[Bug target/96128] New: [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[Bug target/96128] New: [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

            Bug ID: 96128
           Summary: [11 Regression] ICE in do_store_flag, at expr.c:12247
                    since
                    r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
                CC: rguenth at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-linux-gnu
            Target: s390x-linux-gnu

Since my revision the following fails:

$ ./xgcc -B.
/home/marxin/Programming/gcc/gcc/testsuite/c-c++-common/opaque-vector.c
-march=z13 -c
during RTL pass: expand
/home/marxin/Programming/gcc/gcc/testsuite/c-c++-common/opaque-vector.c: In
function ‘f’:
/home/marxin/Programming/gcc/gcc/testsuite/c-c++-common/opaque-vector.c:1:65:
internal compiler error: in do_store_flag, at expr.c:12247
    1 | #define B_TEST(TYPE) { TYPE v __attribute__((vector_size(16)));
(void)((v < v) < v); }
      |                                                                
^~~~~~~~~~~~~~~~~~~
/home/marxin/Programming/gcc/gcc/testsuite/c-c++-common/opaque-vector.c:7:17:
note: in expansion of macro ‘B_TEST’
    7 | #define T(TYPE) B_TEST(TYPE) T_TEST(TYPE)
      |                 ^~~~~~
/home/marxin/Programming/gcc/gcc/testsuite/c-c++-common/opaque-vector.c:16:3:
note: in expansion of macro ‘T’
   16 |   T(short)
      |   ^
0xc498c7 do_store_flag
        /home/marxin/Programming/gcc2/gcc/expr.c:12247
0xc3c9a6 expand_expr_real_2(separate_ops*, rtx_def*, machine_mode,
expand_modifier)
        /home/marxin/Programming/gcc2/gcc/expr.c:9608
0xaa600b expand_gimple_stmt_1
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:3786
0xaa6258 expand_gimple_stmt
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:3847
0xaae714 expand_gimple_basic_block
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:5888
0xab0576 execute
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:6572
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

The change is now that the comparison is expanded in GIMPLE to:
(opaque-vector.c.005t.gimple)

f ()
{
  {
    vector(8) short int v;

    _1 = v > { 0, 0, 0, 0, 0, 0, 0, 0 };
  }
  {
    vector(4) int v;

    _2 = v > { 0, 0, 0, 0 };
  }
  {
    vector(2) long int v;

    _3 = v > { 0, 0 };
  }
  {
    vector(2) long long int v;

    _4 = v > { 0, 0 };
  }
}

while before the revision it was to 

f ()
{
  {
    vector(8) short int v;


  }
  {
    vector(4) int v;


  }
  {
    vector(2) long int v;


  }
  {
    vector(2) long long int v;


  }
}

What kind of target hook is responsible for that?

[Bug target/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-07-09

[Bug target/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Just trace make_ssa_name ...

[Bug target/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> Just trace make_ssa_name ...

So this one is leftover from gimplify.c where we expand the first argument to
assignment to a SSA NAME and so we end up only with the vector comparisons, but
no VEC_COND_EXPR.
Would it be possible to remove the SSA_NAME in gimplify.c when we know the
VEC_COND_EXPR is expanded to no-op?

[Bug target/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

--- Comment #3 from rguenther at suse dot de <rguenther at suse dot de> ---
On Thu, 9 Jul 2020, marxin at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128
> 
> --- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
> (In reply to Richard Biener from comment #1)
> > Just trace make_ssa_name ...
> 
> So this one is leftover from gimplify.c where we expand the first argument to
> assignment to a SSA NAME and so we end up only with the vector comparisons, but
> no VEC_COND_EXPR.
> Would it be possible to remove the SSA_NAME in gimplify.c when we know the
> VEC_COND_EXPR is expanded to no-op?

Somehow switch gimplifying of the operand and evaluating "expanded to 
no-op", other than that, no - the gimplifier doesn't perform DCE.

[Bug target/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #4 from Martin Liška <marxin at gcc dot gnu.org> ---
All right, I have a feasible solution.

[Bug tree-optimization/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Martin Liska <marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:55c9695cbe91c406805ac0cd342949f32f13b779

commit r11-2362-g55c9695cbe91c406805ac0cd342949f32f13b779
Author: Martin Liska <mliska@suse.cz>
Date:   Mon Jul 27 13:55:38 2020 +0200

    Do not expand vector comparison with VEC_COND_EXPR.

    gcc/ChangeLog:

            PR tree-optimization/96128
            * tree-vect-generic.c (expand_vector_comparison): Do not expand
            vector comparison with VEC_COND_EXPR.

    gcc/testsuite/ChangeLog:

            PR tree-optimization/96128
            * gcc.target/s390/vector/pr96128.c: New test.

[Bug tree-optimization/96128] [11 Regression] ICE in do_store_flag, at expr.c:12247 since r11-1445-g502d63b6d6141597bb18fd23c87736a1b384cf8f

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96128

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #6 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed.