Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
11
,
No.
1
,
Febr
uar
y
2021
, pp.
545
~
557
IS
S
N: 20
88
-
8708, DO
I: 10
.11
591/ijece
.v1
1i1
.
pp
545
-
557
545
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
Optimi
zed a
rchitectu
re fo
r SNOW
3G
N.
B.
H
ull
e
1
,
Prathi
ba B
2
,
S
arika R
.
K
hop
e
3
, K
.
A
nur
adha
4
,
Y
og
ini
Bo
role
5
,
D
.
Kota
mbka
r
6
1,2,3
,5
Depa
rtment
of
Elec
tron
ic
s
an
d
Telec
om
m
unicati
on
,
G
H
R
ai
so
ni
Instit
u
te of En
gine
er
ing
and
T
e
chnol
og
y
,
Indi
a
4
Depa
rtment of
Com
pute
r
Scie
n
ce
and Engi
ne
ering
,
Gokar
aj
u
R
a
ngar
aj
u
Instit
u
te
of
Engi
n
ee
rin
g
a
nd
Technol
og
y
,
India
6
Depa
rtment of
El
e
ct
roni
c
s
Desi
gn
Technol
og
y
,
Shri
Ramdeoba
b
a
Col
le
ge
of Eng
ine
er
ing
and
Ma
nage
m
ent
,
Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ve
d
A
pr
3
, 2
0
20
Re
vised
Jun
19
,
20
20
Accepte
d Aug
1
1
, 20
20
SN
O
W
3G
is
a
s
y
nchr
onous,
w
ord
-
orie
nt
ed
stre
am
ci
pher
used
b
y
th
e
3GP
P
standa
rds
as
a
c
onfide
ntial
ity
an
d
int
egr
i
t
y
al
gor
it
hm
s
.
It
is
used
as
first
set
in
long
te
rm
evol
uti
on
(LTE)
and
as
a
sec
ond
set
in
unive
rsal
m
obil
e
te
l
ec
om
m
unic
at
i
ons
sy
stem
(UMT
S)
net
works
.
The
ci
phe
r
uses
128
-
bit
k
e
y
and
128
bit
IV
to
produc
e
32
-
bit
ci
ph
ert
ex
t.
The
pap
er
pre
sents
two
te
chn
ique
s
for
per
form
anc
e
en
hanc
ement
.
The
first
te
chni
qu
e
uses
novel
CLA
arc
h
it
e
ct
ur
e
to
m
ini
m
iz
e
the
prop
aga
t
ion
dela
y
of
the
2
32
m
odulo
adde
rs.
The
sec
ond
technique
u
ses
novel
arc
h
it
e
ct
ure
for
S
-
box
t
o
m
ini
m
iz
e
the
chi
p
area.
The
pre
sen
te
d
work
uses
V
H
DL
la
nguag
e
for
codi
ng
.
The
sam
e
is
implemente
d
on
the
FP
G
A
devi
c
e
Virte
x
xc5vf
x100
e
m
anuf
ac
ture
d
b
y
Xil
inx.
Th
e
p
rese
nte
d
arc
hi
tectur
e
ac
hi
eve
d
a
m
axi
m
um
fre
quency
of
25
4.
9
MH
z and thr
oughput
of
7
.
22
35
Gbps
.
Ke
yw
or
d
s
:
Crypto
gr
a
phy
FPGA
SNO
W
3G
S
tream
c
iph
er
VHDL
W
i
reless
n
et
w
ork
s
ec
uri
ty
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
N.
B.
Hull
e
,
Dep
a
rtm
ent o
f El
ect
ro
nics
and Telec
omm
un
ic
at
ion
,
G H Rai
soni
Insti
tute of
En
gi
neer
i
ng and T
e
chnolo
gy,
Pune, I
nd
ia
.
Em
a
il
:
nag
nath
.hull
e@rais
on
i
.n
et
1.
INTROD
U
CTION
Secu
rity
of
th
e
records
is
i
m
po
rtant
in
the
syst
em
s
w
her
e
per
s
onal
and
fina
ncial
m
at
te
rs
are
involve
d.
Hidi
ng
of
i
nfor
m
ation
f
r
om
un
au
thoriz
ed
use
rs
beco
m
es
essen
ti
al
in
su
ch
syst
e
m
s
a
nd
ser
vices.
Crypto
gr
a
phy is
on
e of the w
i
dely
u
sed
te
ch
niques for
sec
uri
ng infor
m
at
ion
f
r
om
eavesd
r
oppe
rs.
Co
ns
id
erin
g
the
need
to
sec
ur
e
in
form
at
io
n
m
any
research
ers
ar
e
wor
kin
g
in
the
are
a
of
in
form
at
ion
secur
it
y.
To
m
ai
ntain
adv
a
nce
d
network
sec
ur
it
y,
the
c
on
ce
r
n
netw
ork
a
rch
it
ect
ur
e
m
us
t
change
from
tradit
ion
al
sec
uri
ty
to
adv
a
nce
d
sec
uri
ty
. Th
e sam
e m
ay
b
e achie
ve
d by sin
king
hole
s in
the se
cu
rity
w
al
l.
Crypto
gr
a
phy
al
gorithm
s
and
thei
r
ass
oci
at
ed
key
a
re
m
or
e
secur
e
wh
e
n
it
is
im
plem
ented
on
a
hardw
a
re
pla
tfor
m
[1
]
.
Side
-
cha
nnel
at
ta
c
ks
an
d
fa
ult
at
ta
cks
m
a
y
exist.
Howev
e
r,
de
velo
ped
al
gorithm
s
m
us
t
be
fast
e
nough
to
s
upport
a
utono
m
ous
prot
oco
ls.
T
hese
protoc
ols
us
e
dif
fer
e
nt
encr
y
ption
al
gorithm
s
for
a
dif
fer
e
nt
session.
Ma
ny
recent
auto
nom
ou
s
prot
o
c
ols
li
ke
secu
r
e
so
ckets
la
ye
r
(S
S
L)
a
nd
i
nter
net
protoc
ol sec
ur
i
ty
(I
Psec
) use
diff
e
re
nt c
ip
he
rs
f
or
diff
e
re
nt
sessions.
Hardwa
re
im
p
lem
entat
ion
of
the
c
rypt
ogra
ph
ic
al
gorithm
on
FP
GA
de
vi
ces
is
at
tract
ive
s
olu
ti
ons
because
F
PGA
s
are
rec
onfig
urable
[2
-
8]
.
T
hi
s
pro
per
ty
pro
vid
es
fle
xib
il
it
y
for
dynam
ic
syst
e
m
dev
el
opm
ent
and
capa
ble
of
i
m
ple
m
enting
a
wide
ra
ng
e
of
f
unct
ions/
arc
hitec
tures/al
go
rithm
s.
It
seem
s
to
be
si
gn
ific
ant
to
e
m
ph
asi
ze
FP
GA
base
d
im
ple
m
entat
ion
s
of
cry
ptog
raphic
al
gorithm
s,
especial
ly
h
igh
thr
ough
pu
t
arch
it
ect
ures
[
9
].
S
NOW
3G
al
go
rithm
i
s
the
c
or
e
of
the
3rd
ge
ne
rati
on
pa
rtne
rs
hip
pro
j
ect
(
3GPP
)
al
gorithm
s
UEA
2
a
nd
U
I
A2.
