[Bug tree-optimization/113459] New: ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[Bug tree-optimization/113459] New: ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

            Bug ID: 113459
           Summary: ICE: in as_a, at machmode.h:381 with memset() on a
                    _BitInt() at -O1 and above
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
                CC: jakub at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 57119
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57119&action=edit
reduced testcase

Compiler output:
$ cc1 -quiet -O testcase.c 
during GIMPLE pass: fre
testcase.c: In function 'foo':
testcase.c:8:1: internal compiler error: in as_a, at machmode.h:381
    8 | }
      | ^
0x876149 scalar_int_mode as_a<scalar_int_mode>(machine_mode)
        /repo/gcc-trunk/gcc/machmode.h:381
0x876149 scalar_int_mode as_a<scalar_int_mode>(machine_mode)
        /repo/gcc-trunk/gcc/machmode.h:379
0x876149 vn_reference_lookup_3
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:2972
0x16226f7 walk_non_aliased_vuses(ao_ref*, tree_node*, bool, void* (*)(ao_ref*,
tree_node*, void*), void* (*)(ao_ref*, tree_node*, void*, translate_flags*),
tree_node* (*)(tree_node*), unsigned int&, void*)
        /repo/gcc-trunk/gcc/tree-ssa-alias.cc:3914
0x171a4ef vn_reference_lookup(tree_node*, tree_node*, vn_lookup_kind,
vn_reference_s**, bool, tree_node**, tree_node*, bool)
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:4023
0x171e7e7 visit_reference_op_load
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:5740
0x171e7e7 visit_stmt
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:6261
0x171ee6b process_bb
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:8025
0x1720846 do_rpo_vn_1
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:8626
0x17223bb execute
        /repo/gcc-trunk/gcc/tree-ssa-sccvn.cc:8787
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.

$ xgcc -v
Using built-in specs.
COLLECT_GCC=/repo/build-gcc-trunk-amd64A/gcc/xgcc
Target: x86_64-pc-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-valgrind-annotations --disable-nls --enable-checking=yes,rtl,df,extra
--disable-bootstrap --with-cloog --with-ppl --with-isl
--build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=x86_64-pc-linux-gnu --with-ld=/usr/bin/x86_64-pc-linux-gnu-ld
--with-as=/usr/bin/x86_64-pc-linux-gnu-as --disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r14-8193-20240117105849-g3340878009a-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.1 20240117 (experimental) (GCC)

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
   Last reconfirmed|                            |2024-01-18
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
We're relying on INTEGRAL_TYPE_P (vr->type) -> vr->type has integer mode.

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

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

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

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
              unsigned buflen = TREE_INT_CST_LOW (TYPE_SIZE_UNIT (vr->type)) +
1;
              if (INTEGRAL_TYPE_P (vr->type))
                buflen = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (vr->type)) + 1;

there's other spots using the pattern

                      if (INTEGRAL_TYPE_P (type))
                        sz = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (type));

I wonder when GET_MODE_SIZE differs from TYPE_SIZE_UNIT?  PSImode?
Packed bitfields?  r10-6885-g5f9cd512c42786 added the INTEGRAL_TYPE_P
special-casing we now run into with the TYPE_SIZE_UNIT code being there
before.

Jakub, do you remember?

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #2)
>               unsigned buflen = TREE_INT_CST_LOW (TYPE_SIZE_UNIT (vr->type))
> + 1;
>               if (INTEGRAL_TYPE_P (vr->type))
>                 buflen = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (vr->type)) + 1;
> 
> there's other spots using the pattern
> 
>                       if (INTEGRAL_TYPE_P (type))
>                         sz = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (type));
> 
> I wonder when GET_MODE_SIZE differs from TYPE_SIZE_UNIT?  PSImode?

I'm afraid I don't know either.

> Packed bitfields?  r10-6885-g5f9cd512c42786 added the INTEGRAL_TYPE_P
> special-casing we now run into with the TYPE_SIZE_UNIT code being there
> before.
> 
> Jakub, do you remember?

I bet the above comes from what the native_{encode,interpret}_int has been
doing (and which has been tweaked for BITINT_TYPE).
If we don't want to throw it away, we could just change it to
if (INTEGRAL_TYPE_P (...) && TYPE_MODE (...) != BLKmode)

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

--- Comment #4 from rguenther at suse dot de <rguenther at suse dot de> ---
On Thu, 18 Jan 2024, jakub at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459
> 
> --- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
> (In reply to Richard Biener from comment #2)
> >               unsigned buflen = TREE_INT_CST_LOW (TYPE_SIZE_UNIT (vr->type))
> > + 1;
> >               if (INTEGRAL_TYPE_P (vr->type))
> >                 buflen = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (vr->type)) + 1;
> > 
> > there's other spots using the pattern
> > 
> >                       if (INTEGRAL_TYPE_P (type))
> >                         sz = GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (type));
> > 
> > I wonder when GET_MODE_SIZE differs from TYPE_SIZE_UNIT?  PSImode?
> 
> I'm afraid I don't know either.
> 
> > Packed bitfields?  r10-6885-g5f9cd512c42786 added the INTEGRAL_TYPE_P
> > special-casing we now run into with the TYPE_SIZE_UNIT code being there
> > before.
> > 
> > Jakub, do you remember?
> 
> I bet the above comes from what the native_{encode,interpret}_int has been
> doing (and which has been tweaked for BITINT_TYPE).
> If we don't want to throw it away, we could just change it to
> if (INTEGRAL_TYPE_P (...) && TYPE_MODE (...) != BLKmode)

Yeah, that might work.  There are a few occurances in tree-ssa-sccvn.cc
that need such adjustment.

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

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 #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57137
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57137&action=edit
gcc14-pr113459.patch

Untested fix.

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

--- Comment #6 from GCC 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:dcd5601c2b7298155c9a8e1bfb93ee8e952eca0b

commit r14-8276-gdcd5601c2b7298155c9a8e1bfb93ee8e952eca0b
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Fri Jan 19 10:00:16 2024 +0100

    sccvn: Don't use SCALAR_INT_TYPE_MODE on BLKmode BITINT_TYPEs [PR113459]

    sccvn uses GET_MODE_SIZE (SCALAR_INT_TYPE_MODE (type)) for INTEGER_TYPEs,
    most likely because that is what native_{interpret,encode}_int used.
    This obviously doesn't work for larger BITINT_TYPEs which have BLKmode
    and the above ICEs on those.  native_{interpret,encode}_int checks whether
    the BITINT_TYPE is medium/large/huge (i.e. an array of 2+ ABI limbs)
    and uses TYPE_SIZE_UNIT for that case, otherwise SCALAR_INT_TYPE_MODE like
    for the INTEGER_TYPE case.

    The following patch instead just uses SCALAR_INT_TYPE_MODE for non-BLKmode
    TYPE_MODE and TYPE_SIZE_UNIT otherwise.

    2024-01-19  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/113459
            * tree-ssa-sccvn.cc (vn_walk_cb_data::push_partial_def): Use
            TREE_INT_CST_LOW of TYPE_SIZE_UNIT rather than GET_MODE_SIZE
            of SCALAR_INT_TYPE_MODE if type has BLKmode.
            (vn_reference_lookup_3): Likewise.  Formatting fix.

            * gcc.dg/bitint-73.c: New test.

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

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

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

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.

[Bug tree-optimization/113459] ICE: in as_a, at machmode.h:381 with memset() on a _BitInt() at -O1 and above

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113459

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0