[Bug tree-optimization/114068] New: [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[Bug tree-optimization/114068] New: [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

            Bug ID: 114068
           Summary: [14 regression] ICE when building darktable-4.6.1
                    (error: PHI node with wrong VUSE on edge from BB 25)
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sjames at gcc dot gnu.org
  Target Milestone: ---

Created attachment 57505
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57505&action=edit
RawImage.cpp.ii.xz

Initially reported downstream in Gentoo by tdr.

```
# /usr/bin/x86_64-pc-linux-gnu-g++ -DGDK_DISABLE_DEPRECATED
-DGTK_DISABLE_DEPRECATED -DGTK_DISABLE_SINGLE_INCLUDES -DNDEBUG -D_RELEASE
-D_XOPEN_SOURCE=700
-I/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1_build/lib64/darktable/rawspeed/src/librawspeed/common
-I/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common
-I/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1_build/lib64/darktable/rawspeed/src
-I/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/..
-isystem
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/external
 -march=native -O3 -pipe -mavx -msse3 -Wall -Wformat -Wformat-security -Wshadow
-Wtype-limits -Wvla -Wmaybe-uninitialized -Wno-unknown-pragmas
-Wno-error=varargs -Wno-format-truncation -Wno-error=address-of-packed-member
-w -Wall -Wextra -Wcast-qual -Wextra -Wextra-semi -Wformat=2 -Wpointer-arith
-Wvla -Wmissing-format-attribute -Wsuggest-attribute=format
-Wno-unused-parameter -Wno-maybe-uninitialized -Wno-stringop-overflow
-Wno-array-bounds -Wstack-usage=4096 -Wframe-larger-than=4096
-Wlarger-than=32768  -O3 -std=c++20 -fPIC -fvisibility=hidden
-fvisibility-inlines-hidden   -march=native -g3 -ggdb3 -fopenmp -MD -MT
lib64/darktable/rawspeed/src/librawspeed/common/CMakeFiles/rawspeed_common.dir/RawImage.cpp.o
-MF
lib64/darktable/rawspeed/src/librawspeed/common/CMakeFiles/rawspeed_common.dir/RawImage.cpp.o.d
-o
lib64/darktable/rawspeed/src/librawspeed/common/CMakeFiles/rawspeed_common.dir/RawImage.cpp.o
-c
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/RawImage.cpp 
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/RawImage.cpp:
In member function ‘void
rawspeed::RawImageData::clearArea(rawspeed::iRectangle2D)’:
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/RawImage.cpp:325:6:
error: PHI node with wrong VUSE on edge from BB 25
  325 | void RawImageData::clearArea(iRectangle2D area) {
      |      ^~~~~~~~~~~~
.MEM_201 = PHI <.MEM_162(25)>
expected .MEM_155
during GIMPLE pass: vect
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/RawImage.cpp:325:6:
internal compiler error: verify_ssa failed
0x562cb4452725 verify_ssa(bool, bool)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/tree-ssa.cc:1203

/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/common/RawImage.cpp:325:6:
internal compiler error: Segmentation fault
0x562cb4155596 crash_signal
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/toplev.cc:317
0x7fa03922e07f ???
       
/var/tmp/portage/sys-libs/glibc-2.38-r11/work/glibc-2.38/signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0
0x562cb5a2f12f x86_64_fallback_frame_state
        ./md-unwind-support.h:63
0x562cb5a2f12f uw_frame_state_for
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/libgcc/unwind-dw2.c:1013
0x562cb5a300e5 _Unwind_Backtrace
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/libgcc/unwind.inc:303
0x562cb56fe80b backtrace_full
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/libbacktrace/backtrace.c:127
0x562cb562138b diagnostic_context::action_after_output(diagnostic_t)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/diagnostic.cc:781
0x562cb55feb6b diagnostic_action_after_output(diagnostic_context*,
diagnostic_t)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/diagnostic.h:1002
0x562cb55feb6b diagnostic_context::report_diagnostic(diagnostic_info*)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/diagnostic.cc:1633
0x562cb55fefb0 diagnostic_impl
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/diagnostic.cc:1767
0x562cb55ff2b9 internal_error(char const*, ...)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/diagnostic.cc:2225
0x562cb4452725 verify_ssa(bool, bool)
       
/usr/src/debug/sys-devel/gcc-14.0.1_pre20240218/gcc-14-20240218/gcc/tree-ssa.cc:1203
{standard input}: Assembler messages:
{standard input}: Error: open CFI at the end of file; missing .cfi_endproc
directive
x86_64-pc-linux-gnu-g++: internal compiler error: Segmentation fault signal
terminated program cc1plus
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
See <https://bugs.gentoo.org/> for instructions.
```

I can reproduce with just g++ -c RawImage.cpp.ii -std=c++20 -O3 -march=znver2,
but for the original reporter, native was -march=sapphirerapids -mabm -mno-sgx
-mrtm -mshstk --param=l1-cache-line-size=64 --param=l1-cache-size=48
--param=l2-cache-size=76800.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #1 from Sam James <sjames at gcc dot gnu.org> ---
Reducing now...

