[Simh] Implementing a floating point math accelerator simulator

[Bob Supnik]

There's an IEEE-compliant floating point simulator in the Alpha code 
base. However, it only implements what Alpha required, in particular:

- No gradual underflow.
- No 80b floating point (not needed to produce IEEE accurate results for 
single- and double-precision).

You might find John Hauser's SoftFloat to be of more interest, as it is 
very general and implements every possible useless goodie of the 
standard as is existed in the 20th century. It does not have the fifth 
rounding mode that was added more recently, but the chips of that era 
wouldn't have it either.

Bob Supnik