i=n2
i+1
j=n1
p
j
g
i
+
j=n1
p
j
c
0
(2)
where p
i
= x
i
+y
i
and g
i
= x
i
y
i
. This is appropriate (for
delay reduction), however, only when AND and OR happen to
be the basic gates (for example, in CMOS technology). If one
were to use this approach in QCA, the AND and OR terms have
to be realized by majority logic, leading to poor performance
(with respect to area and delay). This is exemplied by
c
8
=(g
7
+p
7
g
6
+p
7
p
6
g
5
+p
7
p
6
p
5
g
4
)
+ [(p
7
p
6
p
5
p
4
)(g
3
+p
3
g
2
+p
3
p
2
g
1
+p
3
p
2
p
1
g
0
)]
+ (p
7
p
6
p
5
p
4
)(p
3
p
2
p
1
p
0
)c
0
. (3)
The c
8
computation requires a delay of (at least) seven ma-
jority gates (since g
i
and p
i
require one majority gate each)
assuming two-input AND and OR gates. Furthermore, the logic
requirement is high. The proposed formulation which forms
the basis of the new decomposition theorems facilitates the
computation of c
n
with less delay (by reducing the major-
ity gates in the critical path) in comparison to direct real-
ization of (2). It relies on two properties of majority logic
(named Properties 1 and 2). The proofs are direct and hence
omitted.
Property 1: If f
1
, f
2
, and f
3
are three binary inputs, then
M(f
1
, f
2
, f
3
) = M(f
1
, f
2
, 0) +M(f
1
, f
2
, 1) f
3
.
An immediate consequence of Property 1 is M(x
2
, y
2
,
M(x
1
, y
1
, M(x
0
, y
0
, z
0
))) = g
2
+p
2
g
1
+p
2
p
1
g
0
+p
2
p
1
p
0
z
0
,
where all the quantities are binary inputs.
Property 2: If f
1
, f
2
, and f
3
are three binary inputs such
that f
1
and f
2
satisfy f
1
f
2
= f
1
and f
1
+f
2
= f
2
, then
M(f
1
, f
2
, f
3
) = f
1
+f
2
f
3
.
Let g
i
=x
i
y
i
=M(x
i
, y
i
, 0) and p
i
=x
i
+y
i
=M(x
i
, y
i
, 1).
The direct calculation of g
i
+p
i
c
i
requires two majority
gates. However, in viewof Property 2, we can obtain g
i
+p
i
c
i
using only one majority gate as M(g
i
, p
i
, c
i
).
Theorem 1 describes the overall delay calculation using the
leading carry (c
n
) since other carries incur less or equal delay.
For a compact description of Theorem 1, we introduce some
additional notation. In particular, we dene two quantities,
denoted by K(i, j) and L(i, j), as shown in (4) and (5), shown
at the bottom of the page. These quantities will also be used to
present Theorem 2.
