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.
10
,
No.
3
,
June
2020
,
pp. 2
951~
2958
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
3
.
pp2951
-
29
58
2951
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Pip
eline
d Vedic
mu
ltip
lier w
i
th manifold
ad
der
co
mp
lexi
ty
l
evels
An
siy
a
Es
ha
c
k,
S
. K
ri
s
hnakumar
Depa
rtment
o
f
E
le
c
troni
cs,
Scho
ol
of Te
chnol
og
y
and
Appl
ie
d
Scie
n
ce
s,
Eda
p
all
y
,
Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
y
13
, 201
9
Re
vised
Jan
3
,
20
20
Accepte
d
Ja
n 8
, 20
20
Rec
en
tly
,
th
e
in
c
rea
sed
us
e
of
por
ta
bl
e
dev
ices,
h
a
s
drive
n
the
rese
arc
h
world
to
design
s
y
ste
m
s
with
low
power
-
consum
pti
on
and
high
thro
ughput.
Vedi
c
m
ult
ipl
ie
r
provi
des
le
ast
dela
y
eve
n
in
com
ple
x
m
ult
ip
li
c
ations
when
compare
d
to
oth
er
conv
ent
ion
al
m
ult
i
pli
ers
.
In
th
is
pape
r
,
a
64
-
bi
t
m
ult
iplier
is
creat
ed
using
the
Urdhav
a
Ti
r
y
akbh
y
am
su
tra
in
Vedi
c
m
at
hemat
ic
s.
The
d
esign
of
t
his
64
-
bit
m
ulti
pli
er
is
implem
ent
ed
in
f
ive
d
i
ffe
ren
t
wa
y
s
with
the
pipe
l
ining
conc
ept
app
lied
at
diffe
r
ent
st
age
s
of
adde
r
co
m
ple
xit
ie
s
.
The
diffe
r
ent
ar
chi
t
ec
tur
es
show
diffe
ren
t
de
lay
and
power
c
onsum
pti
on.
It
is
noti
ce
d
t
hat
as
comple
xity
o
f
adde
rs
in
the
m
ult
ip
li
ers
red
uc
e,
the
s
y
s
te
m
s
show
improved
spee
d
and
l
ea
st
h
ard
ware
uti
l
izati
o
n
.
The
arc
hi
te
c
ture
designe
d
using
2
x
2
-
bit
p
ipe
l
in
ed
Vedi
c
m
ult
ip
li
er
is
then
compare
d
with
exi
sting
Vedi
c
m
ult
ipl
ie
rs
and
conve
nt
iona
l
m
ult
ipliers
and
show
s le
ast
d
el
a
y
.
Ke
yw
or
d
s
:
FPGA
Lo
w
p
ower
Pipeli
ning
Ved
ic
m
ulti
pliers
Urdhava
Tirya
kbhyam
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
An
siy
a E
sh
ac
k
Dep
a
rtm
ent o
f El
ect
ro
nics
,
School
of Tec
hnol
og
y a
nd
Applie
d
Scie
nces,
Eda
pally
, Erna
ku
la
m
6
8202
4, I
nd
ia
Em
a
il
: ansiy
a
@yah
oo.co
m
1.
INTROD
U
CTION
Sp
ee
d
a
nd
po
wer
co
nsum
pti
on
are
the
t
wo
m
a
in
aspects
consi
der
e
d
w
hi
le
desig
ning
a
syst
e
m
in
the
fiel
d
of
co
m
m
un
ic
at
ion
[
1]
.
The
m
ulti
p
li
ers
(
m
or
e
sp
e
ci
fical
ly
the
add
ers
)
w
hich
f
orm
the
m
ajo
r
pa
rt
of
these
syst
e
m
s
aff
ect
it
s
sp
eed
[2]
.
The
m
or
e
com
plex
a
m
ul
ti
plier
or
it
s
relat
ed
add
e
r
is,
the
m
or
e
is
i
ts
eff
ect
on
the
s
peed
[
3]
.
Althou
gh
a
dd
it
io
n
operati
on
s
a
re
si
m
ple,
add
it
ion
of
la
r
ge
bits
will
su
r
el
y
increase
the
tim
e
require
d
f
or
ge
ner
at
io
n
of
t
he
ou
t
pu
t.
As
nu
m
ber
of
bits
of
add
e
nds
inc
re
ases,
ti
m
e
to
com
plete
add
it
ion
al
s
o
incr
eases
,
ulti
m
at
ely
le
ading
to
decr
ea
se
in
thr
ough
pu
t
of
the
m
ulti
plier.
DSPs
a
nd
c
om
m
un
ic
at
ion
s
yst
e
m
s
rely
heav
il
y
on
ad
de
rs
an
d
m
ul
ti
pliers
for
processi
ng
t
he
ir
data
[4
,
5]
.
Existi
ng
m
ulti
plier
syst
em
s
c
on
s
ume
m
or
e p
ower
and tim
e and
are
thu
s
, not s
uitab
le
f
or
DS
P
and
co
m
m
un
ic
at
ion
syst
em
s
[6]
.
This p
ape
r
e
xplores
ways
to
r
edu
ce
the b
it
siz
e
of
the
a
dden
ds
a
nd
eve
ntu
a
ll
y,
reduce
the co
m
plexit
y
of
t
he
a
dd
e
rs
in
the
m
ulti
pliers
to
inc
rea
se
it
s
processi
ng
sp
ee
d
a
nd
lowe
r
the
p
ower
c
ons
um
pt
ion
of
the
syst
em
.
T
he
use
of
a
dders
with
dif
fere
nt
com
plexiti
es
show
s
that
as
ad
der
s
be
com
e
le
ss
com
plex,
the
sp
ee
d
of
th
e
sy
stem
increases.
T
he
desi
gn
e
d
64
-
bit
Ve
dic
m
ult
ipli
ers
(V
M)
,
usi
ng
U
rdha
va
Tiry
ak
bh
ya
m
(U
T
)
Su
t
ra,
em
plo
y
pip
el
ini
ng
at
five
diff
ere
nt
add
e
r
com
plexiti
es
and
the
desig
n
with
th
e
lowest
com
pl
exity
achieves
hi
gh
e
st
thr
oughput.
The
five
desi
gns
i
nclu
de
th
e
us
e
of
4
thi
rty
-
two
-
bit,
16
si
xt
een
-
bit,
64
ei
ght
-
bit,
256
four
-
bit
and
1024
t
wo
-
bi
t
pip
el
ined
V
Ms
resp
ect
i
vely
.
A
re
d
uctio
n
in
the
F
PGA
hard
war
e
us
e
d
is
al
so
ob
s
er
ved
a
s
V
M
are
em
plo
ye
d
in
the
desi
gned
syst
em
.