The
3GPP
is
a
j
oi
nt
at
tem
pt
betwee
n
te
le
com
m
un
ic
at
ion
associat
ions
(T
G)
t
o
m
ake g
lo
bally
app
li
cable
sp
ec
ific
at
ion
s
fo
r
l
ong
te
rm
ev
olut
ion
(LTE
)
m
ob
il
e phone
syst
e
m
s
[10,
11]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
545
-
557
546
The
prese
nted
work
us
es
optim
iz
ed
arch
it
ect
ur
e
f
or
SNO
W
3G
st
rea
m
ci
ph
er.
T
his
arch
it
ect
ure
requires
only
2K
byte
s
of
m
em
or
y
fo
r
im
plem
entat
ion
of
S
-
bo
x
in
place
of
8K
byte
s
of
m
e
m
or
y
requir
ed
f
o
r
the
existi
ng
S
NOW
3G
arc
hitec
tures
[
10
-
16
]
.
T
he
pa
pe
r
is
arr
a
ng
e
d
i
n
the
f
ollo
wing
sect
ions.
Se
ct
ion
-
2
pro
vid
es
init
ia
l
ver
si
on
s
of
S
NOW
stream
ci
ph
e
r
[
17
]
.
Se
c
ti
on
-
3
prov
i
des
the
w
orkin
g
a
nd
desig
n
par
a
m
et
ers
of
t
he
SNO
W
3G
al
gorithm
.
Sect
ion
-
4
li
sts
ex
ist
ing
w
ork
r
el
at
ed
to
prese
nted
te
ch
ni
qu
e
s.
Sect
io
n
-
5
pr
esents
op
ti
m
iz
ed
SN
O
W
3G
arc
hitec
ture
an
d
it
s
analy
sis.
The
resu
lt
s
are
dis
cusse
d
in
Sect
ion
-
6
an
d
Sec
ti
on
-
7
con
cl
ud
e
s the
pr
ese
nted
wo
rk.
2.
INITI
AL V
E
RS
IO
NS OF
SN
O
W
The
resea
rc
her
Patrik
Ek
da
hl
et
.
al
.
pr
op
ose
d
a
stream
cip
he
r
S
NOW
(
SNO
W
1.0)
i
n
the
ye
ar
2000
[
18
]
,
a
fter
two
ye
ar
s
Ha
wk
es
et
.
al
.
de
scribe
d
a
ne
w
at
ta
ck
known
a
s
a
gu
es
s
-
a
nd
-
determ
ine
at
tack
[
19
]
on
SNO
W
1.0
.
SNO
W
1.0
has
tw
o
li
m
it
a
ti
on
s.
T
he
firs
t
lim
it
ation
w
as
finite
sta
te
m
achine
(
F
S
M)
ha
s
a sin
gle in
pu
t,
wh
ic
h
al
lo
ws
t
he
at
ta
cke
r
to
di
sturb the
w
ork
ing
proce
dure i
n
F
SM an
d sec
ond SN
O
W 1.
0 was
li
ttle
un
luc
ky
in
ch
oosin
g
fee
db
ac
k
po
ly
no
m
ia
l.
This
al
lo
ws
creati
ng
bi
twise
corres
pond
e
nce
in
F
S
M
and
wh
ic
h
is t
he ba
se of
disti
nguis
hing a
tt
ack
.
Patrik
E
kd
a
hl
et
.
al
.
the
pr
opos
e
d
a
new
ver
si
on
of
S
NOW
ci
phe
r
as
SNO
W
2.0
[
20
]
with
m
od
ific
at
ion
s
in
S
NOW
1.0
[
18
]
.
T
hey
pr
ovide
d
tw
o
in
puts
to
FSM
a
nd
m
od
ifie
d
fee
db
ac
k
po
ly
no
m
ia
l
in
the
ne
w
ve
rsion
S
N
O
W
2.0
.
The
t
wo
in
pu
t
s
to
FSM
in
S
NOW
2.0
m
a
kes
the
guess
-
a
nd
-
determ
ine
at
ta
ck
m
or
e
diff
ic
ult
because
FSM
update
re
gister
s
R1
an
d
R2
do
not
de
pe
nd
only
on
F
SM
ou
tpu
t.
T
he
poly
no
m
ia
l
sel
ect
ion
in
S
NOW
1.0
was
m
ade
to
sp
eed
up
the
m
ulti
pli
cat
ion
by
le
ft
sh
ift
operati
on
in
LFSR.
T
his
al
lows
the
resu
lt
of
m
ul
ti
plica
ti
on
to
app
ea
r
at
m
any
places
as
a
bit
sh
ifte
d
ver
sio
n
of
th
e
or
igi
nal
wor
d.
S
uch
a
sel
ect
ion
of
po
ly
nom
ial
pr
ovides
a
ba
se
fo
r
c
orrelat
ion
at
ta
ck
in
the
init
ia
l
ver
sion
[
18
]
.
SNO
W
2.0
pro
vid
es
bette
r
distrib
utio
n
of
the
bits
in
fee
db
ac
k
f
un
c
ti
on
by
def
i
ning
fiel
d
(
F_2^3
2)
a
s
an
exte
ns
io
n
ove
r
the
fiel
d
(F_
2^8
)
.
Each
m
ultip
li
cat
ion
was
i
m
ple
m
ented
as
sh
ifti
ng
th
e
co
ntent
by
one
by
te
and
unco
nd
it
ion
al
XO
R wit
h
25
6
po
s
sible
patte
r
ns
.
S
o
S
N
O
W
2
.
0
[
20
]
is
str
ong
a
gain
st
corr
el
at
ion
at
ta
ck
a
s
com
pa
red
to Snow
1.0
[
18
]
.
D
uri
ng
the
e
valuat
ion
of
T
he
E
uro
pea
n
Tel
eco
m
m
un
ic
at
io
ns
Stand
a
r
ds
I
ns
ti
tute
(ET
SI)/
Se
cur
it
y
Algorithm
s
Group
of
E
xp
e
rts
(S
A
GE
),
the
SNO
W
2.0
wa
s
furthe
r
m
od
ifie
d
to
increase
it
s
resist
ance
against
al
gebraic at
ta
cks
a
nd the
n
e
w
d
esi
gn
nam
ed a
s SNO
W 3G [
10
].
3.
SPECIFI
C
AT
ION
S
AN
D
W
ORKIN
G OF
SN
O
W
3G
SNO
W
3G
ge
ner
at
es
a
32
-
bit
ci
ph
erte
xt
pe
r
cl
oc
k
cy
cl
e
with
the
help
of
a
128
-
bit
ke
y
and
128
-
bi
t
init
ia
li
zation
ve
ct
or
(IV)
as
sh
ow
n
i
n
Fi
gure
1.
It
co
nsi
sts
of
the
m
a
in
f
our
m
od
ul
es
init
ia
l
o
perat
ion
s,
li
near
fee
dbac
k
s
hift
re
gister
(LFS
R
),
finite
sta
te
m
achine
(
FSM),
an
d
a
fe
edb
ac
k
path
.
T
he
init
ia
l
opera
ti
on
s
will
div
ide
12
8
bit
Key
into
four
blo
c
ks
as
per
eq
uatio
ns
(
01),
(
02),
(
03)
,
and
(
04).
Sim
il
arly
,
it
al
so
div
ides
128 bit
IV i
nto f
our
blo
c
ks
as
per eq
uatio
ns
(
05), (
06), (
07), an
d (08)
[
10]
.
