<s>
The	O
VP2000	O
was	O
the	O
second	O
series	O
of	O
vector	B-Operating_System
supercomputers	B-Architecture
from	O
Fujitsu	O
.	O
</s>
<s>
The	O
VP2000	O
was	O
succeeded	O
in	O
1995	O
by	O
the	O
VPP300	O
,	O
a	O
massively	B-Operating_System
parallel	I-Operating_System
supercomputer	B-Architecture
with	O
up	O
to	O
256	O
vector	B-Operating_System
processors	I-Operating_System
.	O
</s>
<s>
The	O
VP2000	O
was	O
similar	O
in	O
many	O
ways	O
to	O
their	O
earlier	O
designs	O
,	O
and	O
in	O
turn	O
to	O
the	O
Cray-1	B-Device
,	O
using	O
a	O
register-based	O
vector	B-Operating_System
processor	I-Operating_System
for	O
performance	O
.	O
</s>
<s>
For	O
additional	O
performance	O
the	O
vector	B-Operating_System
units	O
supported	O
a	O
special	O
multiply-and-add	O
instruction	O
that	O
could	O
retire	O
two	O
results	O
per	O
clock	O
cycle	O
.	O
</s>
<s>
This	O
instruction	O
"	O
chain	O
"	O
is	O
particularly	O
common	O
in	O
many	O
supercomputer	B-Architecture
applications	O
.	O
</s>
<s>
Another	O
difference	O
is	O
that	O
the	O
main	O
scalar	O
units	O
of	O
the	O
processor	O
ran	O
at	O
half	O
the	O
speed	O
of	O
the	O
vector	B-Operating_System
unit	O
.	O
</s>
<s>
According	O
to	O
Amdahl	B-Operating_System
's	I-Operating_System
Law	I-Operating_System
computers	O
tend	O
to	O
run	O
at	O
the	O
speed	O
of	O
their	O
slowest	O
unit	O
,	O
and	O
in	O
this	O
case	O
unless	O
the	O
program	O
spent	O
most	O
of	O
its	O
time	O
in	O
the	O
vector	B-Operating_System
units	O
,	O
the	O
slower	O
scalar	O
performance	O
would	O
make	O
it	O
1/2	O
the	O
performance	O
of	O
a	O
Cray-1	B-Device
at	O
the	O
same	O
speed	O
.	O
</s>
<s>
Due	O
to	O
the	O
additional	O
load/store	O
units	O
,	O
adding	O
additional	O
scalar	O
units	O
improved	O
performance	O
by	O
increasing	O
memory	O
bandwidth	O
,	O
as	O
well	O
as	O
allowing	O
several	O
programs	O
to	O
run	O
at	O
the	O
same	O
time	O
and	O
thereby	O
increase	O
the	O
chance	O
there	O
was	O
something	O
to	O
process	O
on	O
the	O
vector	B-Operating_System
unit	O
.	O
</s>
<s>
The	O
machines	O
were	O
supplied	O
with	O
either	O
the	O
Unix-compatible	O
UXP/M	O
or	O
the	O
MVS-compatible	O
VSP/S	O
operating	B-General_Concept
systems	I-General_Concept
,	O
both	O
supplied	O
by	O
Amdahl	O
.	O
</s>
<s>
The	O
later	O
was	O
used	O
for	O
Fortran	B-Application
programs	O
while	O
the	O
former	O
was	O
typically	O
used	O
for	O
C	B-Language
,	O
and	O
vectorizing	B-General_Concept
compilers	I-General_Concept
were	O
supplied	O
for	O
both	O
languages	O
.	O
</s>
<s>
Like	O
most	O
companies	O
,	O
Fujitsu	O
turned	O
to	O
massive	B-Operating_System
parallelism	I-Operating_System
for	O
future	O
machines	O
,	O
and	O
the	O
VP2000	O
family	O
were	O
not	O
on	O
the	O
market	O
for	O
very	O
long	O
.	O
</s>
