Ty

pesofAr
rayPr
ocessor
Ar
rayPr
ocessorper
for
mscomput
ati
onsonl
argear
rayofdat
[Link]
etwo

t
ypes ofAr
ray Pr
ocessor
s:At
tached Ar
ray Pr
ocessor
,and SI
MD Ar
ray

Pr
ocessor
.Thesear
eexpl
ainedasf
oll
owi
ngbel
ow.

[Link]
tachedAr
rayPr
ocessor:
Toi
mpr
ovet
heper
for
manceoft
hehostcomput
eri
nnumer
icalcomput
ati
onal
t
asksauxi
l
iar
ypr
ocessori
sat
tachedt
oit
.

At
tachedar
raypr
ocessorhast
woi
nter
faces:

1.I
nputout
puti
nter
facet
oacommonpr
ocessor
.
2.I
nter
facewi
thal
ocal
memor
y.

Herelocalmemoryint
erconnectsmai nmemor
[Link]
erisgeneral
purposecomputer
.Att
achedpr ocessori
sbackendmachinedri
venbythe
hostcomputer
.
Thearrayprocessori
sconnect
edt hr
oughanI/Ocont
rol
l
ert
othecomput
er&
thecomputertreat
sitasanext
ernali
nter
face.
[Link]:
Thisiscomputerwit
hmul ti
plepr
ocessunitoperat
inginparal
l
elBothty
pesof
arr
ay processor
s,mani pulat
ev ect
ors butt hei
ri nt
ernalorgani
zat
ion i
s
di
ffer
ent.

SI
MDi
sacomput
erwi
thmul
ti
plepr
ocessi
nguni
tsoper
ati
ngi
npar
all
el.
Theprocessi
ngunit
saresy nchroni
zedtoper
formt hesameoper
ationunder
thecontr
olofacommoncont rolunit
.Thusprovidi
ngasingl
ei nstr
ucti
on
str
eam,mul t
ipl
edatast
ream (SIMD)organi
zati
[Link],SIMD
contai
nsasetofi dent
icalprocessi
ngelements(PES)eachhavi
ngal ocal
memor yM.
EachPEi
ncl
udes–
● ALU
● Fl
oat
ingpoi
ntar
it
hmet
icuni
t
● Wor
kingr
egi
ster
s

Mastercontr
olunitcontr
olstheoperati
[Link]
onofmast er
cont
roluniti
st odecodetheinstr
ucti
onanddetermi
nehowt heinst
ruct
ionto
[Link]
nstr
uctionisscal
arorprogr
am contr
olinstr
ucti
onthenit
i
sdirect
lyexecutedwit
hinthemastercontr
oluni
t.
Main memoryis used forst
orage oft
he pr
ogr
am whi
l
e each PE uses
oper
andsst
oredi
ni t
slocalmemory.