Re: Tighten checking of vector comparisons

[Richard Biener <richard dot guenther at gmail dot com>]

On Thu, Nov 22, 2012 at 11:59 AM, Marc Glisse <marc.glisse@inria.fr> wrote:
> (I forgot to send this at the time)
>
>
> On Sun, 4 Nov 2012, Richard Biener wrote:
>
>>> -         else if (!INTEGRAL_TYPE_P (type) && TREE_CODE (type) !=
>>> VECTOR_TYPE)
>>> +         else if (!INTEGRAL_TYPE_P (type) && !VOID_TYPE_P (type)
>>> +                  && TREE_CODE (type) != VECTOR_TYPE)
>
> [...]
>>
>> Ok for this part.
>
>
> Applied, thanks.
>
>
>>> Index: gcc/tree-cfg.c
>>> ===================================================================
>>> --- gcc/tree-cfg.c      (revision 193060)
>>> +++ gcc/tree-cfg.c      (working copy)
>>> @@ -3263,21 +3263,30 @@ verify_gimple_comparison (tree type, tre
>>>        error ("mismatching comparison operand types");
>>>        debug_generic_expr (op0_type);
>>>        debug_generic_expr (op1_type);
>>>        return true;
>>>      }
>>>
>>>    /* The resulting type of a comparison may be an effective boolean
>>> type.
>>> */
>>>    if (INTEGRAL_TYPE_P (type)
>>>        && (TREE_CODE (type) == BOOLEAN_TYPE
>>>           || TYPE_PRECISION (type) == 1))
>>> -    ;
>>> +    {
>>> +      if (TREE_CODE (op0_type) == VECTOR_TYPE
>>> +         || TREE_CODE (op1_type) == VECTOR_TYPE)
>>> +        {
>>> +          error ("vector comparison returning a boolean");
>>> +          debug_generic_expr (op0_type);
>>> +          debug_generic_expr (op1_type);
>>> +          return true;
>>> +        }
>>
>>
>> verify_gimple_* should have "positive" checks, thus, check that
>> if there are vector operands the comparison result should be a
>> vector.  Not complaining about a vector comparison having a
>> boolean result.
>
>
> I wasn't sure what that was supposed to look like, so I dropped it for now.

Ok, looking closer we have

  /* The resulting type of a comparison may be an effective boolean type.  */
  if (INTEGRAL_TYPE_P (type)
      && (TREE_CODE (type) == BOOLEAN_TYPE
          || TYPE_PRECISION (type) == 1))
    ;
  /* Or an integer vector type with the same size and element count
     as the comparison operand types.  */
  else if (TREE_CODE (type) == VECTOR_TYPE
           && TREE_CODE (TREE_TYPE (type)) == INTEGER_TYPE)
    {
...

Thus you are rejecting a boolean valued vector comparison which we
otherwise happily accept.  I suppose that makes sense (even though
at least equality compares can make sense).

Thus that hunk is ok as well.

Thanks,
Richard.

> --
> Marc Glisse