3
=
[
0
]
‖
[
1
]
‖
[
2
]
‖
…
‖
[
31
]
(1)
2
=
[
32
]
‖
[
33
]
‖
[
34
]
‖
…
‖
[
63
]
(2)
1
=
[
64
]
‖
[
65
]
‖
[
66
]
‖
…
‖
[
95
]
(
3)
0
=
[
96
]
‖
[
97
]
‖
[
98
]
‖
…
‖
[
127
]
(4)
3
=
[
0
]
‖
[
1
]
‖
[
2
]
‖
…
‖
[
31
]
(5)
2
=
[
32
]
‖
[
33
]
‖
[
34
]
‖
…
‖
[
63
]
(6)
1
=
[
64
]
‖
[
65
]
‖
[
66
]
‖
…
‖
[
95
]
(7)
0
=
[
96
]
‖
[
97
]
‖
[
98
]
‖
…
‖
[
127
]
(8)
w
he
re
[
0
]
,
[
0
]
are L
S
B par
t a
nd
[
127
]
,
[
127
]
ar
e MSB
p
a
rt
of
the k
ey
a
nd
IV res
pecti
vely
.
In
it
ia
l
operati
ons
a
re
perform
ed
on
key
a
nd
iv
as
pe
r
T
a
ble
1.
The
outp
ut
of
t
he
init
ia
l
op
e
rati
ons
blo
c
k
is
loa
de
d
int
o
LF
SR
be
fore
the
first
cl
ock
cy
cl
e
[
10]
.
T
he
sec
ond
m
od
ule
L
FS
R
con
sist
s
of
sixtee
n
sta
ges
each
ha
ving
pa
rall
el
32
bits.
C
on
te
nts
of
L
FSR
are
sh
ifte
d
from
M
SB
(S15)
to
L
SB
(S0)
in
eac
h
cl
oc
k
cy
cl
e.
S1
5
rec
ei
ves
ne
w
value
f
ro
m
the
fe
edb
ac
k
path
at
each
cl
oc
k
cy
cl
e.
Thir
d
m
odule
FSM
c
on
si
sts
of
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Op
ti
mize
d arc
hitec
ture fo
r
SN
OW 3G
(N
. B
. Hulle
)
547
three
par
al
le
l
32
-
bit
update
r
egiste
r
s
R
1,
R
2,
R
3,
tw
o
S
-
Box
es
S1,
S
2
each
of
4Kbyt
es,
tw
o
32
bit
m
od
ulo
add
e
rs
a
nd
tw
o
32
bit
X
OR
ga
te
s.
The
fi
nal
m
od
ule
is
the
f
eedb
ac
k
pat
h
wh
ic
h
c
onsist
s
of
f
unct
ions
MULα
,
DIVα, a
nd m
a
ny XOR
operat
ion
s
. T
he
tw
o f
un
ct
io
ns M
UL
α
an
d D
I
Vα ar
e i
m
ple
m
ented
as lo
okup ta
bles.
D
i
v
i
d
e
K
E
Y
a
n
d
I
V
1
2
8
3
2
3
2
3
2
3
2
3
2
K
0
K
1
K
2
K
3
I
V
0
I
V
1
I
V
2
I
V
3
I
n
i
t
i
a
l
O
p
e
r
a
t
i
o
n
s
S
1
5
S
1
4
S
1
3
S
1
2
S
1
1
S
1
0
S
9
S
8
S
7
S
6
S
5
S
4
S
3
S
2
S
1
S
0
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
1
2
8
L
F
S
R
S
5
S
1
5
S
2
S
1
1
S
0
F
S
M
3
2
3
2
F
e
e
d
b
a
c
k
3
2
3
2
K
e
y
S
t
r
e
a
m
3
2
3
2
3
2
F
v
K
e
y
I
V
M
o
d
e
0
1
0
:
I
n
i
t
i
a
l
i
z
a
t
i
o
n
m
o
d
e
1
:
K
e
y
s
t
r
e
a
m
m
o
d
e
C
l
k
R
s
t
1
R
s
t
2
Figure
1
.
Exist
ing
SNO
W 3G
a
rc
hitec
tures
Table
1
.
L
FS
R
i
niti
al
iz
at
ion
calc
ulati
on
s
0
=
0
⊕
1
4
=
0
8
=
0
⊕
1
12
=
0
⊕
1
1
=
1
⊕
1
5
=
1
9
=
1
⊕
1
⊕
3
13
=
1
2
=
2
⊕
1
6
=
2
10
=
2
⊕
1
⊕
2
14
=
2
3
=
3
⊕
1
7
=
3
11
=
3
⊕
1
15
=
3
⊕
0
SNO
W
3G
w
orks
into
tw
o
m
od
e
s
of
operati
on,
init
ia
li
zat
ion
m
od
e
a
nd
ke
yst
rea
m
m
od
e.
At
t
he
sta
rt
of
i
niti
al
iz
at
io
n,
t
he
m
od
el
s
yst
e
m
sh
ou
l
d
r
eset
LFSR
a
nd
FSM
us
in
g
te
rm
inals
Rst
1
and
Rst
2
res
pec
ti
vely
.
In
t
he first cl
oc
k
cy
cl
e v
al
ues c
al
culat
ed
in t
he
init
ia
li
zat
ion
, a
m
od
e is loa
de
d
int
o
sixtee
n st
ages
of
LFS
R but
FSM
reg
ist
er
s
sh
oul
d
rem
ai
n
in
a
reset
sta
te
.
In
the
sec
ond
cl
ock
cy
cl
e,
Rst
2=0
an
d
now
LFSR
is
cl
oc
ked.
At
eac
h
cl
oc
k,
32
-
bit
outp
u
t
F
of
FSM
is
c
om
bin
ed
with
S0
,
S
2
&
S
11
in
t
he
feedba
ck
path
by
sel
ect
in
g
m
od
e
0
from
a
sel
ect
li
ne
of
the
m
ulti
ple
xer
a
nd
a
ppli
ed
to
S
15
as
in
te
rm
ediat
e
s
ign
al
v.
T
he
f
ollow
i
ng
equ
at
io
n p
rovi
des
t
he
interm
ediat
e sig
nal v in the
init
ia
li
zation
m
od
e
[
10
]
.
=
(
0
,
1
‖
0
,
2
‖
0
,
3
‖
0
00
)
⊕
(
0
,
0
)
⊕
S2
⊕
(
0
00
‖
11
,
0
‖
11
,
1
‖
11
,
2
)
⊕
(
11
,
3
)
⊕
F
(
9)
Af
te
r
32
cl
oc
k
cy
cl
es,
SN
O
W
3G
ente
rs
into
keyst
ream
m
od
e.
O
pe
rati
on
s
in
this
m
od
e
ar
e
the
sa
m
e
as
init
ia
li
zation
m
od
e
but
th
e
on
ly
dif
fere
nc
e
is
t
hat
ou
tp
ut
F
of
FSM
i
s
no
t
c
om
bin
ed
in
fee
dbac
k
path
by
m
aking
m
od
e
=
1
f
r
om
the
m
ulti
plexer.
T
he
interm
ediat
e
s
ign
al
in
keyst
r
ea
m
m
od
e
is
gi
ven
by
the
f
ollow
i
ng
equ
at
io
n
[
10
]
.
