[Bug c/106030] New: ice in emit_move_insn, at expr.cc:4026

[Bug c/106030] New: ice in emit_move_insn, at expr.cc:4026

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

            Bug ID: 106030
           Summary: ice in emit_move_insn, at expr.cc:4026
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dcb314 at hotmail dot com
  Target Milestone: ---

For this C code:

int safe_mul_func_int8_t_s_s_si2, g_7014, g_11941_l_12290;
char(safe_mul_func_uint8_t_u_u)(ui1, ui2) { return ui1 * ui2; }
void g_11941() {
  char __trans_tmp_1 =
      (safe_mul_func_uint8_t_u_u <= g_7014) * safe_mul_func_int8_t_s_s_si2;
  g_11941_l_12290 = safe_mul_func_uint8_t_u_u(2 != g_11941, __trans_tmp_1);
}

can be compiled like this:

$ /home/dcb/gcc/results/bin/gcc -v
Using built-in specs.
COLLECT_GCC=/home/dcb/gcc/results/bin/gcc
COLLECT_LTO_WRAPPER=/home/dcb/gcc/results.20220617/libexec/gcc/arm-linux-gnueabihf/13.0.0/lto-wrapper
Target: arm-linux-gnueabihf
Configured with: ../trunk/configure --prefix=/home/dcb/gcc/results.20220617
--disable-bootstrap --disable-multilib --disable-werror
--with-pkgversion=1d6044c250e3badf --enable-checking=yes
--enable-languages=c,c++ --with-cpu=cortex-a72 --with-fpu=neon-fp-armv8
--with-float=hard --build=arm-linux-gnueabihf --host=arm-linux-gnueabihf
--target=arm-linux-gnueabihf
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 13.0.0 20220617 (experimental) (1d6044c250e3badf) 
$ /home/dcb/gcc/results/bin/gcc -c  -w q.i
$

but switch on -O1 and this happens:

$ /home/dcb/gcc/results/bin/gcc -c  -w -O1 q.i 
during RTL pass: expand
q.i: In function 'g_11941':
q.i:6:19: internal compiler error: in emit_move_insn, at expr.cc:4026
    6 |   g_11941_l_12290 = safe_mul_func_uint8_t_u_u(2 != g_11941,
__trans_tmp_
1);
      |  
~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~
0x551897 emit_move_insn(rtx_def*, rtx_def*)
        /home/dcb/gcc/working/gcc/../../trunk/gcc/expr.cc:4025
0x54608f expand_and(machine_mode, rtx_def*, rtx_def*, rtx_def*)
        /home/dcb/gcc/working/gcc/../../trunk/gcc/expmed.cc:5482

[Bug c/106030] ice in emit_move_insn, at expr.cc:4026

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #1 from David Binderman <dcb314 at hotmail dot com> ---
I should mention that this code is fine on x86_64, but breaks on arm-32 native
and cross.

[Bug target/106030] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Deobfuscated:

int a, b, c;

char
foo (int x, int y)
{
  return x * y;
}

void
bar (void)
{
  char d = (foo <= b) * a;
  c = foo (2 != bar, d);
}

[Bug middle-end/106030] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
                 CC|                            |jakub at gcc dot gnu.org
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-06-20
          Component|target                      |middle-end

[Bug middle-end/106030] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

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

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

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 53168
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53168&action=edit
gcc13-pr106030.patch

Untested fix.  But it is unclear when it started...

[Bug middle-end/106030] ice in emit_move_insn, at expr.cc:4026

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #4 from Alex Coplan <acoplan at gcc dot gnu.org> ---
Started with r13-673-gd863ba23fb16122bb0547b0c678173be0d98f43c :

commit d863ba23fb16122bb0547b0c678173be0d98f43c
Author: Roger Sayle <roger@nextmovesoftware.com>
Date:   Thu May 19 17:54:38 2022 +0100

    PR middle-end/98865: Expand X*Y as X&-Y when Y is [0,1].

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
            Summary|ice in emit_move_insn, at   |[13 Regression] ice in
                   |expr.cc:4026                |emit_move_insn, at
                   |                            |expr.cc:4026
   Target Milestone|---                         |13.0

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Still it is latent on the release branches.

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #6 from David Binderman <dcb314 at hotmail dot com> ---

Around 22,000 executions of csmith per 24 hours and feeding the results
to trunk gcc are proving their value in flushing out bugs in gcc.

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #7 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:2df1df945fac85d7b3d084001414a66a2709d8fe

commit r13-1185-g2df1df945fac85d7b3d084001414a66a2709d8fe
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Jun 21 11:38:59 2022 +0200

    expand: Fix up expand_cond_expr_using_cmove [PR106030]

    If expand_cond_expr_using_cmove can't find a cmove optab for a particular
    mode, it tries to promote the mode and perform the cmove in the promoted
    mode.

