<s>
In	O
computer	B-General_Concept
architecture	I-General_Concept
,	O
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
(	O
or	O
Barsis	B-Operating_System
's	I-Operating_System
law	I-Operating_System
)	O
gives	O
the	O
speedup	B-Operating_System
in	O
the	O
execution	O
time	O
of	O
a	O
task	O
that	O
theoretically	O
gains	O
from	O
parallel	O
computing	O
,	O
using	O
a	O
hypothetical	O
run	O
of	O
the	O
task	O
on	O
a	O
single-core	O
machine	O
as	O
the	O
baseline	O
.	O
</s>
<s>
It	O
is	O
named	O
after	O
computer	O
scientist	O
John	O
L	O
.	O
Gustafson	O
and	O
his	O
colleague	O
Edwin	O
H	O
.	O
Barsis	O
,	O
and	O
was	O
presented	O
in	O
the	O
article	O
Reevaluating	O
Amdahl	B-Operating_System
's	I-Operating_System
Law	I-Operating_System
in	O
1988	O
.	O
</s>
<s>
Gustafson	O
estimated	O
the	O
speedup	B-Operating_System
of	O
a	O
program	O
gained	O
by	O
using	O
parallel	O
computing	O
as	O
follows	O
:	O
</s>
<s>
is	O
the	O
theoretical	O
speedup	B-Operating_System
of	O
the	O
program	O
with	O
parallelism	O
(	O
scaled	O
speedup	B-Operating_System
)	O
;	O
</s>
<s>
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
addresses	O
the	O
shortcomings	O
of	O
Amdahl	B-Operating_System
's	I-Operating_System
law	I-Operating_System
,	O
which	O
is	O
based	O
on	O
the	O
assumption	O
of	O
a	O
fixed	O
problem	B-General_Concept
size	I-General_Concept
,	O
that	O
is	O
of	O
an	O
execution	O
workload	O
that	O
does	O
not	O
change	O
with	O
respect	O
to	O
the	O
improvement	O
of	O
the	O
resources	O
.	O
</s>
<s>
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
instead	O
proposes	O
that	O
programmers	O
tend	O
to	O
increase	O
the	O
size	O
of	O
problems	O
to	O
fully	O
exploit	O
the	O
computing	O
power	O
that	O
becomes	O
available	O
as	O
the	O
resources	O
improve	O
.	O
</s>
<s>
This	O
gives	O
a	O
linear	O
model	O
between	O
the	O
processor	O
count	O
and	O
the	O
speedup	B-Operating_System
with	O
slope	O
,	O
as	O
shown	O
in	O
the	O
figure	O
above	O
(	O
which	O
uses	O
different	O
notations	O
:	O
for	O
and	O
for	O
)	O
.	O
</s>
<s>
Also	O
,	O
scales	O
linearly	O
with	O
rather	O
than	O
exponentially	O
in	O
the	O
Amdahl	B-Operating_System
's	I-Operating_System
Law	I-Operating_System
.	O
</s>
<s>
The	O
impact	O
of	O
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
was	O
to	O
shift	O
research	O
goals	O
to	O
select	O
or	O
reformulate	O
problems	O
so	O
that	O
solving	O
a	O
larger	O
problem	O
in	O
the	O
same	O
amount	O
of	O
time	O
would	O
be	O
possible	O
.	O
</s>
<s>
After	O
that	O
,	O
another	O
part	O
of	O
the	O
program	O
passes	O
each	O
file	O
to	O
a	O
separate	O
thread	B-Operating_System
for	O
processing	O
.	O
</s>
<s>
Using	O
as	O
the	O
baseline	O
,	O
the	O
speedup	B-Operating_System
for	O
the	O
parallel	O
system	O
is	O
:	O
</s>
<s>
Amdahl	B-Operating_System
's	I-Operating_System
law	I-Operating_System
presupposes	O
that	O
the	O
computing	O
requirements	O
will	O
stay	O
the	O
same	O
,	O
given	O
increased	O
processing	O
power	O
.	O
</s>
<s>
Amdahl	B-Operating_System
's	I-Operating_System
Law	I-Operating_System
reveals	O
a	O
limitation	O
in	O
,	O
for	O
example	O
,	O
the	O
ability	O
of	O
multiple	O
cores	O
to	O
reduce	O
the	O
time	O
it	O
takes	O
for	O
a	O
computer	O
to	O
boot	O
to	O
its	O
operating	O
system	O
and	O
be	O
ready	O
for	O
use	O
.	O
</s>
<s>
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
argues	O
that	O
a	O
fourfold	O
increase	O
in	O
computing	O
power	O
would	O
instead	O
lead	O
to	O
a	O
similar	O
increase	O
in	O
expectations	O
of	O
what	O
the	O
system	O
will	O
be	O
capable	O
of	O
.	O
</s>
<s>
The	O
principal	O
point	O
of	O
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
is	O
that	O
such	O
problems	O
are	O
not	O
likely	O
to	O
be	O
the	O
most	O
fruitful	O
applications	O
of	O
parallelism	O
.	O
</s>
<s>
Algorithms	O
with	O
nonlinear	O
runtimes	O
may	O
find	O
it	O
hard	O
to	O
take	O
advantage	O
of	O
parallelism	O
"	O
exposed	O
"	O
by	O
Gustafson	B-Operating_System
's	I-Operating_System
law	I-Operating_System
.	O
</s>
<s>
Snyder	O
points	O
out	O
an	O
algorithm	O
means	O
that	O
double	O
the	O
concurrency	O
gives	O
only	O
about	O
a	O
26%	O
increase	O
in	O
problem	B-General_Concept
size	I-General_Concept
.	O
</s>
<s>
Furthermore	O
,	O
Woo	O
and	O
Lee	O
studied	O
the	O
implication	O
of	O
energy	O
and	O
power	O
on	O
future	O
many-core	O
processors	O
based	O
on	O
Amdahl	B-Operating_System
's	I-Operating_System
law	I-Operating_System
,	O
showing	O
that	O
an	O
asymmetric	O
many-core	O
processor	O
can	O
achieve	O
the	O
best	O
possible	O
energy	O
efficiency	O
by	O
activating	O
an	O
optimal	O
number	O
of	O
cores	O
given	O
the	O
amount	O
of	O
parallelism	O
is	O
known	O
prior	O
to	O
execution	O
.	O
</s>
<s>
Al-hayanni	O
,	O
Rafiev	O
et	O
al	O
have	O
developed	O
novel	O
speedup	B-Operating_System
and	O
energy	O
consumption	O
models	O
based	O
on	O
a	O
general	O
representation	O
of	O
core	O
heterogeneity	O
,	O
referred	O
to	O
as	O
the	O
normal	O
form	O
heterogeneity	O
,	O
that	O
support	O
a	O
wide	O
range	O
of	O
heterogeneous	O
many-core	O
architectures	O
.	O
</s>