Faste
r
outp
ut
ge
ne
rati
on
w
it
h
lo
w
hard
war
e u
ti
li
zat
ion
certai
nly
bri
ng
s
do
wn
po
wer
co
n
s
um
ption
of
this
novel
s
yst
e
m
.
Thu
s
,
t
he
desig
ne
d
m
ulti
plier
syst
em
s
are
low
po
wer
co
nsum
ing
and
fa
s
t
ou
tp
ut
gen
e
ra
tors
with
inc
re
ase
in
sp
eed
a
nd
re
du
ct
io
n
in
powe
r
co
ns
um
ption
as
the
c
om
plexity
of
the
ad
de
rs
decr
ease
s.
T
he
rest
of
pa
pe
r
is
as
f
ollows:
Re
searc
h
m
eth
od
is
hi
gh
li
gh
te
d
i
n
Sect
ion
2
with
ex
planati
on
on
VMs
,
Pi
pelined
VMs
,
an
d
Mod
i
fied
VM
s.
Sect
io
n
3
gi
ves
the
res
ults
an
d
analy
sis. Finall
y, the c
on
cl
us
i
on is
giv
e
n
in
Sect
ion
4.
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.
10
, No
.
3
,
J
une
2020
:
29
51
-
2958
2952
2.
RESEA
R
CH MET
HO
D
2.1.
Vedic m
ultipl
ie
r
In
the
c
onve
nt
ion
al
m
ulti
plier
s,
‘t
he
car
ry’
m
os
t
of
te
n
ge
ts
prop
a
gated
al
l
the
way
t
o
the
m
os
t
sign
ific
a
nt
bit
,
wh
e
n
bi
nar
y
m
ulti
plica
ti
on
is
perform
ed
[7]
.
This
is
the
m
aj
or
r
easo
n
f
or
hi
gh
e
r
delay
in
these
m
ul
ti
pliers.
Wh
en
Ved
ic
m
ulti
plica
ti
on
te
chn
iq
ue
li
ke
UT
Su
tra
is
us
e
d,
there
is
par
ti
al
product
ge
nerat
ion
and
these
ca
n
be
ad
de
d
up
in
a
pi
peline
d
fash
i
on
[
8,
9]
.
So,
VM
e
m
plo
yi
ng
the
UT
te
ch
niqu
e
f
or
m
ul
ti
plica
ti
on
of
tw
o
nu
m
bers
is
see
n
t
o
produce
outp
uts
faster
t
han
othe
r
m
ulti
pliers
by
re
duci
ng
t
h
e
delay
to
gen
e
rate
the
pr
od
uct
[10]
.
This
te
chn
i
que
us
es
‘V
e
rtic
al
and
Cr
os
s
wise’
m
e
tho
dolo
gy
wh
ic
h
util
iz
es
le
ast
delay
,
and
al
lows
f
or
low
h
ar
dware
us
a
ge
du
ri
ng
proc
essing
[
11]
.
Thu
s
UT
Su
t
ra
al
lows
for
pa
rall
el
processi
ng and
provides
b
et
te
r per
form
ance
[12
–
14]
.
VMs,
i
nvolv
i
ng
th
e
UT
Su
tr
a,
al
lo
w
a
higher
ord
er
bit
m
ul
ti
plica
ti
on
to
be
c
om
pu
te
d
by
br
ea
ki
ng
them
to
lowe
r
order
bits.
G
ener
al
ly
,
an
N
xN
-
bit
VM
is
form
ed
by
four
N/2xN/
2
-
bi
t
VMs
[15]
.
E
ach
of
the
N/2xN/
2
-
bi
t
VM
can
again
be
f
or
m
ed
by
fo
ur
N/4x
N/4
-
bit
VMs
[16
]
.
Thu
s
eac
h
hi
gh
e
r
bit
VM
can
be
form
ed
by
four
l
ow
e
r
bit
VMs
with
t
he
lowe
r
m
ulti
pl
ie
rs
ha
ving
ha
lf
the
bit
siz
e
of
th
e
hi
gh
e
r
on
es
.
This
deco
m
posit
ion
ca
n
c
onti
nu
e
un
ti
l
2x
2
-
bit
VMs
a
r
e
reac
hed,
be
yond
wh
ic
h
de
com
po
sit
ion
is
no
t
po
s
sible.
T
his
is
as
show
n
i
n
F
igure 1
.
A
2x
2
-
bit
UT
S
utra
base
d
VM, h
a
vi
ng
a 2
-
bit
m
ul
ti
plier
11
a
nd
a
2
-
bit
m
ul
ti
plica
nd
11, f
ollow
s
the
s
te
ps
s
how
n
in
F
igure
2
[17]
.
Figure
1. Dec
om
po
sit
ion
of
hig
he
r orde
r bit
V
Ms i
nt
o
lo
we
r order
b
it
VMs
Figure
2. UT
S
utra base
d 2x2
-
bit
VM
2.2.
Desi
gn
e
d s
ystem
This
sect
ion
de
scrib
es
five
a
rch
it
ect
ures
of
the
64x
64
-
bit
VM.
The
diag
ram
m
at
ic
rep
resen
ta
ti
on
of
this i
s g
i
ven in
F
igure
3(
a
)
t
o
(
e).
T
he five
arc
hitec
tures wit
h dif
fer
e
nt
bit siz
ed pipeli
ne
d VM are
as
fo
ll
ow
s:
a.
Mul
t
ipli
er w
it
h f
our 32x
32
-
bit
pip
el
ine
d VM
b.
Mult
ipli
er w
it
h si
xteen
16
x16
-
bit
pip
el
ine
d VM
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
Pipel
ined
Ve
dic mult
ipli
er wi
th
manifold
ad
de
r com
plexi
ty
leve
ls (
Ansiya Esh
ack
)
2953
c.
Mult
ipli
er w
it
h si
xty
-
f
our 8
x8
-
bit
pip
el
ine
d VM
d.
Mul
t
ipli
er w
it
h t
wo
-
fi
fty
-
six
4x4
-
bit pipeli
ne
d VM
e.
Mul
t
ipli
er w
it
h o
ne
-
t
hous
a
nd
-
twenty
-
fo
ur 2x
2
-
bit
pi
pelined
VM
(a)
(b)
(c)
(d)
Figure
3
.
(a
)
M
ul
ti
plier w
it
h f
our 3
2x32
-
bit
pip
el
ine
d
VMs
,
(b)
Mul
t
ipl
ie
r
with si
xteen 1
6x16
-
bit
pip
el
in
ed
VMs,
(c)
M
ul
t
ipli
er
with si
xty
-
f
our
8x8
-
bit
pip
el
i
ned V
Ms
,
(d)
Mul
t
ipli
er
with tw
o
-
fifty
-
six
4x4
-
bit
pi
pe
li
ned
VMs
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.
