[Bug middle-end/101530] New: ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[Bug middle-end/101530] New: ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

            Bug ID: 101530
           Summary: ICE: 'verify_gimple' failed: non-top-level
                    'bit_field_ref' with __builtin_shufflevector()[0]
           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: zsojka at seznam dot cz
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu

Created attachment 51178
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51178&action=edit
reduced testcase

Compiler output:
$ x86_64-pc-linux-gnu-gcc testcase.c 
testcase.c: In function 'foo':
testcase.c:4:1: error: non-top-level 'bit_field_ref'
    4 | foo(V v, int i)
      | ^~~
_2 = VIEW_CONVERT_EXPR<int>(BIT_FIELD_REF <_1, 32, 0>);
testcase.c:4:1: internal compiler error: 'verify_gimple' failed
0x10e0a2d verify_gimple_in_seq(gimple*)
        /repo/gcc-trunk/gcc/tree-cfg.c:5188
0xdd0b46 gimplify_body(tree_node*, bool)
        /repo/gcc-trunk/gcc/gimplify.c:15526
0xdd0ccd gimplify_function_tree(tree_node*)
        /repo/gcc-trunk/gcc/gimplify.c:15597
0xbfa847 cgraph_node::analyze()
        /repo/gcc-trunk/gcc/cgraphunit.c:670
0xbfd327 analyze_functions
        /repo/gcc-trunk/gcc/cgraphunit.c:1234
0xbfdfcd symbol_table::finalize_compilation_unit()
        /repo/gcc-trunk/gcc/cgraphunit.c:2508
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.

$ x86_64-pc-linux-gnu-gcc -v
Using built-in specs.
COLLECT_GCC=/repo/gcc-trunk/binary-latest-amd64/bin/x86_64-pc-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-r12-2398-20210719103439-gf007a638a86-checking-yes-rtl-df-extra-amd64/bin/../libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-valgrind-annotations --disable-nls --enable-checking=yes,rtl,df,extra
--with-cloog --with-ppl --with-isl --build=x86_64-pc-linux-gnu
--host=x86_64-pc-linux-gnu --target=x86_64-pc-linux-gnu
--with-ld=/usr/bin/x86_64-pc-linux-gnu-ld
--with-as=/usr/bin/x86_64-pc-linux-gnu-as --disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r12-2398-20210719103439-gf007a638a86-checking-yes-rtl-df-extra-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20210719 (experimental) (GCC)

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Bet if BIT_FIELD_REF in the middle aren't allowed gimplify_compound_lval could
at least for the non-lvalue case stop walk on BIT_FIELD_REF that isn't the
outermost.  But I have no idea what to do if it is lvalue, i.e. if we want e.g.
VIEW_CONVERT_EXPR <type, BIT_FIELD_REF <var, x, y>> on the lhs.

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2021-07-21

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
I don't think we need V_C_E <BIT_FIELD_REF <>> on the LHS, V_C_E is not
size-changing so you can do

 BIT_FIELD_REF <> = V_C_E <>

instead, thus move the V_C_E to the RHS.  We're restricting BIT_FIELD_REFs
to produce is_gimple_reg_type so for the case in question we're expecting

 D.1234 = BIT_FIELD_REF <_1, 32, 0>;
 _2 = VIEW_CONVERT_EXPR<int> (D.1234);

of course we could have folded the BIT_FIELD_REF which produces a V1SI to
produce a SI, consuming the VIEW_CONVERT_EXPR.

I guess we need similar handling for compound_lval components as we have
for toplevel VIEW_CONVERT_EXPR:

        case VIEW_CONVERT_EXPR:
          if ((fallback & fb_rvalue)
              && is_gimple_reg_type (TREE_TYPE (*expr_p))
              && is_gimple_reg_type (TREE_TYPE (TREE_OPERAND (*expr_p, 0))))
            {
              ret = gimplify_expr (&TREE_OPERAND (*expr_p, 0), pre_p,
                                   post_p, is_gimple_val, fb_rvalue);
              recalculate_side_effects (*expr_p);
              break;
            }

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

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 #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
If we stop walking in gimplify_compound_lval we still gimplify the tail in the
same way, not assigning the register typed BIT_FIELD_REF to a temporary.