    The testcase in the patch ICEs on arm because in that case we pass temp
which
    has the promoted mode (SImode) as target to expand_operands where the
    operands have the non-promoted mode (QImode).
    Later on the function uses paradoxical subregs:
      if (GET_MODE (op1) != mode)
        op1 = gen_lowpart (mode, op1);

      if (GET_MODE (op2) != mode)
        op2 = gen_lowpart (mode, op2);
    to change the operand modes.

    The following patch fixes it by passing NULL_RTX as target if it has
    promoted mode.

    2022-06-21  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/106030
            * expr.cc (expand_cond_expr_using_cmove): Pass NULL_RTX instead of
            temp to expand_operands if mode has been promoted.

            * gcc.c-torture/compile/pr106030.c: New test.

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:d068623e5b109e635e2ec2acfcf15e7c50c7f15c

commit r12-8522-gd068623e5b109e635e2ec2acfcf15e7c50c7f15c
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Jun 21 11:38:59 2022 +0200

    expand: Fix up expand_cond_expr_using_cmove [PR106030]

    If expand_cond_expr_using_cmove can't find a cmove optab for a particular
    mode, it tries to promote the mode and perform the cmove in the promoted
    mode.

    The testcase in the patch ICEs on arm because in that case we pass temp
which
    has the promoted mode (SImode) as target to expand_operands where the
    operands have the non-promoted mode (QImode).
    Later on the function uses paradoxical subregs:
      if (GET_MODE (op1) != mode)
        op1 = gen_lowpart (mode, op1);

      if (GET_MODE (op2) != mode)
        op2 = gen_lowpart (mode, op2);
    to change the operand modes.

    The following patch fixes it by passing NULL_RTX as target if it has
    promoted mode.

    2022-06-21  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/106030
            * expr.cc (expand_cond_expr_using_cmove): Pass NULL_RTX instead of
            temp to expand_operands if mode has been promoted.

            * gcc.c-torture/compile/pr106030.c: New test.

    (cherry picked from commit 2df1df945fac85d7b3d084001414a66a2709d8fe)

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

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

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

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:d9b72d8d03134813d5fa84cffb190027fd29f6d6

commit r11-10353-gd9b72d8d03134813d5fa84cffb190027fd29f6d6
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Jun 21 11:38:59 2022 +0200

    expand: Fix up expand_cond_expr_using_cmove [PR106030]

    If expand_cond_expr_using_cmove can't find a cmove optab for a particular
    mode, it tries to promote the mode and perform the cmove in the promoted
    mode.

    The testcase in the patch ICEs on arm because in that case we pass temp
which
    has the promoted mode (SImode) as target to expand_operands where the
    operands have the non-promoted mode (QImode).
    Later on the function uses paradoxical subregs:
      if (GET_MODE (op1) != mode)
        op1 = gen_lowpart (mode, op1);

      if (GET_MODE (op2) != mode)
        op2 = gen_lowpart (mode, op2);
    to change the operand modes.

    The following patch fixes it by passing NULL_RTX as target if it has
    promoted mode.

    2022-06-21  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/106030
            * expr.c (expand_cond_expr_using_cmove): Pass NULL_RTX instead of
            temp to expand_operands if mode has been promoted.

            * gcc.c-torture/compile/pr106030.c: New test.

    (cherry picked from commit 2df1df945fac85d7b3d084001414a66a2709d8fe)

[Bug middle-end/106030] [13 Regression] ice in emit_move_insn, at expr.cc:4026

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106030

--- Comment #11 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:89741783f6e7127d4d8be5a5e8e4fb7d5785f8a6

commit r10-11332-g89741783f6e7127d4d8be5a5e8e4fb7d5785f8a6
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Jun 21 11:38:59 2022 +0200

    expand: Fix up expand_cond_expr_using_cmove [PR106030]

    If expand_cond_expr_using_cmove can't find a cmove optab for a particular
    mode, it tries to promote the mode and perform the cmove in the promoted
    mode.

    The testcase in the patch ICEs on arm because in that case we pass temp
which
    has the promoted mode (SImode) as target to expand_operands where the
    operands have the non-promoted mode (QImode).
    Later on the function uses paradoxical subregs:
      if (GET_MODE (op1) != mode)
        op1 = gen_lowpart (mode, op1);

      if (GET_MODE (op2) != mode)
        op2 = gen_lowpart (mode, op2);
    to change the operand modes.

    The following patch fixes it by passing NULL_RTX as target if it has
    promoted mode.

    2022-06-21  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/106030
            * expr.c (expand_cond_expr_using_cmove): Pass NULL_RTX instead of
            temp to expand_operands if mode has been promoted.

            * gcc.c-torture/compile/pr106030.c: New test.

    (cherry picked from commit 2df1df945fac85d7b3d084001414a66a2709d8fe)