[Bug middle-end/101861] New: [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[Bug middle-end/101861] New: [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

            Bug ID: 101861
           Summary: [12 Regression] ICE (segfault) via
                    tree-vect-stmts.c:9495's vectorizable_load
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
  Target Milestone: ---
            Target: amdgcn-amdhsa

I see the following new ICE – it did work in yesterday's run:


amdgcn-amdhsa-gcc .../testsuite/gcc.dg/loop-versioning-6.c
-fdiagnostics-plain-output -O3 -fdump-tree-lversion-details -S -o
loop-versioning-6.s

during GIMPLE pass: vect
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/testsuite/gcc.dg/loop-versioning-6.c:
In function 'f1':
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/testsuite/gcc.dg/loop-versioning-6.c:12:1:
internal compiler error: Segmentation fault
0xedbd8f crash_signal
        /scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/toplev.c:328
0x114c130 vec<tree_node*, va_heap, vl_embed>::operator[](unsigned int)
        /scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/vec.h:889
0x114c130 vec<tree_node*, va_heap, vl_ptr>::operator[](unsigned int)
        /scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/vec.h:1495
0x114c130 vectorizable_load
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vect-stmts.c:9495
0x1157a28 vect_transform_stmt(vec_info*, _stmt_vec_info*,
gimple_stmt_iterator*, _slp_tree*, _slp_instance*)
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vect-stmts.c:11062
0x115c9ca vect_transform_loop_stmt
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vect-loop.c:9323
0x11745e5 vect_transform_loop(_loop_vec_info*, gimple*)
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vect-loop.c:9759
0x11a6f9e try_vectorize_loop_1
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vectorizer.c:1103
0x11a77c1 vectorize_loops()
       
/scratch/ci-cs/amdtest/upstream/src/gcc-mainline/gcc/tree-vectorizer.c:1242

 * * *

New FAIL: default/gcc.sum:gcc.dg/loop-versioning-6.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/torture/pr53390.c   -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions  (internal compiler
error)
New FAIL: default/gcc.sum:gcc.dg/torture/pr53390.c   -O3 -g  (internal compiler
error)
New FAIL: default/gcc.sum:gcc.dg/torture/pr91178-2.c   -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions  (internal compiler
error)
New FAIL: default/gcc.sum:gcc.dg/torture/pr91178-2.c   -O3 -g  (internal
compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/O3-pr85794.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/pr53185-2.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/pr53185.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/pr85793.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/pr91403.c (internal compiler error)
New FAIL: default/gcc.sum:gcc.dg/vect/vect-123.c (internal compiler error)
...

[Bug middle-end/101861] [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Target Milestone|---                         |12.0
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-08-11
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

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

[Bug middle-end/101861] [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
OK, so we're having memory_access_type == VMAT_GATHER_SCATTER, gs_info.ifn !=
IFN_LAST but STMT_VINFO_GATHER_SCATTER_P is false.  It looks like vec_offset is
set elsewhere in that case.  Likely here:

  else if (memory_access_type == VMAT_GATHER_SCATTER)
    {
      aggr_type = elem_type;
      vect_get_strided_load_store_ops (stmt_info, loop_vinfo, &gs_info,
                                       &bump, &vec_offset);
    }

so the fix is to re-instante the guard, doing

diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
index ab402b57fb4..cc6c091e41e 100644
--- a/gcc/tree-vect-stmts.c
+++ b/gcc/tree-vect-stmts.c
@@ -9492,7 +9492,8 @@ vectorizable_load (vec_info *vinfo,
                    if (memory_access_type == VMAT_GATHER_SCATTER
                        && gs_info.ifn != IFN_LAST)
                      {
-                       vec_offset = vec_offsets[j];
+                       if (STMT_VINFO_GATHER_SCATTER_P (stmt_info))
+                         vec_offset = vec_offsets[j];
                        tree zero = build_zero_cst (vectype);
                        tree scale = size_int (gs_info.scale);
                        gcall *call;

will push this as obvious.

[Bug middle-end/101861] [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

--- Comment #3 from CVS 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:f19c70afa4aa2416c581484a6eb51c0ac46acd08

commit r12-2849-gf19c70afa4aa2416c581484a6eb51c0ac46acd08
Author: Richard Biener <rguenther@suse.de>
Date:   Wed Aug 11 11:49:10 2021 +0200

    tree-optimization/101861 - fix gather use for non-gather refs

    My previous change broke the usage of gather for strided loads.
    The following fixes it.

    2021-08-11  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/101861
            * tree-vect-stmts.c (vectorizable_load): Fix error in
            previous change with regard to gather vectorization.

[Bug middle-end/101861] [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Should be fixed now.

[Bug middle-end/101861] [12 Regression] ICE (segfault) via tree-vect-stmts.c:9495's vectorizable_load

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101861

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

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

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
*** Bug 101890 has been marked as a duplicate of this bug. ***