M (x
n1
, y
n1
, M (x
n2
, y
n2
, . . . , M (x
1
, y
1
, M(x
0
, y
0
, c
0
)) ))
= M (x
n1
, y
n1
, M(x
n2
, y
n2
, . . . , M(x
i
, y
i
, 0)) )
+p
n1
p
n2
p
i+1
p
i
M (x
i1
, y
i1
, M (x
i2
, y
i2
, . . . , M (x
1
, y
1
, M(x
0
, y
0
, c
0
)) )) , where 0<i <n
c
n
=M (x
n1
, y
n1
, M (x
n2
, y
n2
, . . . , M(x
1
, y
1
, M(x
0
, y
0
, c
0
)) ))
=g
n1
+p
n1
g
n2
+p
n1
p
n2
g
n3
+ +p
n1
p
n2
p
i+2
(g
i+1
+p
i+1
g
i
)
+p
n1
p
n2
p
i+1
p
i
[g
i1
+p
i1
g
i2
+ +p
i1
p
i2
p
1
(g
0
+p
0
c
0
)] [from (2)]
=g
n1
+p
n1
g
n2
+p
n1
p
n2
g
n3
+ +p
n1
p
n2
p
i+3
(g
i+2
+p
i+2
M(x
i+1
, y
i+1
, g
i
))
+p
n1
p
n2
p
i+1
p
i
[g
i1
+p
i1
g
i2
+ +p
i1
p
i2
p
2
(g
1
+p
1
M(x
0
, y
0
, c
0
))] (from Property 1)
=g
n1
+p
n1
g
n2
+p
n1
p
n2
g
n3
+ +p
n1
p
n2
p
i+4
(g
i+3
+p
i+3
M (x
i+2
, y
i+2
, M(x
i+1
, y
i+1
, g
i
)))
+p
n1
p
n2
p
i+1
p
i
[g
i1
+p
i1
g
i2
+ +p
i1
p
i2
p
3
(g
2
+p
2
M (x
1
, y
1
, M(x
0
, y
0
, c
0
)))]
L.H.S. =M (x
n1
, y
n1
, M (x
n2
, y
n2
, . . . , M (x
i+1
, y
i+1
, M(x
i
, y
i
, 0)) ))
+p
n1
p
n2
p
i+1
p
i
M (x
i1
, y
i1
, M (x
i2
, y
i2
, . . . , M (x
1
, y
1
, M(x
0
, y
0
, c
0
)) ))
K(i, j) =
M (x
i
, y
i
, M (x
i1
, y
i1
, . . . , M(x
j
, y
j
, 0)) )) , for 0 j < i
g
i
= M(x
i
, y
i
, 0), for i = j
(4)
L(i, j) =
M(x
i
, y
i
, 1)M(x
i1
, y
i1
, 1) M(x
j
, y
j
, 1), for 0 j < i
p
i
= M(x
i
, y
i
, 1), for i = j
(5)
680 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMSII: EXPRESS BRIEFS, VOL. 59, NO. 10, OCTOBER 2012
The statement of Theorem 1 can be rewritten using (4) and
(5). In particular, c
n
is given by
c
n
=K(n 1, i) +L(n 1, i)c
i
[since c
i
=M(x
i1
, y
i1
,
M(x
i2
, y
i2
, . . . , M(x
0
, y
0
, c
0
)) ))]
K(n1, i)=K(n1, j)+L(n1, j)K(j1, i),
for i < j < n. (6)
We now use (6) to state Theorem 2.
Theorem 2: Let x
0
, x
1
, . . . , x
n2
, x
n1
, y
0
, y
1
, . . . , y
n2
,
y
n1
, and c
0
be 2n + 1 binary inputs. Furthermore, let c
i
be as
dened in Theorem 1. Then, K(n 1, i) +L(n 1, i) c
i
=
M(K(n 1, i), p
n1
, L(n 2, i) c
i
).
Proof:
K(n 1, i) =M (x
n1
, y
n1
,
M(x
n2
, y
n2
, . . . ,
M (x
i+1
, y
i+1
, M(x
i
, y
i
, 0)) ))
=M(x
n1
, y
n1
, 0) +M(x
n1
, y
n1
, 1)
M (x
n2
, y
n2
, . . . ,
M (x
i+1
, y
i+1
, M(x
i
, y
i
, 0)) )
=g
n1
+p
n1
K(n 2, i)
p
n1
=M(x
n1
, y
n1
, 1).
Hence,
K(n 1, i) p
n1
= [g
n1
+p
n1
K(n 2, i)] p
n1
=g
n1
p
n1
+p
n1
p
n1
K(n 2, i)
=g
n1
+p
n1
K(n 2, i)
=K(n 1, i)
K(n 1, i) +p
n1
=g
n1
+p
n1
K(n 2, i) +p
n1
=p
n1
+p
n1
K(n 2, i) = p
n1
.
K(n 1, i) and p
n1
satisfy the two conditions of Property 2.
Therefore, from Property 2, we have
M (K(n 1, i), p
n1
, L(n 2, i) c
i
)
= K(n 1, i) +p
n1
L(n 2, i)c
i
= K(n 1, i) +L(n 1, i) c
i
[since L(n 1, i) = p
n1
L(n 2, i)] .