=
(
0
,
1
‖
0
,
2
‖
0
,
3
‖
0
00
)
⊕
(
0
,
0
)
⊕
S2
⊕
(
0
00
‖
11
,
0
‖
11
,
1
‖
11
,
2
)
⊕
(
11
,
3
)
(10)
In
k
ey
stream
m
od
e,
FSM
is cl
ock
e
d
f
or
on
e
cl
ock
cy
cl
e
a
nd
it
s f
irst
outp
ut
is
disca
rd
e
d
wh
e
n
it
is
cl
oc
ked
for
n
cl
oc
k
cy
cl
es
to
encr
y
pt
n
nu
m
ber
of
32
-
bit
w
ords,
w
her
e
n
=
num
ber
of
32
-
bit
data
w
ords
is
to
be
encr
y
pted [
10
].
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
545
-
557
548
4.
RELATE
D
W
ORK
The
stu
dy
of
existi
ng
arc
hi
te
ct
ur
es
of
S
NOW
3G
ev
olv
e
d
two
c
ha
ll
eng
es.
On
e
m
ini
m
iz
ing
pro
pag
at
io
n
de
la
y
of
the
2
32
m
od
ulo
ad
ders
and
oth
er
is
m
ini
m
iz
ing
the
chip
area
of
S
-
boxe
s.
The
rese
arch
e
r
Kitsos
et
al
.
[
12
]
reali
zed
S
-
bo
xes
us
i
ng
8
look
up
ta
bles.
Each
lo
ok
up
ta
ble
c
on
s
um
es
1
KB
m
e
m
or
y,
so
m
e
m
or
y
us
ed
f
or
S
-
bo
x
re
al
iz
at
ion
is
8
KB.
Jai
ra
j
et
al
.
us
e
d
sym
m
e
try
of
S
-
box
lo
okup
ta
bles
to
m
ini
m
iz
e
cach
e
re
qu
i
rem
ent
in
t
he
s
of
t
wa
re
im
ple
m
enta
ti
on
of
S
NOW
3G
[
21
].
Kit
so
s
et
al
.
[12]
use
d
conve
ntion
al
CLA
for
m
od
ul
o
a
dder
im
ple
m
entat
ion
.
T
he
researc
her
Pai
a
nd
Chen
[
22]
pr
esente
d
a
m
od
ifie
d
CL
A
desig
n
t
o
m
i
nim
iz
e
the
pro
pag
at
io
n
delay
.
Tra
boulsi
et
al
.
[23]
im
ple
m
ented
S
NOW
3G
on
an
em
bed
de
d
platfo
rm
.
The
m
o
ti
ve
of
t
he
desi
gn
wa
s
to
m
ini
m
iz
e
the
m
e
m
or
y
require
d
f
or
S
-
box
i
m
ple
m
entat
io
n.
Re
sea
rc
her
s
us
e
d
2
lo
ok
up
ta
bles
i
n
place
of
8
l
ooku
p
ta
bles
f
or
im
ple
m
entat
ion
of
2
S
-
bo
xes. Ei
ght
-
bit shiftin
g wit
h
cac
he
m
em
or
y i
s u
sed
ef
fici
ently
to
m
i
nim
iz
e
m
e
m
or
y req
uirem
ent
.
5.
PRESENTE
D
SNOW
3G A
RCHITE
CT
U
RE
Con
si
der
i
ng
t
he
chall
en
ges
of
e
xisti
ng
FSM,
the
pro
po
s
ed
im
ple
m
entat
ion
us
es
the
fo
ll
ow
i
ng
ref
inem
ents to im
pr
ove the
p
e
rfor
m
ance of t
he
S
N
O
W
3G
al
gorithm
.
-
Use
of
novel
m
od
ulo
CLA
a
rch
it
ect
ure
ove
r
2
32
to
m
ini
m
i
ze
prop
a
gatio
n
delay
in
FSM,
wh
ic
h
decides
the criti
cal
d
el
ay
o
f
the al
gori
thm
-
Use of
no
vel S
-
Bo
x
a
rch
it
ect
ur
e
to
m
ini
m
ize
chip
area
5
.
1.
N
ovel
m
odul
o CLA
a
r
c
hitecture
over
2
32
Modulo
a
dd
e
rs
are
us
ua
ll
y
i
m
ple
m
ented
by
us
ing
ri
pple
-
carry
ad
ders,
but
this
te
chn
iq
ue
increa
ses
the
pro
pa
gatio
n
delay
of
the
crit
ic
al
path.
T
he
prop
a
gatio
n
delay
of
n
bit
rip
ple
car
ries
add
e
r
is
(2n+
1)
gate
delay
s.
Mo
du
lo
ad
de
r
ov
e
r
2
32
im
plem
ented
by
usi
ng
rip
ple
-
ca
rr
y
a
dd
e
rs
w
il
l
hav
e
dela
y
of
(2*
32
+
1
=
65)
65
gates
delay
,
assum
ing
avera
ge
gate
delay
of
10
ηs
the
tot
al
delay
of
on
e
m
od
ulo
a
dd
e
r
will
be
65*10
=
650
ηs.
FSM
consi
sts
of
tw
o
s
uch
a
dd
e
rs
so
a
total
de
la
y
of
m
odulo
a
dd
e
rs
f
or
sing
le
com
pu
ta
ti
on
w
il
l be 1300
ηs
.
The
pro
pa
gation
delay
of
m
odulo
a
dd
e
rs
c
an
be
m
ini
m
ized
by
us
in
g
C
LA
f
or
it
s
i
m
ple
m
entat
ion
.
Existi
ng
CL
As
are
reali
zed
by
us
ing
basic
gates
i.e.
A
N
D
,
XO
R
,
an
d
O
R
gates,
but
P
ai
et
al
.
reali
zed
CLA
by
usi
ng
un
i
ve
rsal
ga
te
s
i.e.
N
AND
or
N
OR
gates
[
22
]
.
The
sam
e
des
ign
m
ini
m
iz
ed
gate
requirem
ent
as
c
om
par
ed
to
e
xisti
ng
a
rch
it
e
ct
ur
es.
At
the
s
a
m
e
tim
e,
this
CLA
[22
,
24
]
desig
n
s
a
re
fas
te
r
than
c
onve
ntion
a
l
CLA
arc
hitec
tures
.
Adder
a
rch
it
ect
ure
[
25]
de
velo
ped
for
L
ILI
-
I
I
ci
ph
e
r
us
es
dif
f
eren
t
a
ppr
oac
h
f
or
add
ia
ti
on.
Re
du
ct
io
n
in
pro
pag
at
io
n
de
la
y
and
c
hip
area
is
possi
ble
in
existi
ng
arch
it
ect
ures
[
12
-
16,
22
],
so
the
present
ed
resea
rch
w
ork
us
es
un
i
ve
rsal
gates
for
CLA
i
m
ple
m
entat
ion
an
d
oth
e
r
te
chn
i
ques
to
m
ini
m
iz
e
the
nu
m
ber
of
gat
es
require
d.
N
ov
el
m
od
ulo
CLA
arc
hitec
ture
ov
e
r
2
32
use
s
fo
ll
owin
g
t
hr
ee
arch
it
ect
ures in
m
ult
il
evel CLA
desig
ns
f
or
pe
rfor
m
ance im
pro
vem
ent
-
4 bit
CLA at
L
SB (to
calc
ulat
e S0 to
S3)
-
4 bit
CLA at m
idd
le
sta
ges (to
calc
ulate
S
4
to
S27)
-
4 bit
CLA at
MSB
(
to
calc
ul
at
e S28 to S
31)
Using
t
he
a
bove
CL
A
a
rch
it
ect
ur
es
no
vel
arch
it
ect
ure
f
or
m
od
ul
o
CL
A
over
2
32
was
desig
ne
d
as
sh
ow
n
i
n
Fi
gur
e
2.
