<s>
For	O
several	O
years	O
parallel	O
hardware	O
was	O
only	O
available	O
for	O
distributed	B-Architecture
computing	I-Architecture
but	O
recently	O
it	O
is	O
becoming	O
available	O
for	O
the	O
low	O
end	O
computers	O
as	O
well	O
.	O
</s>
<s>
It	O
is	O
quite	O
natural	O
for	O
programmers	O
to	O
think	O
sequentially	O
and	O
hence	O
they	O
are	O
less	O
acquainted	O
with	O
writing	O
multi-threaded	B-Operating_System
or	O
parallel	B-Operating_System
processing	I-Operating_System
applications	O
.	O
</s>
<s>
Parallel	B-Operating_System
programming	I-Operating_System
requires	O
handling	O
various	O
issues	O
such	O
as	O
synchronization	O
and	O
deadlock	B-Operating_System
avoidance	O
.	O
</s>
<s>
The	O
need	O
is	O
also	O
non-trivial	O
because	O
large	O
amount	O
of	O
legacy	O
code	O
written	O
over	O
the	O
past	O
few	O
decades	O
needs	O
to	O
be	O
reused	O
and	O
parallelized	B-Operating_System
.	O
</s>
<s>
Past	O
techniques	O
provided	O
solutions	O
for	O
languages	O
like	O
FORTRAN	B-Application
and	O
C	B-Language
;	O
however	O
,	O
these	O
are	O
not	O
enough	O
.	O
</s>
<s>
These	O
techniques	O
dealt	O
with	O
parallelization	B-Operating_System
sections	O
with	O
specific	O
system	O
in	O
mind	O
like	O
loop	O
or	O
particular	O
section	O
of	O
code	O
.	O
</s>
<s>
Identifying	O
opportunities	O
for	O
parallelization	B-Operating_System
is	O
a	O
critical	O
step	O
while	O
generating	O
multithreaded	O
application	O
.	O
</s>
<s>
This	O
need	O
to	O
parallelize	O
applications	O
is	O
partially	O
addressed	O
by	O
tools	O
that	O
analyze	O
code	O
to	O
exploit	O
parallelism	B-Operating_System
.	O
</s>
<s>
These	O
tools	O
use	O
either	O
compile	B-Application
time	I-Application
techniques	O
or	O
run-time	B-Library
techniques	O
.	O
</s>
<s>
The	O
compiler	O
identifies	O
these	O
language	O
constructs	O
and	O
analyzes	O
the	O
marked	O
code	O
for	O
parallelization	B-Operating_System
.	O
</s>
<s>
Hence	O
a	O
fully	O
automatic	O
tool	O
for	O
converting	O
sequential	O
code	O
to	O
parallel	B-Operating_System
code	I-Operating_System
is	O
required	O
.	O
</s>
<s>
The	O
process	O
starts	O
with	O
identifying	O
code	O
sections	O
that	O
the	O
programmer	O
feels	O
have	O
parallelism	B-Operating_System
possibilities	O
.	O
</s>
<s>
The	O
next	O
stage	O
is	O
to	O
shortlist	O
code	O
sections	O
out	O
of	O
the	O
identified	O
ones	O
that	O
are	O
actually	O
parallelization	B-Operating_System
.	O
</s>
<s>
Sometimes	O
the	O
dependencies	O
are	O
removed	O
by	O
changing	O
the	O
code	O
and	O
this	O
is	O
the	O
next	O
stage	O
in	O
parallelization	B-Operating_System
.	O
</s>
<s>
The	O
last	O
stage	O
in	O
parallelization	B-Operating_System
is	O
generating	O
the	O
parallel	B-Operating_System
code	I-Operating_System
.	O
</s>
<s>
but	O
has	O
additional	O
constructs	O
or	O
code	O
sections	O
which	O
when	O
executed	O
create	O
multiple	O
threads	B-Operating_System
or	O
processes	O
.	O
</s>
<s>
See	O
also	O
main	O
article	O
automatic	O
parallelization	B-Operating_System
.	O
</s>
<s>
There	O
are	O
a	O
number	O
of	O
Automatic	B-Operating_System
Parallelization	I-Operating_System
tools	I-Operating_System
for	O
Fortran	B-Application
,	O
C	B-Language
,	O
C++	B-Language
,	O
and	O
several	O
other	O
languages	O
.	O
</s>
<s>
It	O
takes	O
input	O
as	O
C	B-Language
