source: vis_dev/vis-2.1/examples/fpmpy/README @ 15

$Id: README,v 1.1 1997/03/01 18:54:43 fabio Exp $

This directory contains a simple floating point multiplier,
not exactly IEEE 754-compliant, but largely inspired to the standard.

The significand uses the hidden bit and is between 1 (included) and 2
(excluded).
The exponent uses the excess (2**(n-1) - 1) representation. For single
precision, this is excess 127.
The smallest exponent (0) is used for the represenation of 0. Denormals
are not supported.
The largest exponent is used for infinities and NaNs. Infinities use
the smallest possible significand (all zeroes). Everything else is deemed
a NaN. No distinction is made between signalling and non-signalling NaNs.
When the multiplier generates a NaN, it uses the all-one significand.
One multiplication takes three clock cycles and it is not pipelined.

The CTL formulae check for several properties of the result and
demonstrate the use of vector variables and macros.

Author: Fabio Somenzi <Fabio@Colorado.EDU>