A bit less degenerate testcase:

typedef int __attribute__((__vector_size__ (16))) V;
int
foo(V v, int i)
{
  return __builtin_shufflevector (v, v, 2, 3)[i];
}

We can't really simplify the original GENERIC much here:

  return VIEW_CONVERT_EXPR<int[2]>(BIT_FIELD_REF < VEC_PERM_EXPR < v , v , { 2,
3, 2, 3 } > , 64, 0>)[i];

the BIT_FIELD_REF is used to subset the vector, what we'd need to somehow
ensure is to force that to an rvalue.  I suppose that would be possible
if we'd wrap it in a TARGET_EXPR somehow.  At least the following works for me:

diff --git a/gcc/c-family/c-common.c b/gcc/c-family/c-common.c
index 13341fa315e..4631d50c6f0 100644
--- a/gcc/c-family/c-common.c
+++ b/gcc/c-family/c-common.c
@@ -1243,6 +1243,8 @@ c_build_shufflevector (location_t loc, tree v0, tree v1,
       tree lpartt = build_vector_type (TREE_TYPE (ret_type), mask.length ());
       ret = build3_loc (loc, BIT_FIELD_REF,
                        lpartt, ret, TYPE_SIZE (lpartt), bitsize_zero_node);
+      tree tem = create_tmp_var_raw (lpartt);
+      ret = build4 (TARGET_EXPR, lpartt, tem, ret, NULL_TREE, NULL_TREE);
     }

   if (!c_dialect_cxx () && !wrap)

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

--- Comment #4 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:be59671c5624fe8bf21ddb0192e97ebdfa4db381

commit r12-6403-gbe59671c5624fe8bf21ddb0192e97ebdfa4db381
Author: Richard Biener <rguenther@suse.de>
Date:   Wed Jan 5 15:13:33 2022 +0100

    middle-end/101530 - fix shufflevector lowering

    This makes __builtin_shufflevector lowering force the result
    of the BIT_FIELD_REF lowpart operation to a temporary as to
    fulfil the IL verifier constraint that BIT_FIELD_REFs should
    be always in outermost handled component position.  Trying to
    enforce this during gimplification isn't as straight-forward
    as here where we know we're dealing with an rvalue.

    FAIL: c-c++-common/torture/builtin-shufflevector-1.c   -O0  execution test

    2022-01-05  Richard Biener  <rguenther@suse.de>

            PR middle-end/101530
    gcc/c-family/
            * c-common.c (c_build_shufflevector): Wrap the BIT_FIELD_REF
            in a TARGET_EXPR to force a temporary.

    gcc/testsuite/
            * c-c++-common/builtin-shufflevector-3.c: New testcase.

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

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

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

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.

[Bug middle-end/101530] ICE: 'verify_gimple' failed: non-top-level 'bit_field_ref' with __builtin_shufflevector()[0]

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101530

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:2c4dccb605d485293291878397b42927dc050481

commit r12-6514-g2c4dccb605d485293291878397b42927dc050481
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Jan 12 09:39:21 2022 +0100

    testsuite: Fix up c-c++-common/builtin-shufflevector-3.c testcase
[PR101530]

    This fixes:
    FAIL: c-c++-common/builtin-shufflevector-3.c  -Wc++-compat  (test for
excess errors)
    Excess errors:
    .../gcc/testsuite/c-c++-common/builtin-shufflevector-3.c:6:1: warning: SSE
vector argument without SSE enabled changes the ABI [-Wpsabi]

    2022-01-12  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/101530
            * c-c++-common/builtin-shufflevector-3.c: Add -Wno-psabi to
            dg-options.