not in the IL ...

Well.  I'm going for the first.

Which doesn't fix it because we create an alias-check for a load of a
load appearantly where the latter is not loop invariant.  I presume
this is GATHER support, and eventually the DDR mangling in data-ref
analysis I removed did avoid this to happen.

[Bug tree-optimization/56778] [4.9 Regression] ICE on several benchmarks after r196775.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-02 11:27:09 UTC ---
I have a fix.

[Bug tree-optimization/56778] [4.9 Regression] ICE on several benchmarks after r196775.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56778

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-02 13:31:40 UTC ---
Author: rguenth
Date: Tue Apr  2 13:31:05 2013
New Revision: 197355

URL: http://gcc.gnu.org/viewcvs?rev=197355&root=gcc&view=rev
Log:
2013-04-02  Richard Biener  <rguenther@suse.de>

    PR tree-optimization/56778
    * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
    Runtime alias tests are not supported for gather loads.
    * tree-vect-loop-manip.c (vect_loop_versioning): Insert
    stmts referenced from SSA operands before updating SSA form.

    * gcc.dg/torture/pr56778.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr56778.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-vect-data-refs.c
    trunk/gcc/tree-vect-loop-manip.c