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #2 from Sam James <sjames at gcc dot gnu.org> ---
Note that this is the second time we've seen the weird "double crash" w/
x86_64_fallback_frame_state when unwinding, see
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114048#c3 too.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Sam James from comment #2)
> Note that this is the second time we've seen the weird "double crash" w/
> x86_64_fallback_frame_state when unwinding, see
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114048#c3 too.

See PR 66874 for that ... I think.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
`-std=c++20 -O3 -mavx` is enough to reproduce the ICE.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|                            |2024-02-23
   Target Milestone|---                         |14.0

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  Waiting for reduced testcase.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[14 regression] ICE when    |[14 regression] ICE when
                   |building darktable-4.6.1    |building darktable-4.6.1
                   |(error: PHI node with wrong |(error: PHI node with wrong
                   |VUSE on edge from BB 25)    |VUSE on edge from BB 25)
                   |                            |since r14-8768
           Priority|P3                          |P1
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |tnfchris at gcc dot gnu.org

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r14-8768-g85094e2aa6dba7908f053046f02dd443e8f65d72

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57508
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57508&action=edit
Reduced a lot

This could be reduced more though.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #57508|0                           |1
        is obsolete|                            |

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57509
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57509&action=edit
Little further reduced

I am done reducing this for tonight, if someone wants to take over to manually
reduce it further that would be nice.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #9 from Sam James <sjames at gcc dot gnu.org> ---
```
template <class T> struct Array1DRef {
  T data;
  T *addressOf(int eltIdx) { return &data + eltIdx; }
};
template <class T> struct CroppedArray1DRef {
  Array1DRef<T> base;
  int numElts;
  void getAsArray1DRef() {
    if (numElts)
      __builtin_unreachable();
  }
  T &operator()(int eltIdx) {
    if (eltIdx > numElts)
      __builtin_unreachable();
    return *base.addressOf(eltIdx);
  }
};
struct Array2DRef {
  CroppedArray1DRef<int> __trans_tmp_2;
  void operator[](int) { __trans_tmp_2.getAsArray1DRef(); }
} clearArea_base0;
CroppedArray1DRef<short> __trans_tmp_1;
int clearArea_y, clearArea_posx;
void clearArea() {
  Array2DRef base = clearArea_base0;
  for (int x = clearArea_posx; x; ++x) {
    base.operator[](clearArea_y);
    __trans_tmp_1(x) = 0;
  }
}
```

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #10 from Sam James <sjames at gcc dot gnu.org> ---
last one:
```
struct c {
  int b;
  int *d(int e) { return &b + e; }
};
struct g {
  c base;
  int f;
  void j() {
    if (f)
      __builtin_unreachable();
  }
  int &p(int e) {
    if (e > 1)
      __builtin_unreachable();
    return *base.d(e);
  }
} l;
struct h {
  g i;
  void q() { i.j(); }
} k;
int m;
void n() {
  h a = k;
  for (int o = m; o; ++o) {
    a.q();
    l.p(o) = 1;
  }
}
```

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

Tamar Christina <tnfchris at gcc dot gnu.org> changed:

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

--- Comment #11 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
reduced to

---
struct h {
  int b;
  int f;
} k;

void n(int m) {
  struct h a = k;
  for (int o = m; o; ++o) {
    if (a.f)
      __builtin_unreachable();
    if (o > 1)
      __builtin_unreachable();
    *(&k.b + o) = 1;
  }
}
---

which fails on aarch64 too

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #12 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
looks like the moving of the store didn't update a stray out of block use of
the MEM.

working on patch.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #13 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
Created attachment 57510
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57510&action=edit
candidate-patch1.patch

candidate patch being tested.

I was hoping to correct it during peeling itself when the merge block is
created, but I don't have enough information there to do so.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #14 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
patch submitted
https://gcc.gnu.org/pipermail/gcc-patches/2024-February/646415.html

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

--- Comment #15 from GCC 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:8293df8019adfffae3384cb6fb9cb6f496fe8608

commit r14-9181-g8293df8019adfffae3384cb6fb9cb6f496fe8608
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Feb 26 11:25:50 2024 +0100

    tree-optimization/114068 - missed virtual LC PHI after vect peeling

    When we choose the IV exit to be one leading to no virtual use we
    fail to have a virtual LC PHI even though we need it for the epilog
    entry.  The following makes sure to create it so that later updating
    works.

            PR tree-optimization/114068
            * tree-vect-loop-manip.cc (get_live_virtual_operand_on_edge):
            New function.
            (slpeel_tree_duplicate_loop_to_edge_cfg): Add a virtual LC PHI
            on the main exit if needed.  Remove band-aid for the case
            it was missing.

            * gcc.dg/vect/vect-early-break_118-pr114068.c: New testcase.
            * gcc.dg/vect/vect-early-break_119-pr114068.c: Likewise.

[Bug tree-optimization/114068] [14 regression] ICE when building darktable-4.6.1 (error: PHI node with wrong VUSE on edge from BB 25) since r14-8768

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114068

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

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

--- Comment #16 from Richard Biener <rguenth at gcc dot gnu.org> ---
Should be fixed.