source	O
code	O
which	O
may	O
have	O
multiple	O
source	O
and	O
header	O
files	O
.	O
</s>
<s>
It	O
gives	O
output	O
as	O
transformed	O
multi-threaded	B-Operating_System
parallel	B-Operating_System
code	I-Operating_System
using	O
pthreads	B-Operating_System
functions	O
and	O
OpenMP	B-Application
constructs	O
.	O
</s>
<s>
The	O
YUCCA	O
tool	O
does	O
task	O
and	O
loop	O
level	O
parallelization	B-Operating_System
.	O
</s>
<s>
Par4All	O
is	O
an	O
automatic	O
parallelizing	O
and	O
optimizing	B-Application
compiler	I-Application
(	O
workbench	O
)	O
for	O
C	B-Language
and	O
Fortran	B-Application
sequential	O
programs	O
.	O
</s>
<s>
source-to-source	B-Language
compiler	I-Language
is	O
to	O
adapt	O
existing	O
applications	O
to	O
various	O
hardware	O
targets	O
such	O
as	O
multicore	O
systems	O
,	O
high	B-Architecture
performance	I-Architecture
computers	I-Architecture
and	O
GPUs	B-Architecture
.	O
</s>
<s>
Cetus	O
is	O
written	O
in	O
Java	B-Language
.	O
</s>
<s>
It	O
provides	O
basic	O
infrastructure	O
for	O
writing	O
automatic	B-Operating_System
parallelization	I-Operating_System
tools	I-Operating_System
or	O
compilers	O
.	O
</s>
<s>
A	O
new	O
graphic	B-Application
user	I-Application
interface	I-Application
(	O
GUI	B-Application
)	O
was	O
added	O
in	O
Feb	O
2013	O
.	O
</s>
<s>
Speedup	B-Operating_System
calculations	O
and	O
graph	O
display	O
were	O
added	O
in	O
May	O
2013	O
.	O
</s>
<s>
A	O
Cetus	O
remote	O
server	O
in	O
a	O
client	B-Operating_System
–	I-Operating_System
server	I-Operating_System
model	I-Operating_System
was	O
added	O
in	O
May	O
2013	O
and	O
users	O
can	O
optionally	O
transform	O
C	B-Language
Code	O
through	O
the	O
server	O
.	O
</s>
<s>
PLUTO	O
is	O
an	O
automatic	B-Operating_System
parallelization	I-Operating_System
tool	I-Operating_System
based	O
on	O
the	O
polyhedral	O
model	O
.	O
</s>
<s>
The	O
polyhedral	O
model	O
for	O
compiler	B-Application
optimization	I-Application
is	O
a	O
representation	O
for	O
programs	O
that	O
makes	O
it	O
convenient	O
to	O
perform	O
high-level	O
transformations	O
such	O
as	O
loop	O
nest	O
optimizations	O
and	O
loop	O
parallelization	B-Operating_System
.	O
</s>
<s>
Pluto	O
transforms	O
C	B-Language
programs	I-Language
from	O
source	O
to	O
source	O
for	O
coarse-grained	O
parallelism	B-Operating_System
and	O
data	O
locality	O
simultaneously	O
.	O
</s>
<s>
OpenMP	B-Application
parallel	B-Operating_System
code	I-Operating_System
for	O
multicores	O
can	O
be	O
automatically	O
generated	O
from	O
sequential	O
C	B-Language
program	I-Language
sections	O
.	O
</s>
<s>
The	O
Polaris	O
compiler	O
takes	O
a	O
Fortran77	O
program	O
as	O
input	O
,	O
transforms	O
this	O
program	O
so	O
that	O
it	O
runs	O
efficiently	O
on	O
a	O
parallel	B-Operating_System
computer	I-Operating_System
,	O
and	O
outputs	O
this	O
program	O
version	O
in	O
one	O
of	O
several	O
possible	O
parallel	O
FORTRAN	B-Application
dialects	O
.	O
</s>
<s>
In	O
addition	O
to	O
many	O
commonly	O
known	O
passes	O
,	O
Polaris	O
includes	O
advanced	O
capabilities	O
performing	O
the	O
following	O
tasks	O
:	O
Array	O
privatization	O
,	O
Data	B-Operating_System
dependence	I-Operating_System
testing	O
,	O
Induction	O
variable	O
recognition	O
,	O
Inter	O
procedural	O
analysis	O
,	O
and	O
symbolic	O
program	O