10
, No
.
3
,
J
une
2020
:
29
51
-
2958
2954
Figure
3
.
(e
)
M
ul
t
ipli
er
with
one
-
t
hous
a
nd
-
t
wen
ty
-
fou
r 2x2
-
bit pi
pelined
VMs
(
c
onti
nue
)
These
desig
ne
d
m
ulti
pliers
thu
s
us
e
t
he
UT
S
utra
c
om
bin
ed
with
the
pi
pelinin
g
te
chn
i
qu
e
.
The
desig
ns
a
r
e
create
d
by
usi
ng
t
he
pi
peline
m
et
ho
dolo
gy
to
pro
du
ce
PMs
at
diff
e
re
nt
m
ulti
plier
s
ta
ges.
It
is
ob
ser
ve
d
that
as
the
bi
t
siz
e
of
these
PM
beco
m
es
low,
le
adin
g
to
le
sser
com
plexity
of
the
add
e
rs
involve
d
in
th
e
m
,
the
tim
e
fo
r
t
hro
ughput
al
so
dec
rea
ses
.
The
64x6
4
-
bi
t
VM
desi
gn
e
d
us
in
g
the
2x2
-
bit
pip
el
ine
d
VM
is
seen
t
o
yi
el
d
the
fastest
r
esults
w
he
n
c
om
pa
red
with
tho
se
desi
gn
e
d
us
in
g
4x4
-
bit
PMs,
8x8
-
bit
PMs
,
16x1
6
-
bit
PMs
or
32
x32
-
bit
P
Ms.
As
the
bit
siz
e
of
the
PM
dec
reases,
the
ad
der
s
bec
ome
le
ss
and less c
om
pl
ex
e
ven
t
ually
leadin
g
to
f
ast
e
r
th
rou
ghput.
2.3.1
.
Pipel
ined Ve
dic
m
ulti
pli
er
The
‘
V
e
rtic
al
and
C
r
os
s
wise’
te
ch
nique
fo
l
lowe
d
by
UT
Su
tra
is
as
s
ho
wn
in
F
ig
ur
e
2.
T
he
bits
of
the
input
nu
m
ber
s
are
m
ulti
plied
ve
rtic
al
ly
and
al
so
in
a
cr
os
s
wise
m
ann
er
in
dif
fer
e
nt
ste
ps
a
nd
the
co
ncatena
ti
on
of
the
pa
rtia
l
pr
od
ucts
ob
ta
in
ed
in
these
ind
i
vid
ual
ste
ps
le
ad
t
o
the
ou
tpu
t
of
the
m
ulti
pliers
[18]
.
It
can
be
ob
s
er
ved
that
t
he
UT
Su
t
ra
s
uppo
rts
pi
pelini
ng
an
d
this
m
et
hod
is
f
ollow
e
d
for
the
faster
outp
ut
ge
ner
at
io
n
of
the
m
ulti
pliers
[19
–
21]
.
T
hu
s,
outp
uts
in
m
ulti
pliers
em
pl
oying
t
he
UT
Su
tra
are
go
t
faster t
han no
n
-
pip
e
li
ned m
ulti
pliers
[
22
]
.
The ge
ner
al
ste
ps
fo
ll
owe
d by the
UT Sutra
for a
2
x
2
-
bit P
M i
s as foll
ows
[
23
]
:
Inp
ut 1 = a
2
a
1
;
Input
2
=
b
2
b
1
Step
1(ver
ti
cal
): a
1
* b
1
= a
1
b
1
P1 ;
Step
2 (cr
os
s
-
wise): a
1
* b
2
+
a
2
* b
1
= a
1
b
2
+
a
2
b
1
P
2;
Step
3(ver
ti
cal
): a
2
* b
2
= a
2
b
2
P
3
Ou
t
pu
t:
P
3 P2
P1
The ge
ner
al
ste
ps
fo
ll
owe
d
in
this te
ch
niques
for
a
4 x 4
-
bit PM i
s as
fo
ll
ows:
Inp
ut 1 = a
4
a
3
a
2
a
1
; Inp
ut 2 =
b
4
b
3
b
2
b
1
Step
1(ver
ti
cal
): a
1
* b
1
= a
1
b
1
P1 ;
Step
2(
c
ross
-
w
ise
): a
1
* b
2
+ a
2
*
b
1
=
a
1
b
2
+ a
2
b
1
P2
;
Step
3(
c
ross
-
w
ise
): a
1
* b
3
+ a
2
*
b
2
+
a
3
* b
1
= a
1
b
3
+ a
2
b
2
+
a
3
b
1
P3
;
Step
4(
c
ross
-
w
ise
): a
1
* b
4
+ a
2
*
b
3
+
a
3
* b
2
+
a
4
* b
1
= a
1
b
4
+
a
2
b
3
+ a
3
b
2
+
a
4
b
1
P4
;
Step
5(
c
ross
-
w
ise
): a
2
* b
4
+ a
3
* b
3
+ a
4
* b
2
=
a
2
b
4
+ a
3
b
3
+
a
4
b
2
P5
;
Step
6(
c
ross
-
w
ise
):
a
3
*
b
4
+
a
4
*
b
3
=
a
3
b
4
+
a
4
b
3
P6
;
Step
7(ver
ti
cal
): a
4
* b
4
= a
4
b
4
P7
Ou
t
pu
t:
P
7 P6
P5
P4 P
3
P
2 P
1
The ge
ner
al
ste
ps
fo
ll
owe
d
in
this te
ch
niques
for
a
8 x 8
-
bit PM i
s as
fo
ll
ows:
Inp
ut 1 = a
8
a
7
a
6
a
5
a
4
a
3
a
2
a
1
; Inpu
t
2
=
b
8
b
7
b
6
b
5
b
4
b
3
b
2
b
1
Step
1(ver
ti
cal
): a
1
* b
1
= a
1
b
1
P1
;
Step
2(
c
ross
-
w
ise
): a
1
* b
2
+ a
2
*
b
1
=
a
1
b
2
+ a
2
b
1
P2
;
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
Pipel
ined
Ve
dic mult
ipli
er wi
th
manifold
ad
de
r com
plexi
ty
leve
ls (
Ansiya Esh
ack
)
2955
Step
3(
c
ross
-
w
ise
): a
1
* b
3
+ a
2
*
b
2
+
a
3
