[sv-dc] 4-state real

From: Marq Kole <marq.kole@nxp.com>
Date: Fri Dec 03 2010 - 01:48:47 PST

Dear all,

After reading through the minutes of the last meeting, I couldn't help thinking that we already have a 4-state real. In fact the reals that we have today as IEEE 754-2008 allow multiple states to be defined. According to the standard (see also http://en.wikipedia.org/wiki/IEEE_754-2008) there already three types of values:
- finite numbers (including 0 and 1)
- Infinity (+Inf, -Inf)
- Not-a-Number (NaN)

I would expect that for modeling and verification purposes Z can be mapped to +Inf and X can be mapped on NaN. Already the most important properties of doing real arithmetic with NaNs nicely follows that of doing binary arithmetic with X's. Only the real arithmetic with Inf differs from the binary arithmetic with Z's. However, this is something where the SV real could choose to differ from an IEEE 754-2008 real.

The reference of the IEEE 754 standard in the IEEE 1800-2009 standard does not exclude this: IEEE 754 is only used as a reference for notation and representation, no requirements on the handling of arithmetic are specified. In this effect, the arithmetic operators are already a kind of resolution functions that apply to combining reals and other types of operands - the extension that is required is linking these arithmetic resolution functions to the driver resolutions on wires.

Even if it is useful to maintain Inf and NaN values as real values that can be transferred on a wire, there are multiple NaN patterns that can be used to map X and Z values and that do not get in the way with the numerical or symbolic values (see also http://en.wikipedia.org/wiki/NaN).

Cheers,
Marq

Marq Kole
Product Manager AMSRF Simulation
NXP Semiconductors / Central R&D / Foundation Technology

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
Received on Fri Dec 3 01:49:44 2010

This archive was generated by hypermail 2.1.8 : Fri Dec 03 2010 - 01:49:47 PST