[Bug tree-optimization/18892] missed optimization with & and ==

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-11 07:19 -------
(In reply to comment #5)
> (In reply to comment #4)
> > when I compile this program with mainline.  Isnt this what you claimed it should
> > be compiled to?  or are you claiming it should be optimized to 'return 0'?
> 
> I am claiming it should be compiled to "return 0".  The full testcase which is closer to
> what shows up in GCC is:
> 
> void f(int a)
> {
>   int i = a & -129;
>   if (i == 144)
>     link_error ();
> }

The testcase works but for the wrong reason (we call fold for COND_EXPR after out of ssa because
of tree_cleanup_cfg).

Here is a testcase which fails though:

void g(int)  __attribute__((noinline);
void g(int a) { a+=2; }

void f(int a)
{
  int i = a & -129;
  g(i);
  if (i == 144)
    link_error ();
}

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18892