* b
1
= a
1
b
3
+ a
2
b
2
+
a
3
b
1
P3
;
Step
4(
c
ross
-
w
ise
): a
1
* b
4
+ a
2
*
b
3
+
a
3
* b
2
+
a
4
* b
1
= a
1
b
4
+
a
2
b
3
+ a
3
b
2
+
a
4
b
1
P4
;
Step
5(
c
ross
-
w
ise
): a
1
* b
5
+ a
2
*
b
4
+
a
3
* b
3
+ a
4
* b
2
+ a
5
*
b
1
= a
1
b
5
+ a
2
b
4
+ a
3
b
3
+ a
4
b
2
+
a
5
b
1
P5
;
Step
6(
c
ross
-
w
ise
): a
1
* b
6
+ a
2
*
b
5
+
a
3
* b
4
+ a
4
* b
3
+ a
5
*
b
2
+ a
6
* b
1
= a
1
b
6
+ a
2
b
5
+ a
3
b
4
+ a
4
b
3
+
a
5
b
2
+ a
6
b
1
P
6;
Step
7(
c
r
os
s
-
w
ise
):
a
1
*
b
7
+
a
2
*
b
6
+
a
3
*
b
5
+
a
4
*
b
4
+
a
5
*
b
3
+
a
6
*
b
2
+
a
7
*
b
1
=
a
1
b
7
+
a
2
b
6
+
a
3
b
5
+
a
4
b
4
+ a
5
b
3
+ a
6
b
2
+ a
7
b
1
P
7;
Step
8(
c
ro
ss
-
w
ise
): a
1
*
b
8
+ a
2
*
b
7
+ a
3
* b
6
+ a
4
*
b
5
+ a
5
* b
4
+ a
6
*
b
3
+ a
7
*
b
2
+ a
8
* b
1
=
a
1
b
8
+ a
2
b
7
+ a
3
b
6
+ a
4
b
5
+
a
5
b
4
+
a
6
b
3
+ a
7
b
2
+ a
8
b
1
P
8;
Step 9(cr
os
s
-
w
ise
): a
2
*
b
8
+ a
3
*
b
7
+ a
4
*
b
6
+ a
5
*
b
5
+ a
6
* b
4
+ a
7
*
b
3
+ a
8
*
b
2
= a
2
b
8
+ a
3
b
7
+ a
4
b
6
+
a
5
b
5
+ a
6
b
4
+ a
7
b
3
+ a
8
b
2
P
9;
Step
10
(cross
-
wise):
a
3
*
b
8
+
a
4
*
b
7
+
a
5
*
b
6
+
a
6
*
b
5
+
a
7
*
b
4
+
a
8
*
b
3
=
a
3
b
8
+
a
4
b
7
+
a
5
b
6
+
a
6
b
5
+
a
7
b
4
+ a
8
b
3
P
10
;
Step
11
(cross
-
wise):
a
4
*
b
8
+
a
5
*
b
7
+
a
6
*
b
6
+
a
7
*
b
5
+
a
8
*
b
4
=
a
4
b
8
+
a
5
b
7
+
a
6
b
6
+
a
7
b
5
+
a
8
b
4
P11;
Step
12(cross
-
wise): a
5
* b
8
+
a
6
* b
7
+ a
7
* b
6
+ a
8
* b
5
= a
5
b
8
+ a
6
b
7
+ a
7
b
6
+ a
8
b
5
P
12
;
Step
13(cross
-
wise): a
6
* b
8
+
a
7
* b
7
+ a
8
* b
6
= a
6
b
8
+ a
7
b
7
+ a
8
b
6
P
13
;
Step
14(cross
-
wise): a
7
* b
8
+
a
8
* b
7
= a
7
b
8
+ a
8
b
7
P
14
;
Step
15(
ver
ti
ca
l): a
8
* b
8
= a
8
b
8
P1
5
Ou
t
pu
t:
P
15 P
14 P
13 P
12 P
11 P
10 P
9 P8
P
7
P
6 P5
P
4
P
3 P2 P1
The
sam
e
con
cept
is
app
li
ed
to
f
or
m
16
x16
-
bit
an
d
32x3
2
-
bit
PMs.
It
sh
ould
be
no
te
d
tha
t
the
pip
el
ine
d
VMs
are
no
t
c
r
eat
ed
from
fo
ur
lowe
r
order
bi
t
m
ult
ipli
ers
as
is
the
case
wi
th
a
ge
ner
al
V
M,
i.e.,
an
N
x
N
-
bit
pip
el
ined
VM
ju
st
fo
ll
ows
the
ver
ti
cal
an
d
cr
os
s
-
wi
se
ste
ps
as
li
ste
d
abov
e
and
is
not
buil
t
by
four N/
2
-
bit V
Ms.
2.3.2
.
M
od
ifie
d V
e
dic
m
ulti
pli
er
Ther
e
are
t
wo
ste
ps
f
ollo
we
d
in
m
ulti
plication:
pa
rtia
l
pro
du
ct
gen
e
r
at
ion
a
nd
pa
rtia
l
pr
od
uct
accum
ulati
on
[24,
25]
.
T
he
changes
in
t
he
processin
g
m
et
hod
of
these
two
ste
ps
res
ult
in
differe
nc
e
in
thr
oughput
of
t
he
m
ulti
plier.
I
n
the
desi
gn
e
d
VMs,
t
he
m
eth
od
of
pa
rtia
l
pro
du
ct
acc
umulat
ion
is
e
xploit
ed,
as
descr
i
bed
be
low,
t
o
gai
n
f
ast
er
outp
uts.
Con
si
der
a
4x4
-
bit
VM
f
or
m
ed
f
ro
m
fo
ur
2x2
-
bit
VMs.
T
he
tw
o
4
–
bit
nu
m
ber
s
wr
it
te
n
as
a
4
a
3
a
2
a
1
and
b
4
b
3
b
2
b
1
ca
n
be
pa
rtit
ion
e
d
into
f
ou
r
2
-
bit
nu
m
ber
s
a
4
a
3
,
a
2
a
1
,
b
4
b
3
an
d
b
2
b
1
.
T
hese
2
-
bit
nu
m
ber
s
se
rv
e
as
the
in
pu
ts
to
the
2x2
-
bi
t
VMs.
The
cr
os
s
-
wise
an
d
ve
rtic
al
te
chn
iq
ue
of
UT
S
utra
is
use
d
to
g
ene
rate
the
pro
duct
s
of
these
2x2
-
bit
VMs.
T
he
pa
rt
ia
l
pr
od
ucts
(PP)
of
the
4x
4
-
bi
t
VM
is
as
sh
ow
n
in
F
igure
4.
T
hes
e
par
ti
al
pro
duct
s
wh
ic
h
a
re
f
our
-
bit
wide
ea
ch
are
the
n
a
dded
t
og
et
her
t
o
form
the f
i
nal prod
uc
t.