P
resen
te
d
m
od
ulo
a
dd
e
r
arch
it
ect
ure
is
an
a
rea,
pro
pa
gation
delay
,
a
nd
ene
r
gy
-
ef
fici
ent
as
com
par
ed
t
o
e
xist
ing m
od
ulo C
LA
a
rc
hitec
tures
.
G
a
t
e
r
e
d
u
c
t
i
o
n
a
t
l
e
v
e
l
2
,
S
t
a
g
e
1
U
n
u
s
e
d
l
o
g
i
c
c
i
r
c
u
i
t
r
e
d
u
c
t
i
o
n
F
o
u
r
b
i
t
L
S
B
a
d
d
e
r
F
o
u
r
b
i
t
m
i
d
d
l
e
s
t
a
g
e
a
d
d
e
r
F
o
u
r
b
i
t
M
S
B
a
d
d
e
r
G
a
t
e
r
e
d
u
c
t
i
o
n
a
t
l
e
v
e
l
3
,
s
t
a
g
e
1
FA
_
WC
S
0
G
0
B
0
A
0
FA
_
NAND
S
1
G
1
P
1
B
1
A
1
C
1
FA
_
NAND
S
2
G
2
P
2
B
2
A
2
C
2
FA
_
NAND
S
3
G
3
P
3
B
3
A
3
C
3
GG
0
FA
_
NAND
S
4
G
4
P
4
B
4
A
4
C
4
FA
_
NAND
S
5
G
5
P
5
B
5
A
5
C
5
FA
_
NAND
S
6
G
6
P
6
B
6
A
6
C
6
FA
_
NAND
S
7
G
7
P
7
B
7
A
7
C
7
GG
1
PG
1
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
FA
_
NAND
S
8
G
8
P
8
B
8
A
8
C
8
FA
_
NAND
S
9
G
9
P
9
B
9
A
9
C
9
FA
_
NAND
S
10
G
10
P
10
B
10
A
10
C
10
FA
_
NAND
S
11
G
11
P
11
B
11
A
11
C
11
GG
2
PG
2
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
FA
_
NAND
S
12
G
12
P
12
B
12
A
12
C
12
FA
_
NAND
S
13
G
13
P
13
B
13
A
13
C
13
FA
_
NAND
S
14
G
14
P
14
B
14
A
14
C
14
FA
_
NAND
S
15
G
15
P
15
B
15
A
15
C
15
GG
3
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
PG
3
GG
16
FA
_
NAND
S
16
G
16
P
16
B
16
A
16
FA
_
NAND
S
17
G
17
P
17
B
17
A
17
C
17
FA
_
NAND
S
18
G
18
P
18
B
18
A
18
C
18
FA
_
NAND
S
19
G
19
P
19
B
19
A
19
C
19
GG
4
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
PG
4
FA
_
NAND
S
20
G
20
P
20
B
20
A
20
FA
_
NAND
S
21
G
21
P
21
B
21
A
21
C
21
FA
_
NAND
S
22
G
22
P
22
B
22
A
22
C
22
FA
_
NAND
S
23
G
23
P
23
B
23
A
23
C
23
GG
5
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
PG
5
FA
_
NAND
S
24
G
24
P
24
B
24
A
24
FA
_
NAND
S
25
G
25
P
25
B
25
A
25
C
25
FA
_
NAND
S
26
G
26
P
26
B
26
A
26
C
26
FA
_
NAND
S
27
G
27
P
27
B
27
A
27
C
27
GG
6
N
1
N
3
N
6
N
13
N
12
N
11
N
10
N
14
N
2
N
4
N
5
N
8
N
7
N
9
PG
6
C
16
C
20
C
24
FA
_
NAND
S
28
G
28
P
28
B
28
A
28
FA
_
NAND
S
29
G
29
P
29
B
29
A
29
C
29
FA
_
NAND
S
30
G
30
P
30
B
30
A
30
C
30
FA
_
NAND
_
W
G
I
P
I
S
31
B
31
A
31
C
31
C
28
Figure
2
.
N
ove
l
m
o
du
l
o
CL
A archit
ect
ure
over
2
32
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Op
ti
mize
d arc
hitec
ture fo
r
SN
OW 3G
(N
. B
. Hulle
)
549
5
.
2.
N
ovel
S
-
Box arc
hitect
ure
Tw
o
S
-
boxes
S1
&
S
2
are
us
ed
in
S
N
O
W
3G
a
rch
i
te
ct
ur
e
each
r
equ
i
res
m
e
m
o
ry
of
4
KB
.
The
lo
okup
ta
ble
of
S
1
is
ta
ken
f
ro
m
the
Ri
j
ndael
substi
tuti
on
box
a
nd
a
look
up
ta
bl
e
of
S
2
is
ba
s
ed
on
Dicks
o
n
po
ly
nom
ial
ov
er
GF
-
28.
A
s
per
d
es
ign
sp
eci
ficat
ion,
each
S
-
box
(S
1
or
S2)
is
im
ple
m
ented
by
us
in
g
4
lo
okup
ta
ble
s
an
d
eac
h
lo
okup
ta
ble
has
256
val
ues
ea
ch
of
4
byte
s.
S
o
the
im
ple
m
ent
at
ion
of
eac
h
l
ookup
ta
ble
re
qu
i
res
(
256x4
=
1024
by
te
s
of
m
e
m
or
y).
Eac
h
S
-
box
has
4
lo
okup
t
ables,
s
o
total
m
e
m
or
y
requir
ed
for
the
i
m
ple
m
ent
at
ion
of
S1
or
S2
is
(4x
1024
=
4K
)
4K
B
.
The
total
m
e
m
o
ry
need
e
d
f
or
t
he
reali
zat
ion
of
tw
o
S
-
bo
xes
is
8K
B. Ex
ist
in
g
im
plem
entat
ion
[
10
-
16, 26
-
29
]
us
e
s
S
-
box arc
hitec
ture
a
s s
how
n
i
n
Fi
gure
3.
S
1
_
T
0
S
1
_
T
1
S
1
_
T
2
S
1
_
T
3
3
2
8
8
8
8
3
2
3
2
3
2
3
2
3
2
S
2
_
T
0
S
2
_
T
1
S
2
_
T
2
S
2
_
T
3
3
2
8
8
8
8
3
2
3
2
3
2
3
2
3
2
S
-
b
o
x
S
1
I
m
p
l
e
m
e
n
t
a
t
i
o
n
S
-
b
o
x
S
2
I
m
p
l
e
m
e
n
t
a
t
i
o
n
Figure
3
.
Exist
ing
S
-
boxes
ar
chite
ct
ur
e
The
fou
r
lo
ok
up
ta
bles
of
S
1
i.e.
S
1_T0
t
o
S
1_T
3
as
show
n
i
n
Fig
ur
e
3
has
the
sam
e
co
ntent
but
exist
in
8
bit
sh
ifte
d
f
orm
.
An
al
ogous
is
the
case
of
S
-
box
S
2.
Pr
ese
nted
no
v
el
S
-
box
arc
hitec
tu
res
us
e
a
sing
le
look
up
ta
ble
for
i
m
ple
m
entat
ion
of
S
-
bo
x
(S1
or
S
2).
