<s>
Map	O
is	O
an	O
idiom	O
in	O
parallel	B-Operating_System
computing	I-Operating_System
where	O
a	O
simple	O
operation	O
is	O
applied	O
to	O
all	O
elements	O
of	O
a	O
sequence	O
,	O
potentially	O
in	O
parallel	O
.	O
</s>
<s>
It	O
is	O
used	O
to	O
solve	O
embarrassingly	B-Operating_System
parallel	I-Operating_System
problems	I-Operating_System
:	O
those	O
problems	O
that	O
can	O
be	O
decomposed	O
into	O
independent	O
subtasks	O
,	O
requiring	O
no	O
communication/synchronization	O
between	O
the	O
subtasks	O
except	O
a	O
join	B-Operating_System
or	O
barrier	B-Operating_System
at	O
the	O
end	O
.	O
</s>
<s>
When	O
applying	O
the	O
map	O
pattern	O
,	O
one	O
formulates	O
an	O
elemental	O
function	O
that	O
captures	O
the	O
operation	O
to	O
be	O
performed	O
on	O
a	O
data	O
item	O
that	O
represents	O
a	O
part	O
of	O
the	O
problem	O
,	O
then	O
applies	O
this	O
elemental	O
function	O
in	O
one	O
or	O
more	O
threads	B-Operating_System
of	I-Operating_System
execution	I-Operating_System
,	O
hyperthreads	B-Operating_System
,	O
SIMD	B-General_Concept
lanes	I-General_Concept
or	O
on	O
multiple	B-Architecture
computers	I-Architecture
.	O
</s>
<s>
Some	O
parallel	B-Operating_System
programming	I-Operating_System
systems	O
,	O
such	O
as	O
OpenMP	B-Application
and	O
Cilk	B-Language
,	O
have	O
language	O
support	O
for	O
the	O
map	O
pattern	O
in	O
the	O
form	O
of	O
a	O
parallel	B-Operating_System
for	I-Operating_System
loop	I-Operating_System
;	O
languages	O
such	O
as	O
OpenCL	B-Application
and	O
CUDA	B-Architecture
support	O
elemental	O
functions	O
(	O
as	O
"	O
kernels	B-Operating_System
"	O
)	O
at	O
the	O
language	O
level	O
.	O
</s>
<s>
For	O
example	O
,	O
map	O
combined	O
with	O
category	O
reduction	O
gives	O
the	O
MapReduce	B-Operating_System
pattern	O
.	O
</s>
