Bug in multiprecision arithmetic, libmp

[muha%bbn-unix at sri-unix.UUCP]

From:  Ralph Muha <muha at bbn-unix>

Bingo!  I just tried your calculations on our machine, a BBN C/70,
which has 20 bit words and the uses the 11/70 version of libmp
(15 bits to a word).  I got the correct results.  So whatever
the problem, look for differences between the 11 and VAX versions.
Or maybe a compiler bug???