Figure
4. Mo
dified 4
x
4
-
bit
V
M wit
h
t
he gen
erated
par
ti
al
pro
duct
s and its
accum
ulati
on
4
x
4
b
it
VM
2
x
2
b
it
VM
2
x
2
b
it
VM
2
x
2
b
it
VM
2
x
2
b
it
VM
PP4
=
a
4
a
3
*b
4
b
3
PP3
=
a
4
a
3
*b
2
b
1
PP2
=
a
2
a
1
*b
4
b
3
PP1
=
a
2
a
1
*b
2
b
1
P
P4
(2:1
)
PP3(4
:3
)
PP3
(2:1
)
PP2
(2:1
)
PP1(4
:3
)
PP2
(4:3
)
PP4
(4:3
)
PP1
(2:1
)
Ad
d
er
P
8
-
7
P
6
-
5
P
4
-
3
P
2
-
1
Ad
d
er
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.
10
, No
.
3
,
J
une
2020
:
29
51
-
2958
2956
The
m
od
ifie
d
4x
4
-
bit
VM
is
thu
s
co
ns
tr
uc
te
d
f
ro
m
four
2x2
-
bit
VMs
by
us
i
ng
the
pr
i
nciple
of
pro
du
c
t
accum
ulati
on
as
gi
ve
n
belo
w.
T
he
f
our
pa
rtia
l
pr
oducts
PP
1
-
PP
4
and
the
fi
nal
pro
duct
can
be
wr
it
te
n
as foll
ows:
PP1
=
a
2
a
1
* b
2
b
1
PP2
=
a
2
a
1
* b
4
b
3
PP3
=
a
4
a
3
* b
2
b
1
PP4
=
a
4
a
3
*
b
4
b
3
------
------
------
------
------
------
------
------
------
------
------
---
Final p
r
oduct
=
P
8
-
7
P
6
-
5
P
4
-
3
P
2
-
1
------
------
------
------
------
------
------
------
------
------
------
---
wh
e
re
P
8
-
7
=
P
P4
(
4:
3)
,
the
fir
st
two
bits
of
PP4,
P
6
-
5
=
PP
4(2:1
)
+
PP
3(4
:3)
+
PP
2(4:3
),
the
s
um
of
th
e
first
two
bits
of
P
P
3
an
d
PP2
&
t
he
la
st
two
bits
of
PP
4,
P
4
-
3
=
PP3
(
2:1)
+
PP2(2
:
1)
+
PP
1(4:3
),
the
sum
of
first
two
bits
of
PP
1
an
d
la
st
tw
o
bits
of
PP
3
&
PP2
a
nd
P
2
-
1
=
PP1(2
:
1),
the
l
ast
two
bits
of
PP1.
T
he
car
ry
of
P
4
-
3
is
add
ed
to
the
su
m
of
P
6
-
5
and
that
of
P
6
-
5
is
add
e
d
to
P
8
-
7
.
Mod
ifie
d
8
x8
-
bit,
16x1
6
-
bit
and
32x3
2
-
bit
VMs
al
l
fo
ll
ow
the
s
a
m
e
pr
inci
ple
of
pa
rtia
l
pro
duct
accum
ulati
on
f
or
fin
al
pr
oduct
gen
e
rati
on.
T
his
m
od
if
ic
at
ion
al
lows
for fa
st
er
gen
e
rati
on
of the
prod
ucts
and th
us
reduc
es the
delay
of
the syst
em
.
3.
RESU
LT
S
AND A
N
ALYSIS
Fiv
e
di
ff
e
ren
t
arch
it
ect
ures
of
64
x
64
-
bit
V
Ms,
with
var
yi
ng
bit
siz
es
of
the
pip
el
in
ed
VMs,
hav
e
been
desig
ne
d
and
im
ple
m
ent
ed.
T
he
desi
gns
con
sist
of
m
od
ifie
d
VMs
an
d
pip
el
i
ned
V
Ms
at
diff
ere
nt
add
e
r
com
plexiti
es.
The
bit
siz
e
of
the
pi
peline
d
VMs
ha
s
a
dir
ect
relat
ion
with
the
c
om
plexity
of
the
a
dd
e
rs
a
nd
the
thr
ough
pu
t
of
the
syst
em
.
It
is
obse
rv
e
d
that
as
m
ulti
pli
er
an
d
t
he
a
dd
e
r
com
plexity
r
edu
ce
s,
the
del
ay
for
gen
e
rati
ng
the
product
al
so
decr
eases
.
F
igure
5
gi
ves
the
delay
fo
r
the
five
di
ff
ere
nt
desi
gns
of
the 64
x64
–
bit
m
ul
ti
plier archi
te
ct
ur
es.
Figure
5. Dela
y for the
f
i
ve d
iffer
e
nt arc
hite
ct
ur
es
of
64x6
4
-
bit VM
The
num
ber
of
lo
ok
-
up
ta
bles
(L
UTs
),
of
the
FP
GA,
util
iz
ed
by
the
five
dif
fere
n
t
pip
el
ine
d
arch
it
ect
ures of 64
x
64
-
bit V
M are as g
i
ven in Tab
le
1.
It i
s see
n
that t
he
arch
it
ect
ure wi
th 32x
32
-
bit pi
peline
d
VM u
ti
li
zes the h
igh
e
st nu
m
ber
o
f
L
UTs
. The u
ti
li
zat
ion
o
f LUTs
decr
ease
s as the
le
vel o
f
pip
el
ini
ng
re
du
ce
s
and
the
m
ultip
li
er
wit
h
2x
2
-
bit
pi
peline
d
VM
ha
s
the
l
east
nu
m
ber
of
L
UTs.
The
reason
for
i
nc
reased
thr
oughpu
t
of
the
arc
hitec
tur
es
with
decr
ea
se
in
bit
siz
e
of
pi
pelined
V
M
is
du
e
t
o
th
e
fact
that
as
bit
siz
e
reduces,
the
c
om
plexity
of
add
e
rs
al
so
de
creases.
The
vo
l
um
e
of
addi
ti
on
s
an
d
the
siz
e
of
the
add
e
nds
decr
ease
s
a
nd
this
le
ads
t
o
de
crease
in
the
tim
e
to
produc
e
the
ou
t
pu
t.
This
is
t
he
rea
so
n
f
or
re
duct
ion
in
the
nu
m
ber
of
LUTs
us
ed
by
the
m
ult
ipli
er
desig
ns
.
T
he
de
la
y
of
the
des
ign
e
d
m
ulti
plier
syst
e
m
e
m
pl
oying
2x
2
-
bit
pip
el
in
ed
VM
th
us
yi
el
ds
the
best
re
su
lt
in
te
rm
s
of
ha
rdwar
e
util
iz
at
ion
of
FP
G
A
a
nd
s
yst
em
sp
ee
d.
