[Bug c/116791] New: ICE in operator[], at vec.h:910

[Bug c/116791] New: ICE in operator[], at vec.h:910

[manuel.lauss at googlemail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

            Bug ID: 116791
           Summary: ICE in operator[], at vec.h:910
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: manuel.lauss at googlemail dot com
  Target Milestone: ---

Created attachment 59160
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=59160&action=edit
compressed unreduced testcase

The attached unreduced testcase ICEs on AMD64:

gcc version 15.0.0 20240919 (experimental)
33ccc1314dcdb0b988a9276ca6b6ce9b07bea21e (Gentoo 15.0.9999 p, commit
d676863b64b58bf73a7f26eea70f81aeba517b13)

# gcc -c nine_ff.i -O2 -mavx2
during GIMPLE pass: vect
../mesa-9999/src/gallium/frontends/nine/nine_ff.c: In function
'nine_ff_update':
../mesa-9999/src/gallium/frontends/nine/nine_ff.c:2073:1: internal compiler
error: in operator[], at vec.h:910
 2073 | nine_ff_update(struct NineDevice9 *device)
      | ^~~~~~~~~~~~~~
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
See <https://bugs.gentoo.org/> for instructions.


-O2 -mavx2 are required to trigger it; -O1 or plain -mavx make it go away.

Thanks!
     Manuel

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|ICE in operator[], at       |[15 Regression] ICE in
                   |vec.h:910                   |operator[], at vec.h:910
   Target Milestone|---                         |15.0

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Looks very recent as it worked at r15-3721-ga2746e43470 .
And it was reported against r15-3737-g33ccc1314dcdb0 .

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[manuel.lauss at googlemail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

--- Comment #2 from Manuel Lauss <manuel.lauss at googlemail dot com> ---
(In reply to Andrew Pinski from comment #1)
> Looks very recent as it worked at r15-3721-ga2746e43470 .
> And it was reported against r15-3737-g33ccc1314dcdb0 .

I suspect it's 4eba48a684  or  664e0ce5, because it broke it the last ~10
hours.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Full backtrace for r15-3741-gee3efe06c9c49c :
apinski@xeond:~/src/upstream-gcc/gcc/objdir/gcc$ ./cc1 ../../../t.i -quiet -O2
-mavx2
during GIMPLE pass: vect
../mesa-9999/src/gallium/frontends/nine/nine_ff.c: In function
‘nine_ff_update’:
../mesa-9999/src/gallium/frontends/nine/nine_ff.c:2073:1: internal compiler
error: in operator[], at vec.h:910
0x259abce internal_error(char const*, ...)
        /home/apinski/src/upstream-gcc/gcc/gcc/diagnostic-global-context.cc:517
0x9d5e87 fancy_abort(char const*, int, char const*)
        /home/apinski/src/upstream-gcc/gcc/gcc/diagnostic.cc:1657
0x8e72c0 vec<edge_def*, va_gc, vl_embed>::operator[](unsigned int)
        /home/apinski/src/upstream-gcc/gcc/gcc/vec.h:910
0x8e7dea vec<tree_node*, va_heap, vl_embed>::operator[](unsigned int)
        /home/apinski/src/upstream-gcc/gcc/gcc/tree.h:3788
0x8e7dea vec<tree_node*, va_heap, vl_ptr>::operator[](unsigned int) const
        /home/apinski/src/upstream-gcc/gcc/gcc/vec.h:1590
0x8e7dea vect_transform_slp_perm_load_1
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:9943
0x1445338 vect_transform_slp_perm_load(vec_info*, _slp_tree*, vec<tree_node*,
va_heap, vl_ptr> const&, gimple_stmt_iterator*, poly_int<1u, unsigned long>,
bool, unsigned int*, unsigned int*, bool)
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10051
0x13ef3dd vectorizable_load
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-stmts.cc:10524
0x13faa37 vect_transform_stmt(vec_info*, _stmt_vec_info*,
gimple_stmt_iterator*, _slp_tree*, _slp_instance*)
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-stmts.cc:13515
0x1448688 vect_schedule_slp_node
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10683
0x1460d1e vect_schedule_slp_node
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10470
0x1460d1e vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10921
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x1460ca3 vect_schedule_scc
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:10902
0x146132c vect_schedule_slp(vec_info*, vec<_slp_instance*, va_heap, vl_ptr>
const&)
        /home/apinski/src/upstream-gcc/gcc/gcc/tree-vect-slp.cc:11066
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |needs-bisection,
                   |                            |needs-reduction

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Working on reducing this.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-09-20
             Status|UNCONFIRMED                 |NEW
           Keywords|needs-reduction             |

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reduced testcase:
```
struct nine_context {
  unsigned tex_stage[8][33];
};
struct fvec4 {
  float x[2];
};
void f(struct fvec4 *dst, struct nine_context *context) {
  unsigned s;
  for (s = 0; s < 8; ++s)
    {
      float *rgba = &dst[s].x[0];
      unsigned color = context->tex_stage[s][0];
      rgba[0] = (float)((color >> 16) & 0xFF) / 0xFF;
      rgba[1] = (float)((color >> 8) & 0xFF) / 0xFF;
    }
}
```

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu.org,
                   |                            |sjames at gcc dot gnu.org
            Summary|[15 Regression] ICE in      |[15 Regression] ICE in
                   |operator[], at vec.h:910    |operator[], at vec.h:910
                   |                            |since
                   |                            |r15-3735-g664e0ce580a8f2

--- Comment #6 from Sam James <sjames at gcc dot gnu.org> ---
r15-3735-g664e0ce580a8f2

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

--- Comment #7 from Sam James <sjames at gcc dot gnu.org> ---
libjxl as well.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

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

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 116797 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Mine.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

--- Comment #10 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:723f7b6db841c1a101a2f5b3b6273d8449dae39e

commit r15-3791-g723f7b6db841c1a101a2f5b3b6273d8449dae39e
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Sep 23 10:13:17 2024 +0200

    tree-optimization/116791 - Elementwise SLP vectorization

    The following restricts the elementwise SLP vectorization to the
    single-lane case which is the reason I enabled it to avoid regressions
    with non-SLP.  The PR shows that multi-line SLP loads with elementwise
    accesses require work, I'll open a new bug to track this for the
    future.

            PR tree-optimization/116791
            * tree-vect-stmts.cc (get_group_load_store_type): Only
            fall back to elementwise access for single-lane SLP, restore
            hard failure mode for other cases.

            * gcc.dg/vect/pr116791.c: New testcase.

[Bug tree-optimization/116791] [15 Regression] ICE in operator[], at vec.h:910 since r15-3735-g664e0ce580a8f2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116791

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

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

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
Should be fixed.