[Bug tree-optimization/113024] New: Nested cast not optimized out in GIMPLE

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113024

            Bug ID: 113024
           Summary: Nested cast not optimized out in GIMPLE
           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: jakub at gcc dot gnu.org
                CC: jakub at gcc dot gnu.org, unassigned at gcc dot gnu.org,
                    zsojka at seznam dot cz
        Depends on: 112941
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

+++ This bug was initially created as a clone of Bug #112941 +++

The above PR leads to a question, why don't we optimize:
unsigned int
foo (signed char x)
{
  unsigned long long y = x;
  return y;
}

unsigned int
bar (signed char x)
{
  return (unsigned long long) x;
}

We do optimize the bar case in convert.cc's
          CASE_CONVERT:
            {
              tree argtype = TREE_TYPE (TREE_OPERAND (expr, 0));
              /* Don't introduce a "can't convert between vector values
                 of different size" error.  */
              if (TREE_CODE (argtype) == VECTOR_TYPE
                  && maybe_ne (GET_MODE_SIZE (TYPE_MODE (argtype)),
                               GET_MODE_SIZE (TYPE_MODE (type))))
                break;
            }
            /* If truncating after truncating, might as well do all at once.
               If truncating after extending, we may get rid of wasted work. 
*/
            return convert (type, get_unwidened (TREE_OPERAND (expr, 0),
type));
This is the truncating after extending case.
Now, for
unsigned int
baz (unsigned char x)
{
  unsigned long long y = x;
  return y;
}
we optimize it in the match.pd
    /* Likewise, if the intermediate and initial types are either both
       float or both integer, we don't need the middle conversion if the
       former is wider than the latter and doesn't change the signedness
       (for integers).  Avoid this if the final type is a pointer since
       then we sometimes need the middle conversion.  */
    (if (((inter_int && inside_int) || (inter_float && inside_float))
         && (final_int || final_float)
         && inter_prec >= inside_prec
         && (inter_float || inter_unsignedp == inside_unsignedp))
     (ocvt @0))
case.

In the foo case, we have inter_int && inside_int && final_int && inside_prec <
final_prec && final_prec < inter_prec && (!inside_unsignedp) && inter_unsignedp
&& final_unsignedp, so don't trigger the above condition.
Slightly later, there is a comment which describes the reason why 2 conversions
would be needed:
    /* Two conversions in a row are not needed unless:
        - some conversion is floating-point (overstrict for now), or
        - some conversion is a vector (overstrict for now), or
        - the intermediate type is narrower than both initial and
          final, or
        - the intermediate type and innermost type differ in signedness,
          and the outermost type is wider than the intermediate, or
        - the initial type is a pointer type and the precisions of the
          intermediate and final types differ, or
        - the final type is a pointer type and the precisions of the
          initial and intermediate types differ.  */
and I believe none of those bullets apply here, while intermediate type and
innermost type differ in signedness, the outermost type is not wider than the
intermediate.
But the actually implemented rule below it has 7 cases rather than 6.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112941
[Bug 112941] during GIMPLE pass: bitintlower ICE: in handle_operand_addr, at
gimple-lower-bitint.cc:2126 (gimple-lower-bitint.cc:2134) at -O with _BitInt()