[Bug d/110193] New: d_signed_or_unsigned_type is invoked for vector types

[Bug d/110193] New: d_signed_or_unsigned_type is invoked for vector types

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

            Bug ID: 110193
           Summary: d_signed_or_unsigned_type is invoked for vector types
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: d
          Assignee: ibuclaw at gdcproject dot org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

It looks like this function (via a langhook) is invoked on VECTOR_TYPE and
the implementation assumes it works on integer types only:

  if (TYPE_PRECISION (type) == TYPE_PRECISION (d_cent_type))
    return unsignedp ? d_ucent_type : d_cent_type;

  if (TYPE_PRECISION (type) == TYPE_PRECISION (d_long_type))
    return unsignedp ? d_ulong_type : d_long_type;
...

TYPE_PRECISION for vectors is TYPE_VECTOR_SUBPARTS so in this case
for any V8mode vector we'd return d_{u,}cent_type.  The C familiy
frontend similar routine does

  type1 = TYPE_MAIN_VARIANT (type);
  if (type1 == signed_char_type_node || type1 == char_type_node || type1 ==
unsigned_char_type_node)
    return unsignedp ? unsigned_char_type_node : signed_char_type_node;
  if (type1 == integer_type_node || type1 == unsigned_type_node)
    return unsignedp ? unsigned_type_node : integer_type_node;

...

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2023-06-09
           Keywords|                            |wrong-code

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

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[ibuclaw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

ibuclaw at gcc dot gnu.org changed:

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

--- Comment #2 from ibuclaw at gcc dot gnu.org ---
Just having a cursory look, the only references to d_signed_or_unsigned_type is
via d_unsigned_type and d_signed_type.

The former is only called by d_gimplify_unsigned_rshift_expr.

So I guess you must mean you've noticed this with something like.

    __vector(int[4]) unsigned_rshift(__vector(int[4]) v)
    {
        return v >>> 2;
    }

It turns out that this accident doesn't affect the intent on x86.

https://d.godbolt.org/z/99fK7fWaW

Gimple dump:
---
    __vector(int[4]) rshift (__vector(int[4]) v)
    {
      __vector(int[4]) D.1795;

      _1 = VIEW_CONVERT_EXPR<vector(4) unsigned int>(v);
      _2 = _1 >> 2;
      D.1795 = VIEW_CONVERT_EXPR<__vector(int[4])>(_2);
      return D.1795;
    }

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[ibuclaw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

--- Comment #3 from ibuclaw at gcc dot gnu.org ---
(In reply to ibuclaw from comment #2)
> Gimple dump:
> ---
>     __vector(int[4]) rshift (__vector(int[4]) v)
>     {
>       __vector(int[4]) D.1795;
> 
>       _1 = VIEW_CONVERT_EXPR<vector(4) unsigned int>(v);
>       _2 = _1 >> 2;
>       D.1795 = VIEW_CONVERT_EXPR<__vector(int[4])>(_2);
>       return D.1795;
>     }
I've just realized that I wasn't reading this code dump properly.  Everything
looks correct here.

I've a look again and picked at random vector E_V8HImode, that has a precision
of 3.  Whereas a cent has a precision of 128.

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[ibuclaw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

--- Comment #4 from ibuclaw at gcc dot gnu.org ---
A shortcut to signed_or_unsigned_type_for for vector types seems reasonable
nonetheless.

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
I noticed this by instrumenting TYPE_PRECISION and ICEing when applied on
VECTOR_TYPEs.

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Iain Buclaw <ibuclaw@gcc.gnu.org>:

https://gcc.gnu.org/g:9757e4440bd8755d327601a60a73d57d712583ed

commit r14-2168-g9757e4440bd8755d327601a60a73d57d712583ed
Author: Iain Buclaw <ibuclaw@gdcproject.org>
Date:   Wed Jun 28 17:38:16 2023 +0200

    d: Fix d_signed_or_unsigned_type is invoked for vector types (PR110193)

    This function can be invoked on VECTOR_TYPE, but the implementation
    assumes it works on integer types only.  To fix, added a check whether
    the type passed is any `__vector(T)' or non-integral type, and return
    early by calling `signed_or_unsigned_type_for()' instead.

    Problem was found by instrumenting TYPE_PRECISION and ICEing when
    applied on VECTOR_TYPEs.

            PR d/110193

    gcc/d/ChangeLog:

            * types.cc (d_signed_or_unsigned_type): Handle being called with
any
            vector or non-integral type.

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[ibuclaw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

ibuclaw at gcc dot gnu.org changed:

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

--- Comment #7 from ibuclaw at gcc dot gnu.org ---
Fix committed.

[Bug d/110193] d_signed_or_unsigned_type is invoked for vector types

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110193

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

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

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
*** Bug 110480 has been marked as a duplicate of this bug. ***