This
arc
hitec
ture
desig
n
is
co
m
par
ed
with
ot
her
e
xisti
ng
64
x
64
-
bit
VMs
as
sh
ow
n
in
Fig
ur
e
6.
The
desi
gn
e
d
64x
64
-
bit
m
ult
ipli
er
syst
e
m
with
le
ast
adder
com
plexity
,
i.
e.
e
m
plo
yi
ng
the
2x
2
-
bit
pip
el
ine
d
VM,
sh
ow
s
the
le
ast
delay
wh
e
n
c
om
par
ed
w
it
h
t
he
exi
sti
ng
m
ulti
plier
s.
Fi
gure
7
s
hows
t
he
delay
com
par
ison
be
tween
the con
ven
ti
onal
m
ulti
pliers an
d t
he desi
gn
e
d
m
ulti
plier syst
e
m
.
Table
1.
L
UT
util
iz
at
ion
of t
he five
dif
fer
e
nt ar
c
hitec
tures
of
64x64
-
bit
VM
Ty
p
es o
f
VM
3
2
bit
p
ip
elin
ed
VM
1
6
bit
p
ip
elin
ed
VM
8
bit
p
ip
elin
ed
VM
4
bit
p
ip
elin
ed
VM
2
bit
p
ip
elin
ed
VM
No
.
o
f
L
o
o
k
-
u
p
tables
1
3
1
5
1
1
2
8
0
5
1
2
2
5
5
1
1
9
0
5
1
1
3
7
8
1
5
1
.
7
2
ns
7
4
.
3
4
ns
4
2
.
9
4
ns
3
3
.
7
7
ns
3
2
.
0
8
ns
20
40
60
80
1
0
0
1
2
0
1
4
0
1
6
0
3
2
bi
t
pi
peline
d
VM
1
6
bi
t
pi
peline
d
VM
8
bi
t
pi
peli
ned
VM
4
bi
t
pi
peli
ned
VM
2
bi
t
pi
peli
ned
VM
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
Pipel
ined
Ve
dic mult
ipli
er wi
th
manifold
ad
de
r com
plexi
ty
leve
ls (
Ansiya Esh
ack
)
2957
Figure
6. Dela
y com
par
ison
of d
es
i
gn
e
d sy
stem
w
it
h
existi
ng 64
x
64
-
bit
VMs
Figure
7. Dela
y com
par
ison
of d
esi
gn
e
d sy
stem
with
conve
ntion
al
64
x
64
-
bit m
ultip
li
ers
4.
CONCL
US
I
O
N
The
pa
pe
r
pro
po
s
es
the
desi
gn
of
five
syst
e
m
arch
it
ect
ures
of
the
64x
64
-
bit
VM.
The
desig
ns
sho
w
that
the
syst
e
m
design
ed
usi
ng
t
he
2x
2
-
bit
pip
el
ine
d
m
od
ifie
d
VM
give
s
the
highest
thr
oughput.
Also
thi
s
2
x2
-
bit
pip
el
in
ed
m
od
ifie
d
V
M
desig
n
util
iz
es
le
ast
am
ou
nt
of
FP
GA
ha
rdwar
e
out
of
al
l
the
fi
ve
de
signs
.
This p
r
oves
th
at
as
the b
it
siz
e
of
the
pi
pelin
ed
VM
re
duces
,
the
c
om
plexity
of
the
syst
em
re
du
ces
,
a
nd
t
his
in
tur
n
increa
ses
it
s
co
m
pu
ta
ti
on
s
pee
d
a
nd
decr
ease
s
it
s
powe
r
co
nsum
ption.
Th
e
bes
t
arch
it
ect
ur
e
a
m
on
g
the
fi
ve
desig
ns
of
the
64x
64
-
bit
VM
ha
s
bee
n
c
om
pared
with
oth
e
r
existi
ng
sim
i
lar
bit
-
siz
ed
V
Ms
an
d
sta
nd
a
rd m
ultip
li
ers
a
nd
is
observ
e
d
t
o hav
e
the lea
st delay
.
REFERE
NCE
S
[1]
M.
Jham
b,
Gari
m
a,
and
H.
Lohani,
“
Design,
imp
le
m
ent
a
ti
on
and
per
form
anc
e
co
m
par
ison
of
m
ult
ipl
i
er
topo
logi
e
s
in
power
-
de
lay
spac
e,”
In
te
rnati
onal
Journal
En
gin
ee
ring
Scienc
e
and
Te
chnol
o
gy,
v
o
l.
19,
no.
1,
pp.
355
–
363
,
Mar
2016.
[2]
C.
Chaitan
y
a
,
C
.
Sundare
san
,
P.
V
enka
t
eswara
n,
and
K.
Prasad
,
“
As
ic
design
of
low
power
-
dela
y
produ
ct
car
r
y
pre
-
computation
base
d
m
ult
ip
lie
r,
”
Indon
esian
J
ournal
of
E
lectr
ic
al
Engi
n
ee
ring
and
Computer
Sci
en
ce
,
vol
.
13,
no.
2
,
pp
.
845
–
8
52,
2019
.
[3]
T.
Gupt
a
and
J.
B.
Sharm
a,
“
Han
–
Carl
son
add
er
base
d
h
igh
-
spe
ed
Vedi
c
m
ultip
li
er
for
complex
m
ult
ipl
i
cation,”
Mic
ros
yste
m Technologie
s
,
vol
.
2
4,
no
.
9
,
pp
.
390
1
–
3906,
Sep
201
8.
[4]
V.
K.
Rao
and
K.
La
van
y
a
,
“
An
are
a
eff
i
ci
en
t
Q
-
form
at
m
ult
ipl
ie
r
with
h
igh
p
erf
orm
anc
e
for
digi
tal
proc
essing
appl
i
c
ations,”
in
IEEE
Asia
Pa
c
if
ic
Conf
ere
nc
e
on
Postgraduate
Re
search
in
Mic
roel
ec
troni
cs
and
El
ec
troni
cs
(
PrimeA
sia)
,
pp.
137
–
141,
2015.
[5]
M.
Bara
kat,
W
.
Saad,
and
M.
Shokair,
“
Im
ple
m
ent
a
ti
on
of
eff
i
c
ie
nt
m
ult
iplier
f
or
high
spee
d
appl
icati
ons
usin
g
FP
GA
,
”
in
13th
Inte
rnational
Co
nfe
renc
e
on
Co
mputer
Engi
n
ee
r
ing
and
S
yste
ms
(
ICCES
)
,
pp.
21
1
–
214,
2018
.
[6]
K.
D.
Rao
,
C
.
G
anga
dhar
,
and
P.
K.
