[Bug tree-optimization/110117] New: ICE on valid code at -O1 with "": verify_ssa failed

[Bug tree-optimization/110117] New: ICE on valid code at -O1 with "": verify_ssa failed

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

            Bug ID: 110117
           Summary: ICE on valid code at -O1 with "": verify_ssa failed
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

It appears to be a recent regression.

Compiler Explorer: https://godbolt.org/z/fsxExbW9P

[584] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/home/suz/suz-local/software/local/gcc-trunk/bin/../libexec/gcc/x86_64-pc-linux-gnu/14.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror
--disable-multilib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 14.0.0 20230604 (experimental) [master r14-924-gd709841ae0f] (GCC)
[585] %
[585] % gcctk -O1 -ftree-vrp -fno-tree-ccp -fno-tree-forwprop small.c
during RTL pass: expand
small.c: In function ‘main’:
small.c:9:12: internal compiler error: in as_a, at machmode.h:381
    9 |     c = ~h - (-g & f || e);
      |         ~~~^~~~~~~~~~~~~~~
0x7564c4 scalar_int_mode as_a<scalar_int_mode>(machine_mode)
        ../../gcc-trunk/gcc/machmode.h:381
0x758382 scalar_int_mode as_a<scalar_int_mode>(machine_mode)
        ../../gcc-trunk/gcc/tree.h:3669
0x758382 expand_single_bit_test
        ../../gcc-trunk/gcc/expr.cc:12914
0x758382 do_store_flag
        ../../gcc-trunk/gcc/expr.cc:13173
0xbbd259 expand_expr_real_2(separate_ops*, rtx_def*, machine_mode,
expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:10251
0xbc4ccd expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:10800
0xbc7324 expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-trunk/gcc/expr.h:310
0xbc7324 expand_operands(tree_node*, tree_node*, rtx_def*, rtx_def**,
rtx_def**, expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:8577
0xbbcd0a expand_expr_real_2(separate_ops*, rtx_def*, machine_mode,
expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:10591
0xbc4ccd expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:10800
0xbbd57a expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:8999
0xbbd57a expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-trunk/gcc/expr.h:310
0xbbd57a expand_expr_real_2(separate_ops*, rtx_def*, machine_mode,
expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:9334
0xbc4ccd expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:10800
0xbc733c expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-trunk/gcc/expr.h:310
0xbc733c expand_operands(tree_node*, tree_node*, rtx_def*, rtx_def**,
rtx_def**, expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:8578
0xbbcd0a expand_expr_real_2(separate_ops*, rtx_def*, machine_mode,
expand_modifier)
        ../../gcc-trunk/gcc/expr.cc:10591
0xbc4ccd expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:10800
0xbc185e expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-trunk/gcc/expr.cc:8999
0xbc185e expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-trunk/gcc/expr.h:310
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.
[586] %
[586] % cat small.c
int a, b, d;
unsigned c;
int main() {
  char e = -10;
  int f = 1, g = 0;
  if (a) {
    char h = e;
  i:
    c = ~h - (-g & f || e);
    int j = b % c;
    g = j % 9;
    if (c) {
      if (d)
        e = 0;
      while (!g)
        ;
      int k = 0;
    l:
      if (k)
        goto i;
    }
  }
  if (e > -10) {
    if (g)
      f = 0;
    goto l;
  }
  return 0;
}

[Bug tree-optimization/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Target Milestone|---                         |14.0
            Summary|ICE on valid code at -O1    |[14 Regression] ICE on
                   |with "-ftree-vrp            |valid code at -O1 with
                   |-fno-tree-ccp               |"-ftree-vrp -fno-tree-ccp
                   |-fno-tree-forwprop": in     |-fno-tree-forwprop": in
                   |as_a, at machmode.h:381     |as_a, at machmode.h:381
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|                            |2023-06-04

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Mine.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
  _4 = 0;
  _5 = _4 & 1;
  _39 = _5 != 0;


The reason why it worked before was that when creating trees, it simplify down
to 0 and then expand that. Now we are expanding directly to rtl, we get:
(const_int 0) for inner and there is no mode associated with that.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Anyways imode should just be operand_mode really.

But that does not solve the issue either because extract_bit_field does not
know how to handle const_int :(.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This was "fixed" by r14-1534-g908e5ab5c11c64 as we don't have a nonzero bits on
_5 in :

  # RANGE [irange] int [-2147483647, +INF]
  _4 = 0;
  _5 = _4 & 1;
  _39 = _5 != 0;

Since we don't have a non-zero bits we don't look back at the TER to see it was
&1 and we don't call into expand_single_bit_test .

I am still going to fix expand_single_bit_test to do the correct thing for
const_int but we no longer ICEs in the original testcase.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 55255
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55255&action=edit
Patch which fixes this

I am going to split this patch into 2, one for the do_store_flag change and one
for the patch which fixes this part. The do_store_flag change is needed to hit
this bug again.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Simplier testcase:
```
int f()
{
  int t = 0;
  return (t & 1) != 0;
}
```
Options are:
`-O1 -fno-tree-dominator-opts -fno-tree-vrp -fno-tree-ccp -fno-tree-forwprop
-fno-tree-fre -fno-tree-copy-prop` .

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch
                URL|                            |https://gcc.gnu.org/piperma
                   |                            |il/gcc-patches/2023-June/62
                   |                            |0618.html

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Patch submitted here:
https://gcc.gnu.org/pipermail/gcc-patches/2023-June/620618.html

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:e60593f3881c72a96a3fa4844d73e8a2cd14f670

commit r14-1600-ge60593f3881c72a96a3fa4844d73e8a2cd14f670
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sun Jun 4 19:21:05 2023 -0700

    Improve do_store_flag for single bit when there is no non-zero bits

    In r14-1534-g908e5ab5c11c, I forgot you could turn off CCP or
    turn off the bit tracking part of CCP so we would lose out
    what TER was able to do before hand. This moves around the
    TER code so that it is used instead of just the nonzerobits.
    It also makes it easier to remove the TER part of the code
    later on too.

    OK? Bootstrapped and tested on x86_64-linux-gnu.

    Note it reintroduces PR 110117 (which was accidently fixed after
    r14-1534-g908e5ab5c11c). The next patch in series will fix that.

    gcc/ChangeLog:

            * expr.cc (do_store_flag): Rearrange the
            TER code so that it overrides the nonzero bits
            info if we had `a & POW2`.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:3f085e45755643f13d4fa45a12a6ade45be98f95

commit r14-1601-g3f085e45755643f13d4fa45a12a6ade45be98f95
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sun Jun 4 19:42:08 2023 -0700

    Handle const_int in expand_single_bit_test

    After expanding directly to rtl instead of
    creating a tree, we could end up with
    a const_int which is not ready to be handled
    by extract_bit_field.
    So need to the constant folding here instead.

    OK? bootstrapped and tested on x86_64-linux-gnu with no regressions.

            PR middle-end/110117

    gcc/ChangeLog:

            * expr.cc (expand_single_bit_test): Handle
            const_int from expand_expr.

    gcc/testsuite/ChangeLog:

            * gcc.dg/pr110117-1.c: New test.
            * gcc.dg/pr110117-2.c: New test.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

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

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

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed.

[Bug middle-end/110117] [14 Regression] ICE on valid code at -O1 with "-ftree-vrp -fno-tree-ccp -fno-tree-forwprop": in as_a, at machmode.h:381

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110117

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |19373742 at buaa dot edu.cn

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 110272 has been marked as a duplicate of this bug. ***