analysis	O
.	O
</s>
<s>
The	O
auto-parallelization	O
feature	O
of	O
the	O
Intel	B-Language
C++	I-Language
Compiler	I-Language
automatically	O
translates	O
serial	O
portions	O
of	O
the	O
input	O
program	O
into	O
semantically	O
equivalent	O
multi-threaded	B-Operating_System
code	O
.	O
</s>
<s>
Automatic	O
parallelization	B-Operating_System
determines	O
the	O
loops	O
that	O
are	O
good	O
work	O
sharing	O
candidates	O
,	O
</s>
<s>
programming	O
with	O
OpenMP	B-Application
directives	O
.	O
</s>
<s>
The	O
is	O
a	O
vectorization	O
optimization	O
and	O
thread	B-Operating_System
prototyping	O
tool	O
.	O
</s>
<s>
is	O
a	O
tool	O
which	O
can	O
automatically	O
insert	O
OpenMP	B-Application
pragmas	O
into	O
input	O
serial	O
C/C	O
++	O
codes	O
.	O
</s>
<s>
For	O
input	O
programs	O
with	O
existing	O
OpenMP	B-Application
directives	O
,	O
the	O
tool	O
will	O
double	O
check	O
the	O
correctness	O
when	O
the	O
right	O
option	O
is	O
turned	O
on	O
.	O
</s>
<s>
Compared	O
to	O
conventional	O
tools	O
,	O
AutoPar	O
can	O
incorporate	O
user	O
knowledge	O
(	O
semantics	O
)	O
to	O
discover	O
more	O
parallelization	B-Operating_System
opportunities	O
.	O
</s>
<s>
This	O
tool	O
provides	O
users	O
with	O
the	O
assistance	O
needed	O
for	O
OpenMP	B-Application
parallelization	B-Operating_System
of	O
a	O
sequential	O
program	O
.	O
</s>
<s>
All	O
the	O
activities	O
related	O
to	O
program	O
parallelization	B-Operating_System
,	O
such	O
as	O
selecting	O
a	O
target	O
portion	O
of	O
the	O
program	O
,	O
invoking	O
an	O
assistance	O
command	O
,	O
and	O
modifying	O
the	O
program	O
based	O
on	O
the	O
assistance	O
information	O
shown	O
by	O
the	O
tool	O
,	O
can	O
be	O
handled	O
in	O
the	O
source	O
program	O
editor	O
environment	O
.	O
</s>
<s>
SUIF	O
is	O
a	O
fully	O
functional	O
compiler	O
that	O
takes	O
both	O
Fortran	B-Application
and	O
C	B-Language
as	O
input	O
languages	O
.	O
</s>
<s>
The	O
parallelized	B-Operating_System
code	O
is	O
output	O
as	O
an	O
SPMD	B-Operating_System
(	O
Single	B-Operating_System
Program	I-Operating_System
Multiple	I-Operating_System
Data	I-Operating_System
)	O
parallel	O
C	B-Language
version	O
of	O
the	O
program	O
that	O
can	O
be	O
compiled	O
by	O
native	O
C	B-Language
compilers	O
on	O
a	O
variety	O
of	O
architectures	O
.	O
</s>
<s>
It	O
translates	O
C	B-Language
and	O
Fortran	B-Application
programs	O
with	O
OpenMP	B-Application
pragmas	O
into	O
C	B-Language
code	O
suitable	O
for	O
compiling	O
with	O
a	O
native	O
compiler	O
linked	O
with	O
the	O
Omni	O
OpenMP	B-Application
runtime	B-Library
library	O
.	O
</s>
<s>
It	O
does	O
for	O
loop	O
parallelization	B-Operating_System
.	O
</s>
<s>
It	O
uses	O
a	O
simulation	O
based	O
approach	O
to	O
improve	O
task	O
allocation	O
and	O
task	O
parallelization	B-Operating_System
to	O
multiple	O
cores	O
.	O
</s>
<s>
The	O
compiler	O
extracts	O
better	O
locality	O
with	O
fine	O
-	O
and	O
coarse-grained	O
parallelism	B-Operating_System
for	O
C/C	O
++	O
applications	O
.	O
</s>
<s>
SequenceL	B-Operating_System
is	O
a	O
general-purpose	O
functional	O
programming	O
language	O
and	O
auto-parallelizing	O
tool	O
set	O
,	O
whose	O
primary	O
design	O
objectives	O
are	O
performance	O
on	O
multi-core	O
processor	O
hardware	O
,	O
ease	O
of	O
programming	O
,	O
platform	O