Pr
ese
nted
r
esearch
wor
k
use
s
two
arc
hitec
tures
for
S
-
box
im
ple
m
entat
ion
.
Fi
rst
arch
it
ect
ure
as
sh
ow
n
in
Figure
4,
co
nsum
es
few
er
re
so
urces
bu
t
use
fu
l
to
low
-
f
reque
ncy
app
li
cat
ion
s
only
.
Seco
nd
a
r
chite
ct
ur
e
as
s
how
n
in
Fig
ure
5,
co
nsum
es
few
er
resou
rc
es
as
com
par
ed
t
o
e
xisti
ng arc
hitec
tures b
ut r
e
quir
ed
m
or
e re
sour
ces as c
om
par
ed
to
No
vel S
-
box arc
hitec
ture
-
1.
A
d
d
r
1
A
d
d
r
2
A
d
d
r
3
A
d
d
r
4
D
a
t
a
O
u
t
1
D
a
t
a
O
u
t
2
D
a
t
a
O
u
t
3
D
a
t
a
O
u
t
4
8
8
8
8
3
2
3
2
3
2
3
2
M
u
l
t
i
p
o
r
t
R
O
M
8
3
2
3
2
B
i
t
L
a
t
c
h
3
2
3
2
3
2
3
2
3
2
B
i
t
L
a
t
c
h
3
2
B
i
t
L
a
t
c
h
3
2
B
i
t
L
a
t
c
h
2
B
i
t
C
o
u
n
t
e
r
Figure
4
.
N
ove
l S
-
Bo
x
a
rc
hit
ect
ur
e
1
Pr
ese
nted
desi
gn
s
re
quire
2
KB
of
m
e
m
or
y
for
t
he
reali
zat
ion
of
S
-
bo
xe
s
S
1
&
S
2.
T
hese
desi
gn
s
save
6
KB
of
m
e
m
or
y
as
co
m
par
ed
to
e
xis
ti
ng
desig
ns
.
S
-
bo
x
arc
hitec
tu
re
-
1
sa
ves
6
K
B
m
e
m
or
y
at
t
he
c
os
t
of
s
om
e
add
it
ion
al
ha
r
dw
a
r
e
(S
in
gle
2
-
bit
coun
te
r,
t
w
o
4
I/p
m
ultip
le
xe
rs,
a
nd
f
our
32
bit
la
tc
hes).
This
a
rch
it
ect
ure
is
4
ti
m
es
sl
ow
e
r
t
han
co
nventio
nal
a
rch
i
te
ct
ur
es
a
nd
use
fu
l
for
l
ow
-
f
r
equ
e
ncy
a
pp
li
c
at
ion
s.
S
-
bo
x
arc
hite
ct
ur
e
-
2
has
the
sam
e
sp
eed
as
c
onve
ntion
al
a
rch
i
te
ct
ur
es
bu
t
us
es
4
ad
dit
ion
al
256:1
m
ulti
plexer
s
.
T
he
seco
nd
a
rc
hitec
ture
can
be
us
e
d
f
or
l
ow
a
nd
hi
gh
-
s
pee
d
ap
plic
at
ion
s
dep
e
nding
on
cost a
nd sp
ee
d t
rad
e
offs
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
545
-
557
550
L
O
C
1
L
O
C
0
L
O
C
2
5
5
M
U
X
M
U
X
M
U
X
O
u
t
1
f
o
r
A
1
O
u
t
2
f
o
r
A
2
O
u
t
4
f
o
r
A
4
A
d
d
r
e
s
s
A
1
A
d
d
r
e
s
s
A
2
A
d
d
r
e
s
s
A
4
3
2
3
2
3
2
8
8
8
Figure
5
.
N
ove
l S
-
Bo
x
a
rc
hitec
ture 2
5
.
3.
Op
timi
z
e
d SNOW
3G
ar
chitecture
Op
ti
m
iz
ed
SNO
W
3G
arc
hit
ect
ur
e
as
sho
wn
i
n
Fi
gure
6
is
de
sig
ned
us
in
g
novel
m
odulo
CLA
arch
it
ect
ure
a
nd
novel
S
-
B
ox
a
rch
it
ect
ure
as
discusse
d
in
the
pre
viou
s
sect
io
n.
SNO
W
3G
arc
hitec
ture
desig
ne
d
us
in
g
S
-
Bo
x
a
rc
hitec
ture
-
1
is
us
e
d
for
l
ow
-
fr
e
quency
app
li
cat
io
ns
a
nd
nee
d
s
tw
o
cl
oc
k
arr
a
ng
em
ents.
Wh
ereas
S
NOW
3G
a
r
chite
ct
ur
e
des
ign
e
d
us
in
g
S
-
Bo
x
a
rc
hitec
ture
-
2
is
use
d
for
high
-
fr
e
quency
app
li
cat
ion
s
and
needs
a
sing
le
cl
ock.
I
nter
nal
blo
c
k
diag
ram
of
optim
iz
ed
SN
O
W
3G
arch
it
ect
ure as
sh
ow
n
in
Fi
gur
e 7
.
D
i
v
i
d
e
K
E
Y
a
n
d
I
V
1
2
8
3
2
3
2
3
2
3
2
3
2
K
0
K
1
K
2
K
3
I
V
0
I
V
1
I
V
2
I
V
3
I
n
i
t
i
a
l
O
p
e
r
a
t
i
o
n
s
S
1
5
S
1
4
S
1
3
S
1
2
S
1
1
S
1
0
S
9
S
8
S
7
S
6
S
5
S
4
S
3
S
2
S
1
S
0
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
1
2
8
L
F
S
R
S
5
S
1
5
S
2
S
1
1
S
0
F
e
e
d
b
a
c
k
3
2
3
2
K
e
y
S
t
r
e
a
m
3
2
3
2
3
2
F
v
K
e
y
I
V
M
o
d
e
0
1
0
:
I
n
i
t
i
a
l
i
z
a
t
i
o
n
m
o
d
e
1
:
K
e
y
s
t
r
e
a
m
m
o
d
e
C
l
k
f
a
s
t
R
s
t
1
S
1
a
n
d
S
2
L
o
g
i
c
2
b
i
t
R
e
g
i
s
t
e
r
C
l
k
s
l
o
w
O
t
h
e
r
l
o
g
i
c
i
n
F
S
M
R
s
t
2
F
S
M
Figure
6
.
To
p m
od
ule of r
e
fi
ned S
NOW
3G
arc
hitec
ture
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Op
ti
mize
d arc
hitec
ture fo
r
SN
OW 3G
(N
. B
. Hulle
)
551
3
2
3
2
3
2
S
0
S
1
S
2
S
5
S
1
4
S
1
5
.
.
.
.
.
.
L
F
S
R
3
2
R
3
R
1
S
2
-
B
O
X
R
2
S
1
-
B
O
X
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
F
e
e
d
b
a
c
k
F
S
M
M
o
d
e
0
1
3
2
S
1
1
D
i
v
i
d
e
K
E
Y
a
n
d
I
V
1
2
8
K
0
K
1
K
2
K
3
I
V
0
I
V
1
I
V
2
I
V
3
I
n
i
t
i
a
l
O
p
e
r
a
t
i
o
n
s
S
1
5
S
1
4
S
1
3
S
1
2
S
1
1
S
1
0
S
9
S
8
S
7
S
6
S
5
S
4
S
3
S
2
S
1
S
0
1
2
8
K
E
Y
I
V
K
e
y
S
t
r
e
a
m
O
u
t
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
D
I
V
α
>
>
8
3
2
3
2
3
2
3
2
M
U
L
α
<
<
8
3
2
0
:
I
n
i
t
i
a
l
i
z
a
t
i
o
n
m
o
d
e
1
:
K
e
y
s
t
r
e
a
m
m
o
d
e
0
1
F
v
M
o
d
C
L
A
M
o
d
C
L
A
Figure
7
.