Korra
i,
“
FPGA
implementa
t
ion
of
compl
ex
m
ult
ipl
ie
r
using
m
ini
m
um
del
a
y
Vedic
real
m
ult
ipl
i
er
arc
h
it
e
cture,
”
in
IE
EE
Utta
r
Pradesh
Sec
t
ion
Inte
rnat
ional
Confe
renc
e
on
El
ectric
al
,
Computer
and
E
le
c
tronic
s E
ng
in
ee
ring (
UPCON)
,
pp.
580
–
584
,
2016.
[7]
E.
Prabhu
,
H.
Manga
la
m
,
and
P.
R.
Gokul
,
“
A
del
a
y
eff
icien
t
Vedic
m
ult
ipli
er,
”
Proceedi
ng
s
of
the
Na
ti
on
al
Ac
ademy
of
Scie
nce
s,
India
S
ecti
on
A: Phy
si
cal
S
ci
en
ce
s
,
vol. 89
,
no
.
2
,
pp
.
257
–
2
68,
Jun
2019
.
[8]
Y.
S.
Rao,
M.
Kam
ara
ju,
and
D.
V.
S.
Ramanja
n
e
y
u
lu,
“
An
FP
G
A
implementa
ti
o
n
of
high
spee
d
and
area
eff
i
ci
en
t
double
-
pre
ci
sion
floa
ti
ng
poin
t
m
ult
ipl
ie
r
usin
g
Urdhva
Ti
r
y
a
gbh
y
am
techniq
ue,
”
in
Con
fe
re
nce
on
Powe
r,
Control,
Comm
u
nic
ati
on
and
Co
mputati
onal
Tec
hnologi
es
for
Su
stainabl
e
Gr
owth
(
PCCCTSG
)
,
pp.
2
7
1
-
2
7
6
,
2
0
1
5
.
[9]
A.
Sai
Ram
y
a
,
B.
S.
S.
V
.
Ram
esh
Babu,
E.
Sri
kal
a
,
M.
Pav
an,
P.
Unita,
and
A.
V.
S.
Sw
at
hi
,
“
Perform
anc
e
o
f
o
pti
m
iz
ed
rev
ersibl
e
Vedic
m
ult
i
pli
ers,”
in
Comp
.
Comm
.
,
Ne
tw
.
and
Int
.
Se
c
.
,
pp
.
587
–
593,
2017.
[10]
S.
G.
M.
E.
At
re
and
M.
A.
M.
Alshewim
y
,
“
Design
and
implementa
ti
on
of
new
del
a
y
-
ef
fic
i
ent
/
conf
igur
a
ble
m
ult
ipl
ie
r
using
FP
GA
,
”
in
12th
Int
.
Con
fe
ren
ce
on
Computer
En
gine
ering
and
S
y
stems (
ICCES)
,
pp.
8
–
13
,
2017
.
[11]
C.
Chai
t
an
y
a
,
C.
Sundare
san,
P.
Venka
te
sw
ara
n
,
and
K.
Prasad,
“
Design
of
m
odifi
ed
booth
base
d
m
ult
ipl
ie
r
wi
th
ca
rr
y
p
re
-
compu
ta
ti
on
,
”
Ind
.
Jou
r
.
of El
ec
.
l Eng
.
and
Comp
.
S
ci
.
,
vo
l.
13
,
no
.
3
,
pp
.
1048
–
1055
,
20
19.
[12]
S.
L.
G.
Mos
es
and
M.
Thi
la
g
ar,
“
VHDL
implem
ent
a
ti
on
of
high
per
form
anc
e
RC
6
al
gorit
hm
using
anc
ie
n
t
Indian
V
edi
c
m
at
h
emat
ic
s,”
in
3rd Int
ernati
onal
Con
fe
re
nce
on
E
le
c
troni
cs
Computer
Te
c
hnology
,
vo
l
.
4
,
p
p
.
1
4
0
–
143,
2
0
1
1
.
[13]
P.
Tuwa
nuti
a
nd
N.
Thongb
ai
,
“
Im
ple
m
ent
at
ion
of
V
edic
m
ult
ipl
i
er
t
echnique
on
m
ul
ti
cor
e
proc
essor,”
in
TENCON
201
4
-
2014
IEEE R
egi
on
10
Confe
r
enc
e
,
pp
.
1
–
6
,
2
014.
[14]
K.
Pichhode,
M.
D.
Pat
il
,
D
.
Shah,
and
B.
C
.
Rohit
,
“
FP
GA
implementa
t
ion
of
eff
i
ci
en
t
V
e
dic
m
ult
iplier
,
”
in
Int
ernati
onal
Confe
renc
e
on
I
nformation
Proc
essing
(
ICIP)
,
pp.
565
–
570
,
201
5.
[15]
T.
Gupta
and
J.
B.
Sharm
a,
“
A
C
SA
-
base
d
arc
hitect
u
re
of
Vedic
m
ult
ipl
i
er
for
complex
m
ul
ti
pli
c
at
ion
,
”
in
Ambi
ent
Com
municat
ions a
nd
Computer
Syst
e
ms
,
pp.
41
–
52
,
2
018.
[16]
Jinesh
S,
Rames
h
P,
and
J.
Thomas,
“
I
m
ple
m
ent
at
ion
o
f
64Bit
h
igh
spee
d
m
ult
ipl
ie
r
for
DS
P
app
li
c
at
ion
-
base
d
o
n
V
edi
c
m
at
h
emat
ic
s,”
in
TENCON
2015
-
2015
I
EE
E
Re
g
ion
10
Confe
renc
e
,
pp.
1
–
5,
2015
.
44
.
87
n
s
40
.
33
n
s
36
.
68
n
s
32
.
08
n
s
30
36
42
48
M.
Ram
alat
ha
[1
7
]
V
.
K.
Rao [4
]
M.
Yuva
r
a
j
[2
4
]
Des
i
gn
ed
V
M
wit
h 2x2
bi
t
46
.
11
n
s
36
.
55
n
s
32
.
08
n
s
30
36
42
48
Modi
f
i
e
d Bo
oth
Al
gorit
hm [1
0
]
Ar
r
ay
Mu
l
t
i
p
l
i
er
[1
5
]
Des
i
gn
ed
V
M w
i
t
h
2
x
2
bi
t
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.
10
, No
.
3
,
J
une
2020
:
29
51
-
2958
2958
[
17]
M.
Ramala
th
a,
K.
D.
Da
y
al
an
,
P.
Dhara
ni,
and
S.
D.
Pri
y
a
,
“
High
spee
d
ene
rg
y
eff
ic
i
ent
ALU
d
esign
using
Vedi
c
m
ult
ipl
icati
on
technique
s,
”
in
20
09
Inte
rnationa
l
Confe
renc
e
on
A
dvanc
es
in
Com
putat
ional
Tools
for
Engi
ne
ering
Appl
ic
a
ti
ons
,
pp.
