[Bug middle-end/83653] [6/7/8 Regression] GCC fails to remove a can't-happen call on ia64

[law at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83653

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
                 CC|                            |law at redhat dot com
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=72785
         Resolution|---                         |INVALID

--- Comment #10 from Jeffrey A. Law <law at redhat dot com> ---
So look for something similar to this:

 <bb 29> [local count: 96151367]:
  # i_280 = PHI <0(28), 0(30)>
  if (i_280 < nr_300)
    goto <bb 81>; [92.50%]
  else
    goto <bb 40>; [7.50%]

The subscripts may change, but I think that's what you're looking for.

i_280 will be replaced with zero resulting in

if (0 < nr_300)
  true
else
  false


And because nr is unsigned that will turn into an equality comparison

if (nr_300 != 0)
  true
else
  false

At that point the compiler knows that nr_300 has the value 0 on the false arm
of the conditional.


Anyway, this isn't a GCC bug.  I'll note that it may be worth reviewing pr72785
 which is essentially the same issue with misunderstanding how
builtin_constant_p works.