[Bug rtl-optimization/103485] New: [12 Regression] ICE: in emit_move_insn, at expr.c:4013 with -O -fno-tree-dce -fno-tree-dse

[Bug rtl-optimization/103485] New: [12 Regression] ICE: in emit_move_insn, at expr.c:4013 with -O -fno-tree-dce -fno-tree-dse

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103485

            Bug ID: 103485
           Summary: [12 Regression] ICE: in emit_move_insn, at expr.c:4013
                    with -O -fno-tree-dce -fno-tree-dse
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu

Created attachment 51905
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51905&action=edit
reduced testcase

Compiler output:
$ x86_64-pc-linux-gnu-gcc -O -fno-tree-dce -fno-tree-dse testcase.c 
during RTL pass: expand
testcase.c: In function 'foo':
testcase.c:5:1: internal compiler error: in emit_move_insn, at expr.c:4013
    5 | foo (void)
      | ^~~
0x6bfb55 emit_move_insn(rtx_def*, rtx_def*)
        /repo/gcc-trunk/gcc/expr.c:4013
0xe517aa expand_gimple_stmt_1
        /repo/gcc-trunk/gcc/cfgexpand.c:3994
0xe517aa expand_gimple_stmt
        /repo/gcc-trunk/gcc/cfgexpand.c:4028
0xe57508 expand_gimple_basic_block
        /repo/gcc-trunk/gcc/cfgexpand.c:6069
0xe59697 execute
        /repo/gcc-trunk/gcc/cfgexpand.c:6795
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/bin/x86_64-pc-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-r12-5590-20211129212232-g6c7d489a1e6-checking-yes-rtl-df-extra-nobootstrap-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
--disable-bootstrap --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-r12-5590-20211129212232-g6c7d489a1e6-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20211130 (experimental) (GCC)

[Bug rtl-optimization/103485] [12 Regression] ICE: in emit_move_insn, at expr.c:4013 with -O -fno-tree-dce -fno-tree-dse

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103485

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-11-30
             Status|UNCONFIRMED                 |ASSIGNED
           Priority|P3                          |P1
     Ever confirmed|0                           |1
             Target|                            |x86_64-*-*
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
   Target Milestone|---                         |12.0

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

(gdb) p debug_gimple_stmt (stmt)
_12 = (vector(8) signed int) foo_v256u8_0.0_1;
$1 = void
(gdb) p debug_tree (gimple_assign_lhs (stmt))
 <ssa_name 0x7ffff66ab3a8
    type <vector_type 0x7ffff66a5348
        type <integer_type 0x7ffff66a52a0 SI
            size <integer_cst 0x7ffff6543f90 constant 32>
            unit-size <integer_cst 0x7ffff6543fa8 constant 4>
            align:32 warn_if_not_align:0 symtab:0 alias-set 2 canonical-type
0x7ffff66a52a0 precision:32 min <integer_cst 0x7ffff6562438 -2147483648> max
<integer_cst 0x7ffff66b0570 2147483647>>
        BLK
        size <integer_cst 0x7ffff65620c0 constant 256>
        unit-size <integer_cst 0x7ffff65621b0 constant 32>
        align:256 warn_if_not_align:0 symtab:0 alias-set 2 canonical-type
0x7ffff66a5348 nunits:8>
    visited
    def_stmt _12 = (vector(8) signed int) foo_v256u8_0.0_1;
    version:12>
$2 = void
(gdb) p debug_tree (gimple_assign_rhs1 (stmt))
 <ssa_name 0x7ffff654ed80
    type <vector_type 0x7ffff667abd0
        type <integer_type 0x7ffff655e690 unsigned int public unsigned SI
            size <integer_cst 0x7ffff6543f90 constant 32>
            unit-size <integer_cst 0x7ffff6543fa8 constant 4>
            align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff655e690 precision:32 min <integer_cst 0x7ffff6543fc0 0> max <integer_cst
0x7ffff6543f78 4294967295>
            pointer_to_this <pointer_type 0x7ffff656e1f8>>
        unsigned BLK
        size <integer_cst 0x7ffff65620c0 constant 256>
        unit-size <integer_cst 0x7ffff65621b0 constant 32>
        align:256 warn_if_not_align:0 symtab:0 alias-set 1 canonical-type
0x7ffff667abd0 nunits:8
        pointer_to_this <pointer_type 0x7ffff66a51f8>>
    visited
    def_stmt foo_v256u8_0.0_1 = foo_v256u8_0;
    version:1>
$3 = void

the issue is that we introduce a non-V_C_E vector conversion in
forwprop3 that is not supported by the target and vector lowering is
not supporting lowering of such conversions.

/* Fold -((int)x >> (prec - 1)) into (unsigned)x >> (prec - 1).  */
(simplify
 (negate (convert? (rshift @0 INTEGER_CST@1)))
 (if (tree_nop_conversion_p (type, TREE_TYPE (@0))
      && wi::to_wide (@1) == element_precision (type) - 1)
  (with { tree stype = TREE_TYPE (@0);
          tree ntype = TYPE_UNSIGNED (stype) ? signed_type_for (stype)
                                             : unsigned_type_for (stype); }
   (convert (rshift:ntype (convert:ntype @0) @1)))))

is the pattern responsible.  For vector types this needs to check availability
of (convert ...) support or alternatively use VIEW_CONVERT.

[Bug rtl-optimization/103485] [12 Regression] ICE: in emit_move_insn, at expr.c:4013 with -O -fno-tree-dce -fno-tree-dse

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103485

--- Comment #2 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:e2521cd2d26661cfcfceaabf9bd281ef316fd3fc

commit r12-5593-ge2521cd2d26661cfcfceaabf9bd281ef316fd3fc
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Nov 30 08:19:24 2021 +0100

    middle-end/103485 - fix conversion kind for vectors

    This makes sure to use a VIEW_CONVERT_EXPR for converting
    vector signedness in the -((int)x >> (prec - 1)) to (unsigned)x >> (prec -
1)
    simplification.

    2021-11-30  Richard Biener  <rguenther@suse.de>

            PR middle-end/103485
            * match.pd (-((int)x >> (prec - 1)) to (unsigned)x >> (prec - 1)):
            Use VIEW_CONVERT_EXPR for vectors.

            * gcc.dg/pr103485.c: New testcase.

[Bug rtl-optimization/103485] [12 Regression] ICE: in emit_move_insn, at expr.c:4013 with -O -fno-tree-dce -fno-tree-dse

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103485

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

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

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.