[Bug middle-end/94935] New: Failure to emit call to absvsi2 for __builtin_abs with -ftrapv

[Bug middle-end/94935] New: Failure to emit call to absvsi2 for __builtin_abs with -ftrapv

[gabravier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94935

            Bug ID: 94935
           Summary: Failure to emit call to absvsi2 for __builtin_abs with
                    -ftrapv
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gabravier at gmail dot com
  Target Milestone: ---

unsigned f(int x)
{
    return __builtin_abs(x);
}

This should emit a call to __absvsi2, not get "inlined" into a call to
__subvsi3

[Bug middle-end/94935] Failure to emit call to absvsi2 for __builtin_abs with -ftrapv

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94935

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |x86_64-*-*
   Last reconfirmed|                            |2020-05-04
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
           Keywords|                            |wrong-code

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  We're expanding via

rtx
expand_abs_nojump (machine_mode mode, rtx op0, rtx target,
                   int result_unsignedp)
{
...
  /* If this machine has expensive jumps, we can do integer absolute
     value of X as (((signed) x >> (W-1)) ^ x) - ((signed) x >> (W-1)),
     where W is the width of MODE.  */

since the machine has no absvsi2 pattern.  It smells like a correctness
issue (though you don't explicitely say so?)