600
–
603,
200
9.
[18]
R.
K.
Bari
k,
M.
Pradha
n,
an
d
R.
Panda,
“
Ti
m
e
eff
icien
t
signed
Vedic
m
ult
ipl
ie
r
using
red
undant
bin
ar
y
rep
rese
nt
at
ion
,
”
The
Journal
o
f E
ngine
ering
,
vol
.
2017,
no
.
3
,
pp
.
60
–
68,
2017
.
[19]
V.
Ja
y
apr
ak
asa
n
,
S.
Vi
jay
akuma
r,
and
V.
S.
Ka
ncha
na
Bhaa
ska
ra
n,
“
Ev
al
ua
ti
on
of
the
conv
ent
i
onal
vs.
ancie
n
t
computat
ion
m
e
thodol
og
y
for
e
ner
g
y
eff
ic
i
ent
ari
thmet
ic
arc
h
itect
ur
e
,
”
in
Inte
r
nati
onal
Con
fe
r
enc
e
on
Proc
ess
Aut
omation
,
Co
ntrol
and
Comp
uti
ng
,
pp.
1
–
4,
2
011.
[20]
K.
Morghade
an
d
P.
Dakhole,
“
Design
of
fast
V
edi
c
m
ultipli
er
with
fau
l
t
di
agn
ostic
c
apa
bi
li
t
ie
s
,
”
in
Inte
rnat
ion
al
Confe
renc
e
on
C
omm
unic
ati
on
a
nd
Signal P
roc
e
ss
ing
,
pp.
0416
–
0419,
2016
.
[21]
A.
Eshac
k
and
S
.
Krishnakum
ar,
“
Low
p
ower
32
x
32
–
bi
t
r
ev
er
sible
Vedi
c
m
ultipli
er
,
”
Inte
rnat
i
onal
Journal
o
f
Innov
ative
Te
ch
nology
and
E
xplor
ing
Engi
n
ee
ri
ng
(
IJI
TEE)
,
vol
.
8
,
no
.
10
,
Aug
2019.
[22]
V.
Gow
ree
sriniva
s
and
P.
Sam
undiswary
,
“
Resourc
e
eff
ic
i
en
t
single
pre
c
isio
n
floa
ti
ng
poin
t
m
ult
ipl
ie
r
usin
g
Kara
tsuba al
gor
i
th
m
,
”
Ind
.
Jour
.
of
E
le
c
.
Eng
.
an
d
Inf
.
(
IJE
EI)
,
v
ol.
6
,
no
.
3
,
pp
.
3
33
–
342,
2018
.
[23]
A.
Jais
and
P.
P
al
sodkar,
“
Design
and
implementa
ti
on
of
64
bit
m
ult
iplier
using
V
edi
c
a
lgori
thm,”
in
Inte
rnational
Confe
renc
e
on
C
omm
unic
ati
on
a
nd
Signal P
roc
e
ss
ing
(
ICCSP)
,
pp.
0775
–
0779
,
2
016.
[24]
M.
Yuva
raj,
B.
J.
Kailath,
and
N.
Bhaskhar
,
“
Design
of
opti
m
iz
ed
MA
C
uni
t
using
int
egr
at
ed
V
edi
c
m
ultipli
e
r,
”
in
Int
ernati
onal
conf
ere
n
ce on
M
ic
roel
ec
troni
c
D
ev
i
ce
s,
Circuits
and
Syste
ms
(
ICMDCS)
,
pp.
1
–
6,
2017.
[25]
A.
Eshac
k
and
S
.
Krishnakum
ar,
“
Reve
rsible
lo
g
i
c
in
pipelined l
o
w power
Vedic
m
ult
ipl
ie
r
,”
Indo
nesian
Journal
o
f
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Computer
Sc
ie
nc
e
,
v
ol
.
16
,
no
.
3
,
pp
.
1265
–
1272,
De
c
2019.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
An
siy
a
Esh
ack
rec
e
ive
d
her
B
.
Te
ch
from
MES
Coll
ege
of
Enginee
ring
,
Kutti
pu
ram
and
M.T
ech
from
Model
Engi
nee
r
ing
Coll
e
ge,
Kochi
.
She
is
pre
sently
an
As
socia
te
Prof
essor
at
KM
EA
Engi
ne
eri
ng
Col
le
ge
,
Erna
ku
la
m
.
She
is
working
on
her
PhD
in
El
e
ct
roni
cs
at
M
aha
tma
Gandhi
Univer
sit
y
,
Kot
tay
am.
Her
res
ea
rch
in
te
r
ests
inc
lud
e
Low
Pow
er
VLSI
De
sign,
Embedde
d
S
y
stems
,
Com
m
unic
a
ti
on
S
y
ste
m
s a
nd
DS
P Ap
pli
c
at
ions.
S.
Kr
ish
nak
um
ar
complet
ed
hi
s
M.Sc.
in
Phy
s
ic
s
with
El
e
ct
ro
nic
s
spec
ialization
in
1987
fro
m
Maha
tma
Gand
hi
Univer
sit
y
,
K
ott
a
y
am.
He
wa
s
awa
rde
d
Ph.D
.
in
Thi
n
Film
Devic
es
in
199
5
from
Maha
tma
Gandhi
Univer
si
t
y
,
Kot
tay
am.
He
re
ce
iv
ed
h
is
M.T
e
ch.
in
Com
pu
te
r
Sc
ie
nc
e
from
Alla
hab
ad
Agric
ult
ura
l
Inst
it
ut
e
(Dee
m
ed
Univer
sit
y
)
in
2006
and
his
MCA
fro
m
I
GN
OU
in
2010.
He
serve
d
as
th
e
Regi
on
al
Dir
e
ct
or
a
t
the
Scho
ol
of
Technol
og
y
a
nd
Applie
d
S
ci
en
ce
s
(STAS
),
during
the
p
eri
o
d
2014
-
17.
Curr
ent
l
y
h
e
is
working
in
the
ST
AS
,
Maha
tma
Gandhi
Univer
s
i
t
y
Resea
rch
Cen
tre,
Eda
pa
lly
.
His
r
ese
arc
h
in
te
r
est
fie
lds
include
AN
N,
VLSI,
Analog
ci
rcu
i
t
design
and
Im
age
proc
essing.
Dr.
S.
Krishnakum
ar
is
an
As
socia
te
m
ember
of
Insti
tut
e
of
Engi
n
ee
r
s,
India
.
He
was
a
m
ember
of
Boar
d
of
studie
s
of
Univer
sit
y
of
C
a
li
cut
and
a
m
ember
of
Ac
ade
m
ic
counc
i
l
of
Maha
tma
Gandhi
Uni
ver
sit
y
,
Kott
a
y
a
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.