I
nter
nal b
l
ock d
ia
gra
m
o
f o
pti
m
iz
e
d
S
N
O
W
3G a
rch
it
ect
ure
FSM
of
S
NOW
3G
a
rch
it
e
ct
ur
e
c
onsist
s
of
t
wo
m
odulo
ad
de
rs
a
nd
t
wo
S
-
B
ox
e
s.
Tw
o
m
od
ul
o
add
e
rs
will
de
ci
de
the
sp
e
ed
of
the
al
gorithm
and
two
S
-
bo
xes
will
decide
ha
rdwar
e
util
iz
at
ion
of
the
al
gorithm
.
The
us
e
of
novel
m
od
ulo
C
LA
over
2
32
m
ini
m
iz
es
pro
pag
at
io
n
delay
an
d
the
use
of
novel
S
-
bo
x
a
rch
it
ec
ture
m
ini
m
iz
es
hard
war
e
util
iz
at
ion
.
Thes
e
ref
inem
ents
he
lp
to
im
pr
ove
the
pe
rfor
m
an
ce
of
the S
NOW
3G
al
gorithm
in
te
rm
s o
f
th
rou
ghpu
t a
nd a
rea.
Op
ti
m
iz
ed
SN
O
W
3G
a
rch
it
ect
ur
e
use
s
V
HD
L
la
ngua
ge
fo
r
c
odin
g.
T
he
sam
e
is
i
mp
le
m
ented
on
the
FPGA
de
vice
Virtex
xc
5vf
x100e
m
anu
fact
ur
e
d
by
Xili
nx
[
30
]
.
T
he
pr
e
sente
d
arch
it
ect
ure
achiev
e
d
a
m
axi
m
u
m
fr
equ
e
ncy
of
25
4.9
MHz
a
nd
thr
oughput
of
7.223
5
Gbps.
Table
2
s
how
s
pa
rtic
ulars
a
bout
the
te
ch
nolo
gy
us
e
d.
Fig
ur
e
8
a
nd
Fig
ure
9
s
how
RTL
s
chem
at
ic
and
ou
t
pu
t
wa
vefo
rm
of
the
pres
ented
arch
it
ect
ure re
s
pecti
v
el
y.
Table
2
.
T
ech
nolo
gy used
d
et
ai
ls
SNO
W
3
GNE
W
Pr
o
ject Status
(
0
5
/0
3
/2
0
2
0
-
1
1
:3
4
:4
1
)
Project File:
SNO
W
3
GOPT
.xis
e
Parser
E
r
rors:
No
E
r
rors
Mod
u
le Na
m
e:
SNO
W
3
G
I
m
p
le
m
en
tatio
n
St
ate:
Sy
n
th
esized
Tar
g
et
Dev
ice:
x
c5
v
fx1
0
0
t
-
3
ff
1
1
3
6
Er
rors:
No
E
r
rors
P
rod
u
ct Ver
sio
n
:
ISE
13
.2
W
arnin
g
s:
No
W
a
rnin
g
s
Desig
n
Goal:
Balan
ced
Ro
u
tin
g
Resu
lts:
Desig
n
Str
ateg
y
:
Xilin
x
Defau
lt (
u
n
lo
ck
ed
)
Ti
m
in
g
Co
n
strain
ts:
Env
iron
m
en
t:
Sy
ste
m
Setting
s
Fin
al T
i
m
in
g
Scor
e:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
545
-
557
552
Figure
8
.
RTL
schem
at
ic
Figure
9
.
O
utput wa
ve
form
6.
RESU
LT
A
N
D DIS
CUSSI
ONS
The
f
ollow
i
ng
sect
ion
disc
us
ses
the
res
ult
in
te
r
m
s
of
area
,
pro
pa
gation
delay
,
through
pu
t
,
and m
e
m
or
y uti
li
zed
for pres
e
nted SN
O
W 3
G
a
rch
it
ect
ure.
6
.
1.
The
are
a
6.1.1.
N
ov
el
m
od
ul
o CLA
ar
chitecture
ove
r 2
32
Pr
ese
nted
nove
l
m
od
ul
o
CL
As
a
re
us
e
d
as
m
od
ulo
ad
der
s
ove
r
2
32
in
O
ptim
iz
ed
SNO
W
3G
arch
it
ect
ure.
A
com
par
iso
n
of
dev
ic
e
util
iz
at
ion
of
existi
ng
[
13,
22
]
an
d
pr
ese
nte
d
ar
chite
ct
ur
es
is
sh
ow
n
in Figu
re
10.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Op
ti
mize
d arc
hitec
ture fo
r
SN
OW 3G
(N
. B
. Hulle
)
553
Figure
10
.
C
om
par
ison
s
o
f
hard
war
e
u
ti
li
zat
ion
for
CL
A
a
rch
it
ect
ures
6.1.2.
N
ov
el
S
-
Box arc
hitect
ure
Op
ti
m
iz
ed
SNO
W
3G
a
rch
it
ect
ur
e
us
es
N
ov
el
S
-
box
a
rc
hitec
ture
t
o
av
oid
redu
nd
a
nc
y
of
l
ookup
ta
bles.
P
rese
nted
Novel
S
-
B
ox
arc
hitec
ture
-
1
is
s
uitable
fo
r
lo
w
-
f
reque
ncy
ap
plica
ti
ons
a
nd
Novel
S
-
Box
arch
it
ect
ure
-
2
is
us
efu
l
f
or
high
-
fr
e
quency
app
li
cat
ion
s
.
The
us
e
of
th
ese
novel
arc
hitec
tures
m
in
i
m
iz
es
hard
war
e
re
quirem
ent
as
sh
ow
n
in
t
he
F
ig
ure
11
.
T
he
co
m
par
ison
sho
w
s
that
the
hard
war
e
re
sourc
es
us
ed
in
the
pr
ese
nted
arch
it
ect
ures
a
re
le
ss
t
han
e
xisti
ng
arc
hitec
ture
s
[
12
-
16
]
.
The
re
duct
io
n
in
are
a
is
possible
because
S
-
bo
x i
s d
esi
gne
d usi
ng one l
ookup
ta
ble i
n place
of fo
ur lo
okup t
ables.
Figure
11
.
C
om
par
ison
s
of ha
rdwar
e
u
ti
li
zat
ion
for
S
-
box
6
.
1.3.
Op
timi
zed S
NOW 3
G
architec
tu
re
Op
ti
m
iz
ed
SN
O
W
3G
arc
hitec
ture
use
s
re
fine
d
m
od
ulo
CLA
over
2
32
and
ref
in
ed
S
-
bo
x
to
f
or
perform
ance
im
pr
ov
em
ent.
Hardwa
re
re
sources
us
e
d
by
op
ti
m
iz
ed
SN
O
W
3G
arc
hitec
ture
are
pr
e
s
ented
in
Table
3
an
d
T
able
4
sho
ws
c
om
p
arisons
of
hard
war
e
res
ources
us
e
d
by
op
ti
m
iz
ed
SNO
W
3G
a
nd
e
xisti
ng
arch
it
ect
ures
[
12
-
16
]
.
