[Bug tree-optimization/126538] Compile time hog with gimple-range-phi and bitint

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126538

--- Comment #3 from Andrew Macleod <amacleod at redhat dot com> ---
(In reply to Jakub Jelinek from comment #2)
> sizeof (long) doesn't look like a good idea, that would make the compilation
> dependent on the host size of long, 64-bit cc1 would compile something
> different from 32-bit cc1 targetting the same target.  Did you mean size of
> the target long (or wordsize or something similar)?

Yeah, just some relevant cap for the target that will cover typical ints. 
Suggestions welcome :-)    This never affects correctness, just gives PHIs a
better range if the modifier converges during the iteration.