‹header›

‹date/time›

Click to edit Master text styles

Second level

Third level

Fourth level

Fifth level

‹footer›

‹#›

In this talk I will present the software
system we have developed at UC Berkeley called MVSIS for multi-valued logic
optimization.

Present the MVSIS group, which include the following members.

1.Time allocation:

1.Motivation ¼

2.Specification and
optimization ½

3.Examples and demo
¼

We identified immediate application in
these four areas.

1.Current design
process;

2.Alternative design
flow using multi-valued logic;

3.Previous work and why
they did not succeed;

1.Optimization
hard;

2.Encoding hard

4.MVSIS uniqueness and
contribution.

1.Synchronous
programming as in the embedded system design industry;

2.Design flow for
Esterel;

3.Previous work and what
is lacking;

4.MVSIS
contribution.

FIX THE REFERENCE FOR BERRY!

1.New design paradigm
and its benefits;

2.It’s basic building
blocks;

This shows NMOS only and the same applies for PMOS.

This shows NMOS only and the same applies for PMOS.

3.This paper shows the
design of a fast multiplier. 50% faster than binary CMOS under the same
voltage and power consumption.

1.Synchronous hardware
computation model;

2.Boolean algebra;
however, it can be easily converted into a multi-valued algebra by imposing a
priority ordering among the different values.

1.Literal: Boolean
function and evaluates to 1 if the variable takes on value 0, or 2;

2.Cube: Boolean product
of a set of literals and evaluates to 1 if all literals evaluate to 1.

3.SOP and MDD

1. Exdc output always binary.

The optimization sequence
is:

1.Derive the
kernel/cubes;

2.Divide it into the
original expressions;

3.Substitute into other
expressions;

Different flavors of commands
developed around this concept, like in SIS.

Define: non-determinism is the
specification allows the same input minterm to produce different values at the
primary output.

Define: don’t care is a minterm
that is not specified and the output is allowed to produce any value.

Value of a part/node: the cost
associated with this part/node, if eliminated, the cost being literal count or
cube count.

1.New design paradigm
and its benefits;

2.It’s basic building
blocks;

This shows NMOS only and the same applies for PMOS.

This shows NMOS only and the same applies for PMOS.

3.This paper shows the
design of a fast multiplier. 50% faster than binary CMOS under the same
voltage and power consumption.