[Bug middle-end/114099] New: [14 regression] ICE when building darktable-4.6.1

[Bug middle-end/114099] New: [14 regression] ICE when building darktable-4.6.1

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

            Bug ID: 114099
           Summary: [14 regression] ICE when building darktable-4.6.1
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sjames at gcc dot gnu.org
  Target Milestone: ---

Created attachment 57525
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57525&action=edit
PentaxDecompressor.cpp.ii

This was found after applying the patch from PR114068. I can hit it without too
though.

Initially reported downstream in Gentoo by tdr.

```
$ gcc -c PentaxDecompressor.cpp.ii -O3 -march=sapphirerapids -std=c++20
during GIMPLE pass: vect
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/decompressors/PentaxDecompressor.cpp:
In static member function ‘static
rawspeed::HuffmanCode<rawspeed::BaselineCodeTag>
rawspeed::PentaxDecompressor::SetupPrefixCodeDecoder_Modern(rawspeed::ByteStream)’:
/var/tmp/portage/media-gfx/darktable-4.6.1/work/darktable-4.6.1/src/external/rawspeed/src/librawspeed/decompressors/PentaxDecompressor.cpp:81:1:
internal compiler error: Segmentation fault
0x55743ad30b10 crash_signal
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/toplev.cc:319
0x55743bb50b1f find_uses_to_rename_use
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-ssa-loop-manip.cc:414
0x55743bb50b1f find_uses_to_rename_bb(basic_block_def*, bitmap_head**,
bitmap_head*, int) [clone .constprop.0]
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-ssa-loop-manip.cc:489
0x55743b993aed find_uses_to_rename
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-ssa-loop-manip.cc:521
0x55743b993aed rewrite_into_loop_closed_ssa_1
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-ssa-loop-manip.cc:588
0x55743b993aed rewrite_into_loop_closed_ssa(bitmap_head*, unsigned int)
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-ssa-loop-manip.cc:628
0x55743bd1e8d9 execute
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/tree-vectorizer.cc:1360
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://bugs.gentoo.org/> for instructions.
```

[Bug middle-end/114099] [14 regression] ICE when building darktable-4.6.1

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

--- Comment #1 from Sam James <sjames at gcc dot gnu.org> ---
Created attachment 57526
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57526&action=edit
reduced.ii

Attached reduced version from tdr too. I can reproduce with `gcc -c
PentaxDecompressor.cpp.ii -O3 -mavx -std=c++20`.

[Bug middle-end/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #57526|0                           |1
        is obsolete|                            |

--- Comment #2 from Sam James <sjames at gcc dot gnu.org> ---
Created attachment 57527
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57527&action=edit
reduced.ii

[Bug middle-end/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note the reduced testcase is undefined code.
c::j is not initialized when c::k() is called.

[Bug middle-end/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57532
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57532&action=edit
Reduced further but still has undefined code in it

[Bug middle-end/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #57527|0                           |1
        is obsolete|                            |
  Attachment #57532|0                           |1
        is obsolete|                            |

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 57533
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57533&action=edit
Reduced all the way without undefined code in it

This ICEs on aarch64 with just -O3.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-02-25
           Keywords|                            |needs-bisection
             Status|UNCONFIRMED                 |NEW
          Component|middle-end                  |tree-optimization
     Ever confirmed|0                           |1
   Target Milestone|---                         |14.0

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tnfchris at gcc dot gnu.org

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note the difference between noreturn vs __builtin_unreachable BB is the
noreturn bb has:
  # _21 = PHI <_2(4)>
  m = _21;

But the one with unreachable does not.
It feels like that difference is causing the issue here. Replacing unreachable
with another noreturn function removes the ICE even.

I am 99% sure this is related to the early out vect changes too.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

--- Comment #8 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
Created attachment 57537
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57537&action=edit
uses.patch

new code seems sensitive to visitation order as get_virtual_phi returns NULL
for blocks which don't define or use a virtual PHI.

testing patch.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

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 #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Let me handle this as well.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

--- Comment #10 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:fb68e2cac1283f731a3a979cb714621afb1ddfcc

commit r14-9182-gfb68e2cac1283f731a3a979cb714621afb1ddfcc
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Feb 26 12:27:42 2024 +0100

    tree-optimization/114099 - virtual LC PHIs and early exit vect

    In some cases exits can lack LC PHI nodes for the virtual operand.
    We have to create them when the epilog loop requires them which also
    allows us to remove some only halfway correct fixups.  This is the
    variant triggering for alternate exits.

            PR tree-optimization/114099
            * tree-vect-loop-manip.cc (slpeel_tree_duplicate_loop_to_edge_cfg):
            Create and fill in a needed virtual LC PHI for the alternate
            exits.  Remove code dealing with that missing.

            * gcc.dg/vect/vect-early-break_120-pr114099.c: New testcase.

[Bug tree-optimization/114099] [14 regression] ICE in find_uses_to_rename_use when building darktable-4.6.1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114099

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

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

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.