<s>
In	O
computing	O
,	O
multiple	B-Operating_System
instruction	I-Operating_System
,	I-Operating_System
multiple	I-Operating_System
data	I-Operating_System
(	O
MIMD	B-Operating_System
)	O
is	O
a	O
technique	O
employed	O
to	O
achieve	O
parallelism	O
.	O
</s>
<s>
Machines	O
using	O
MIMD	B-Operating_System
have	O
a	O
number	O
of	O
processors	O
that	O
function	O
asynchronously	B-Operating_System
and	O
independently	O
.	O
</s>
<s>
MIMD	B-Operating_System
architectures	O
may	O
be	O
used	O
in	O
a	O
number	O
of	O
application	O
areas	O
such	O
as	O
computer-aided	O
design/computer	O
-aided	O
manufacturing	O
,	O
simulation	B-Application
,	O
modeling	O
,	O
and	O
as	O
communication	O
switches	O
.	O
</s>
<s>
MIMD	B-Operating_System
machines	O
can	O
be	O
of	O
either	O
shared	B-Operating_System
memory	I-Operating_System
or	O
distributed	B-Operating_System
memory	I-Operating_System
categories	O
.	O
</s>
<s>
These	O
classifications	O
are	O
based	O
on	O
how	O
MIMD	B-Operating_System
processors	O
access	O
memory	O
.	O
</s>
<s>
Shared	B-Operating_System
memory	I-Operating_System
machines	O
may	O
be	O
of	O
the	O
bus-based	B-Architecture
,	O
extended	O
,	O
or	O
hierarchical	O
type	O
.	O
</s>
<s>
Distributed	B-Operating_System
memory	I-Operating_System
machines	O
may	O
have	O
hypercube	B-Operating_System
or	O
mesh	B-Architecture
interconnection	O
schemes	O
.	O
</s>
<s>
An	O
example	O
of	O
MIMD	B-Operating_System
system	O
is	O
Intel	B-General_Concept
Xeon	I-General_Concept
Phi	I-General_Concept
,	O
descended	O
from	O
Larrabee	B-Architecture
microarchitecture	O
.	O
</s>
<s>
Most	O
parallel	O
computers	O
,	O
as	O
of	O
2013	O
,	O
are	O
MIMD	B-Operating_System
systems	O
.	O
</s>
<s>
In	O
shared	B-Operating_System
memory	I-Operating_System
model	O
the	O
processors	O
are	O
all	O
connected	O
to	O
a	O
"	O
globally	O
available	O
"	O
memory	O
,	O
via	O
either	O
software	O
or	O
hardware	O
means	O
.	O
</s>
<s>
The	O
operating	B-General_Concept
system	I-General_Concept
usually	O
maintains	O
its	O
memory	B-Operating_System
coherence	I-Operating_System
.	O
</s>
<s>
From	O
a	O
programmer	O
's	O
point	O
of	O
view	O
,	O
this	O
memory	O
model	O
is	O
better	O
understood	O
than	O
the	O
distributed	B-Operating_System
memory	I-Operating_System
model	O
.	O
</s>
<s>
Another	O
advantage	O
is	O
that	O
memory	B-Operating_System
coherence	I-Operating_System
is	O
managed	O
by	O
the	O
operating	B-General_Concept
system	I-General_Concept
and	O
not	O
the	O
written	O
program	O
.	O
</s>
<s>
Two	O
known	O
disadvantages	O
are	O
:	O
scalability	O
beyond	O
thirty-two	O
processors	O
is	O
difficult	O
,	O
and	O
the	O
shared	B-Operating_System
memory	I-Operating_System
model	O
is	O
less	O
flexible	O
than	O
the	O
distributed	B-Operating_System
memory	I-Operating_System
model	O
.	O
</s>
<s>
There	O
are	O
many	O
examples	O
of	O
shared	B-Operating_System
memory	I-Operating_System
(	O
multiprocessors	O
)	O
:	O
UMA	O
(	O
uniform	B-Operating_System
memory	I-Operating_System
access	I-Operating_System
)	O
,	O
COMA	O
(	O
cache-only	B-Operating_System
memory	I-Operating_System
access	I-Operating_System
)	O
.	O
</s>
<s>
MIMD	B-Operating_System
machines	O
with	O
shared	B-Operating_System
memory	I-Operating_System
have	O
processors	O
which	O
share	O
a	O
common	O
,	O
central	O
memory	O
.	O
</s>
<s>
This	O
means	O
that	O
every	O
machine	O
with	O
shared	B-Operating_System
memory	I-Operating_System
