[Bug middle-end/111082] New: ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[Bug middle-end/111082] New: ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[manuel.lauss at googlemail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

            Bug ID: 111082
           Summary: ICE in expand_fn_using_insn, at internal-fn.cc:214
                    (AVX512 related)
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: manuel.lauss at googlemail dot com
  Target Milestone: ---

Created attachment 55766
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55766&action=edit
compressed unreduced testcase

gcc version 14.0.0 20230820 (experimental)
d77c280454cfba48ef38357145cecdabc8c1b05c

# gcc -O2 -march=znver4 -mtune=znver4 -pipe -std=gnu++11  gcc14-bf.ii
In file included from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/panodata/PanoramaData.h:40,
                 from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/nona/Stitcher.h:51,
                 from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/nona/Stitcher1.cpp:26:
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/panodata/PanoramaVariable.h:160:31:
warning: 'template<class _Arg, class _Result> struct std::unary_function' is
deprecated [-Wdeprecated-declarations]
  160 | struct PrintVar : public std::unary_function<Variable, void>
      |                               ^~~~~~~~~~~~~~
In file included from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/string:49,
                 from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/bits/locale_classes.h:40,
                 from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/bits/ios_base.h:41,
                 from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/ios:44,
                 from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/istream:40,
                 from
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/sstream:40,
                 from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/nona/Stitcher.h:32:
/usr/lib/gcc/x86_64-pc-linux-gnu/14/include/g++-v14/bits/stl_function.h:120:12:
note: declared here
  120 |     struct unary_function
      |            ^~~~~~~~~~~~~~
during RTL pass: expand
In file included from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/vigra_ext/StitchWatershed.h:30,
                 from
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/nona/Stitcher.h:47:
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/vigra_ext/BlendPoisson.h:
In function 'void vigra_ext::poisson::BuildSeamPyramid(const Image&,
vigra::ImagePyramid<PyramidImage>&, int) [with Image =
vigra::BasicImage<unsigned char>; PyramidImage = vigra::BasicImage<signed
char>]':
/tmp-ram/portage/media-gfx/hugin-2022.0.0/work/hugin-2022.0.0/src/hugin_base/vigra_ext/BlendPoisson.h:650:6:
internal compiler error: in expand_fn_using_insn, at internal-fn.cc:214
  650 | void BuildSeamPyramid(const Image& input,
vigra::ImagePyramid<PyramidImage>& seams, const int minLength)
      |      ^~~~~~~~~~~~~~~~
0x5652df7942af internal_error(char const*, ...)
        ???:0
0x5652ddd04ed6 fancy_abort(char const*, int, char const*)
        ???:0


ICE goes away with -mno-avx512f appended.

Attached us an unreduced testcase.  This failure is fairly new; it did not
break 1 week ago.

[Bug middle-end/111082] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[manuel.lauss at googlemail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

--- Comment #1 from Manuel Lauss <manuel.lauss at googlemail dot com> ---
Created attachment 55767
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55767&action=edit
another compressed unreduced testcase

Another testcase.  it triggers with "-march=zvner1" but not e.g. "core2" but
that is probably due to the source being preprocessed for znver4.

# g++ -march=znver1 -mtune=znver1 -O2 -c qxmlstream.ii 
during RTL pass: expand
In file included from
/tmp-ram/portage/dev-qt/qtcore-5.15.10-r1/work/qtbase-everywhere-src-5.15.10/src/corelib/serialization/qxmlstream.cpp:76:
/tmp-ram/portage/dev-qt/qtcore-5.15.10-r1/work/qtbase-everywhere-src-5.15.10/src/corelib/serialization/qxmlstream_p.h:
In member function 'QStringView QXmlStreamReaderPrivate::symView(int) const':
/tmp-ram/portage/dev-qt/qtcore-5.15.10-r1/work/qtbase-everywhere-src-5.15.10/src/corelib/serialization/qxmlstream_p.h:949:17:
internal compiler error: in expand_fn_using_insn, at internal-fn.cc:214
  949 |     QStringView symView(int index) const
      |                 ^~~~~~~
0x555774d492af internal_error(char const*, ...)
        ???:0
0x5557732b9ed6 fancy_abort(char const*, int, char const*)

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|ICE in                      |[14 Regression] ICE in
                   |expand_fn_using_insn, at    |expand_fn_using_insn, at
                   |internal-fn.cc:214 (AVX512  |internal-fn.cc:214 (AVX512
                   |related)                    |related)
           Keywords|                            |ice-on-valid-code
            Version|unknown                     |14.0
   Target Milestone|---                         |14.0

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reducing ...

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-08-20
             Status|UNCONFIRMED                 |NEW
             Target|                            |x86_64-linux-gnu
     Ever confirmed|0                           |1
           Keywords|                            |needs-bisection

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reduced testcase at `-O3 -mavx512f`:
```
long minarray2(const long *input) {
  if (input[0] < input[1])
    return input[0] ;
  return input[1];
}
```

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
I think that's mine.

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Interestingly we have CODE_FOR_reduc_smin_scal_v2di at
vectorizable_bb_reduc_epilogue time where we check

          && (!reduction_fn_for_scalar_code (reduc_code, &reduc_fn)
              || reduc_fn == IFN_LAST
              || !direct_internal_fn_supported_p (reduc_fn,
                                                  vectype, OPTIMIZE_FOR_BOTH)))

which passes.  But at that point we have a signed MIN.  I have a patch.

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

--- Comment #6 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:e10cb804e658dbd1e9d58f528f3985362e4e72e7

commit r14-3354-ge10cb804e658dbd1e9d58f528f3985362e4e72e7
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Aug 21 11:07:18 2023 +0200

    tree-optimization/111082 - bogus promoted min

    vectorize_slp_instance_root_stmt promotes operations with undefined
    overflow to unsigned arithmetic but fails to consider operations
    that do not overflow like MIN which it turned into MIN with wrong
    signedness and in the case of the PR an unsupported operation.
    The following rectifies this.

            PR tree-optimization/111082
            * tree-vect-slp.cc (vectorize_slp_instance_root_stmt): Only
            pun operations that can overflow.

            * gcc.dg/pr111082.c: New testcase.

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.

[Bug middle-end/111082] [14 Regression] ICE in expand_fn_using_insn, at internal-fn.cc:214 (AVX512 related)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111082

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dcb314 at hotmail dot com

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 111092 has been marked as a duplicate of this bug. ***