CVS-19981209: Patch for "unsigned HOST_WIDE_INT" on Solaris2.6

[Jeffrey A Law]

  In message < 199812160612.WAA01282@shade.twinsun.com >you write:
  > OK, you convinced me.  I assume that the ``less intrusive change'' is
  > to add new macros HOST_WIDEST_INT and unsigned_HOST_WIDEST_INT that
  > expand to `long long' and `unsigned long long' (or perhaps to intmax_t
  > and uintmax_t :-), and to have cpp and any other interested parties
  > use these macros?
Yea.  Something along those lines.  autoconf is our friend.

It also gives us the ability to make certain optimizations less pessimistic
if we're so inclined -- there are *some* places in the optimization phases
that would benefit from having access to HOST_WIDEST_INT or whatever we
call it.


  > While we're on the subject, it's too bad that gcc has to use slow
  > double-precision routines like add_double on hosts that have perfectly
  > serviceable longer ints, but I guess that's life.
This might be one of the places that could benefit from HOST_WIDEST_INT :-)

jeff