shares	O
a	O
specific	O
CM	O
,	O
common	O
bus	O
system	O
for	O
all	O
the	O
clients	O
.	O
</s>
<s>
MIMD	B-Operating_System
machines	O
with	O
hierarchical	O
shared	B-Operating_System
memory	I-Operating_System
use	O
a	O
hierarchy	O
of	O
buses	O
(	O
as	O
,	O
for	O
example	O
,	O
in	O
a	O
"	O
fat	B-Protocol
tree	I-Protocol
"	O
)	O
to	O
give	O
processors	O
access	O
to	O
each	O
other	O
's	O
memory	O
.	O
</s>
<s>
In	O
distributed	B-Operating_System
memory	I-Operating_System
MIMD( multiple instruction, multiple data )	O
machines	O
,	O
each	O
processor	O
has	O
its	O
own	O
individual	O
memory	O
location	O
.	O
</s>
<s>
Since	O
there	O
is	O
no	O
shared	B-Operating_System
memory	I-Operating_System
,	O
contention	O
is	O
not	O
as	O
great	O
a	O
problem	O
with	O
these	O
machines	O
.	O
</s>
<s>
Systems	O
were	O
designed	O
to	O
reduce	O
this	O
time	O
loss	O
and	O
hypercube	B-Operating_System
and	O
mesh	B-Architecture
are	O
among	O
two	O
of	O
the	O
popular	O
interconnection	O
schemes	O
.	O
</s>
<s>
Examples	O
of	O
distributed	B-Operating_System
memory	I-Operating_System
(	O
multiple	O
computers	O
)	O
include	O
MPP	O
(	O
massively	B-Operating_System
parallel	I-Operating_System
processors	O
)	O
,	O
COW	O
(	O
clusters	O
of	O
workstations	O
)	O
and	O
NUMA	B-Operating_System
(	O
non-uniform	B-Operating_System
memory	I-Operating_System
access	I-Operating_System
)	O
.	O
</s>
<s>
Examples	O
include	O
hypercube	B-Operating_System
and	O
mesh	B-Architecture
interconnections	O
.	O
</s>
<s>
In	O
an	O
MIMD	B-Operating_System
distributed	B-Operating_System
memory	I-Operating_System
machine	O
with	O
a	O
hypercube	B-Operating_System
system	O
interconnection	O
network	O
containing	O
four	O
processors	O
,	O
a	O
processor	O
and	O
a	O
memory	O
module	O
are	O
placed	O
at	O
each	O
vertex	O
of	O
a	O
square	O
.	O
</s>
<s>
In	O
a	O
hypercube	B-Operating_System
system	O
with	O
eight	O
processors	O
and	O
each	O
processor	O
and	O
memory	O
module	O
being	O
placed	O
in	O
the	O
vertex	O
of	O
a	O
cube	O
,	O
the	O
diameter	O
is	O
3	O
.	O
</s>
<s>
In	O
general	O
,	O
a	O
system	O
that	O
contains	O
2^N	O
processors	O
with	O
each	O
processor	O
directly	O
connected	O
to	O
N	O
other	O
processors	O
,	O
the	O
diameter	O
of	O
the	O
system	O
is	O
N	O
.	O
One	O
disadvantage	O
of	O
a	O
hypercube	B-Operating_System
system	O
is	O
that	O
it	O
must	O
be	O
configured	O
in	O
powers	O
of	O
two	O
,	O
so	O
a	O
machine	O
must	O
be	O
built	O
that	O
could	O
potentially	O
have	O
many	O
more	O
processors	O
than	O
is	O
really	O
needed	O
for	O
the	O
application	O
.	O
</s>
<s>
In	O
an	O
MIMD	B-Operating_System
distributed	B-Operating_System
memory	I-Operating_System
machine	O
with	O
a	O
mesh	B-Architecture
interconnection	O
network	O
,	O
processors	O
are	O
placed	O
in	O
a	O
two-dimensional	O
grid	O
.	O
</s>
<s>
Wrap	O
around	O
connections	O
may	O
be	O
provided	O
at	O
the	O
edges	O
of	O
the	O
mesh	B-Architecture
.	O
</s>
<s>
One	O
advantage	O
of	O
the	O
mesh	B-Architecture
interconnection	O
network	O
over	O
the	O
hypercube	B-Operating_System
is	O
that	O
the	O
mesh	B-Architecture
system	O
need	O
not	O
be	O
configured	O
in	O
powers	O
of	O
two	O
.	O
</s>
<s>
A	O
disadvantage	O
is	O
that	O
the	O
diameter	O
of	O
the	O
mesh	B-Architecture
network	I-Architecture
is	O
greater	O
than	O
the	O
hypercube	B-Operating_System
for	O
systems	O
with	O
more	O
than	O
four	O
processors	O
.	O
</s>