The
com
par
is
on
s
hows
that
op
ti
m
iz
ed
SN
O
W
3G
arc
hitec
ture
util
iz
es
m
ini
m
u
m
resour
ces
as
com
par
ed
to
a
rch
it
ect
ure
pr
e
sented
Kitsos
et
al
.
[
13
]
,
Ma
dan
i
an
d
an
ouga
st
[15]
and
Ma
dan
i
et
al
.
[16]
.
T
he
arc
hitec
ture
pr
ese
nted
by
Kitsos
et
al
.
[
12
]
is
AS
IC,
so
the
com
pa
rison
is
diff
ic
ult.
The
arc
hitec
ture
pr
ese
nted
by
Z
hang
et
al
.
[
14
]
us
es
le
ss
ha
r
dw
a
re
as
com
par
ed
to
pro
po
s
ed
r
efine
d
a
rc
hi
te
ct
ur
e
beca
use
only
on
e
m
od
e im
pl
e
m
ented
on
ha
rdwar
e
.
94
52
55
102
57
57
103
57
55
0
20
40
60
80
100
120
4 i
nput
LUT
s
O
ccupi
ed Sl
i
ces
Sli
ces co
ntainin
g R
elated
Logic
Presented Novel Modulo CLA
Exisist
ing
Conve
nti
ona
l CLA [13]
Exisist
ing
Modif
ie
d
CLA
[22]
534
1152
2048
748
2304
4164
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Proposed Low Frequency
Archi
t
ecture
Proposed High frequency
Archi
t
ecture
Existing Arc
hit
e
ct
ur
es
[12, 13, 14,
15,
16]
N
o. of s
l
ic
es used
N
umber o
f 4
i
np
ut LUTs
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
545
-
557
554
Table
3
.
De
vice u
ti
li
zat
ion
s
um
m
ary of
o
pti
m
iz
ed
SNO
W 3G arc
hitec
tur
e
Dev
ice Utiliz
atio
n
Su
m
m
a
ry (esti
m
at
ed
valu
es)
Log
ic Utilization
Used
Av
ailab
le
Utilizatio
n
1.
Nu
m
b
e
r
o
f
Slice
Reg
ist
ers
870
6
4
0
0
0
1%
2.
Nu
m
b
e
r
o
f
Slice
L
UTs
1208
6
4
0
0
0
1%
3.
Nu
m
b
e
r
o
f
f
u
lly
us
ed
L
UT
-
FF pai
rs
680
1398
48%
4.
Nu
m
b
e
r
o
f
bo
n
d
ed
I
OBs
325
640
50%
5.
Nu
m
b
e
r
o
f
BUFG/
BUFGC
TRLs
10
32
31%
Table
4
.
C
om
par
iso
n of ha
rdwar
e
r
es
ources
for
dif
fer
e
nt arc
hitec
tures
Sr.
No.
Ar
ch
itectu
res
Hardwar
e r
eso
u
rce
s u
sed
1
Prop
o
sed
Ref
in
ed
Architectu
re
8
7
0
Slice
Reg
ister
s an
d
12
0
8
slice L
UTs on
Vir
t
ex
5
2
The ar
ch
itec
tu
re
p
r
o
p
o
sed
by
P.
Kitso
s et al.
[
1
2
]
ASIC i
m
p
le
m
en
ta
t
io
n
us
ed
2501
6
eq
u
iv
alen
t gates
3
The ar
ch
itec
tu
re
p
r
o
p
o
sed
by
P
.
Kitso
s
et al.
[
1
3
]
Slices u
sed
35
5
9
o
n
Spartan
3 Fa
m
ily
4
The ar
ch
itec
tu
re
p
r
o
p
o
sed
by
L.
Z
h
an
g
et
al.
[
1
4
]
On
ly
on
e
m
o
d
e i
m
p
le
m
en
ted
on
hard
ware
to in
c
rease
th
rou
g
h
p
u
t
with
m
in
i
m
u
m
har
d
ware
r
eso
u
rces,
3
5
6
slices o
n
Vir
tex
5
5
The ar
ch
itec
tu
re
p
r
o
p
o
sed
b
y
M
ah
d
i M
ad
an
i and
Ca
m
el
Tano
u
g
ast [
1
5
]
1
0
2
0
Slice
Reg
isters an
d
88
9
Slice
L
UTs on
Vir
t
ex
5
6
The ar
ch
itec
tu
re
p
r
o
p
o
sed
M
ah
d
i M
ad
an
i,
Ily
as
Ben
k
h
ad
d
ra
et al.
[
1
6
]
9
1
2
Slice
Reg
ister
s an
d
11
0
8
Slice
LUT
s o
n
Vir
t
ex
5
6
.
2.
Pr
opagati
on
del
ay
6.2.1.
N
ov
e
l
m
od
ul
o CLA
ov
er 2
32
Pr
opa
gatio
n
de
la
y
com
par
iso
n
of
pr
opos
e
d
ref
ine
d
CLA
a
nd
existi
ng
C
L
A
a
rch
it
ect
ure
s
[
13,
22
]
is
sh
ow
n
in
Fig
ur
e
12
.
Prop
a
gation
delay
evaluati
on
sho
ws
that
delay
of
prese
nted
novel
m
od
ul
o
CL
A
arch
it
ect
ure
is
few
e
r
tha
n
exi
sti
ng
CL
A
ar
c
hitec
tures.
The
pr
ese
nted
CL
A
arc
hitec
ture
will
help
to
im
pro
ve
the th
rou
ghput
of Opti
m
iz
ed
S
NOW
3G ar
chi
te
ct
ur
e.
Figure
12
.
Del
ay
co
m
par
isons o
f novel a
nd
existi
ng CLA
a
rch
it
ect
ures
6.2.2.
N
ov
el
S
-
Box arc
hitect
ure
The
c
om
bin
at
ion
al
path
dela
y
com
par
ison
s
of
pro
pose
d
r
efine
d
S
-
bo
x
arch
it
ect
ures
a
nd
existi
ng
S
-
bo
x
arc
hitec
tures
[12
-
16]
are
sho
wn
i
n
th
e
T
able
5
.
T
he
com
par
iso
n
s
hows
t
hat
the
pro
pag
at
io
n
de
la
y
of
pro
po
se
d
lo
w
-
fr
e
qu
e
ncy
arc
hitec
ture
is
m
or
e
as
com
pa
red
to
oth
e
r
arch
it
ect
ures,
with
le
s
s
hardw
a
re.
Si
m
il
arly
,
the
pro
pag
at
io
n
delay
of
pro
pose
d
hi
gh
-
f
requen
cy
a
rch
it
ec
ture
is
le
ss
as
com
par
ed
to
othe
r
arch
it
ect
ures
with
m
od
erate
har
dwa
re
util
iz
at
ion
.
The
path
delay
of
existi
ng
arc
hi
te
ct
ur
es
is
m
or
e
as
com
par
ed
t
o
pr
esented
arc
hitec
ture
1
bu
t
le
ss
as
com
par
e
d
t
o
arc
hitec
tur
e
2.
T
he
ha
rdwa
re
res
ource
s
us
ed
by
existi
ng arc
hitec
ture
a
re m
or
e
as com
par
ed
to
o
the
r
a
rch
it
ect
ur
es
.
Evaluation Warning : The document was created with Spire.PDF for Python.