portability/optimization	O
,	O
and	O
code	O
clarity	O
and	O
readability	O
.	O
</s>
<s>
Its	O
main	O
advantage	O
is	O
that	O
it	O
can	O
be	O
used	O
to	O
write	O
straightforward	O
code	O
that	O
automatically	O
takes	O
full	O
advantage	O
of	O
all	O
the	O
processing	O
power	O
available	O
,	O
without	O
programmers	O
needing	O
to	O
be	O
concerned	O
with	O
identifying	O
parallelisms	B-Operating_System
,	O
specifying	O
vectorization	O
,	O
avoiding	O
race	O
conditions	O
,	O
and	O
other	O
challenges	O
of	O
manual	O
directive-based	O
programming	O
approaches	O
such	O
as	O
OpenMP	B-Application
.	O
</s>
<s>
Programs	O
written	O
in	O
SequenceL	B-Operating_System
can	O
be	O
compiled	O
to	O
multithreaded	O
code	O
that	O
runs	O
in	O
parallel	O
,	O
with	O
no	O
explicit	O
indications	O
from	O
a	O
programmer	O
of	O
how	O
or	O
what	O
to	O
parallelize	O
.	O
</s>
<s>
As	O
of	O
2015	O
,	O
versions	O
of	O
the	O
SequenceL	B-Operating_System
compiler	O
generate	O
parallel	B-Operating_System
code	I-Operating_System
in	O
C++	B-Language
and	O
OpenCL	O
,	O
which	O
allows	O
it	O
to	O
work	O
with	O
most	O
popular	O
programming	O
languages	O
,	O
including	O
C	B-Language
,	O
C++	B-Language
,	O
C#	O
,	O
Fortran	B-Application
,	O
Java	B-Language
,	O
and	O
Python	O
.	O
</s>
<s>
A	O
platform-specific	O
runtime	B-Library
manages	O
the	O
threads	B-Operating_System
safely	O
,	O
automatically	O
providing	O
parallel	O
performance	O
according	O
to	O
the	O
number	O
of	O
cores	O
available	O
.	O
</s>
<s>
OMP2MPI	O
Automatically	O
generates	O
MPI	B-Application
source	O
code	O
from	O
OpenMP	B-Application
.	O
</s>
<s>
OMP2HMPP	O
,	O
a	O
tool	O
that	O
,	O
automatically	O
translates	O
a	O
high-level	O
C	B-Language
source	O
code(OpenMP )	O
code	O
into	O
HMPP	O
.	O
</s>
<s>
The	O
generated	O
version	O
rarely	O
will	O
differs	O
from	O
a	O
hand-coded	O
HMPP	O
version	O
,	O
and	O
will	O
provide	O
an	O
important	O
speedup	B-Operating_System
,	O
near	O
113%	O
,	O
that	O
could	O
be	O
later	O
improved	O
by	O
hand-coded	O
CUDA	B-Architecture
.	O
</s>
<s>
is	O
a	O
source-to-source	O
parallelization	B-Operating_System
tool	O
combined	O
with	O
an	O
interactive	O
GUI	B-Application
developed	O
by	O
emmtrix	O
Technologies	O
GmbH	O
.	O
</s>
<s>
It	O
takes	O
C	B-Language
,	O
MATLAB	B-Language
,	O
Simulink	B-Application
,	O
Scilab	B-Application
or	O
Xcos	B-Application
source	O
code	O
as	O
input	O
and	O
generates	O
parallel	O
C	B-Language
code	O
as	O
output	O
.	O
</s>
<s>
It	O
relies	O
on	O
static	O
schedule	O
and	O
a	O
message	O
passing	O
API	O
for	O
the	O
parallel	B-Operating_System
program	I-Operating_System
.	O
</s>
<s>
The	O
whole	O
parallelization	B-Operating_System
process	O
is	O
controlled	O
and	O
visualized	O
in	O
an	O
interactive	O
GUI	B-Application
enabling	O
parallelization	B-Operating_System
decisions	O
by	O
the	O
end	O
user	O
.	O
</s>
<s>
It	O
targets	O
embedded	O
multicore	O
architectures	O
combined	O
with	O
GPU	B-Architecture
and	O
FPGA	O
accelerators	O
.	O
</s>
<s>
The	O
translates	O
Fortran	B-Application
programs	O
with	O
claw	O
pragmas	O
into	O
Fortran	B-Application
code	O
suitable	O
for	O
a	O
specific	O
supercomputer	B-Architecture
target	O
augmented	O
with	O
OpenMP	B-Application
or	O
OpenACC	O
pragmas	O
.	O
</s>
<s>
PaSH	O
is	O
parallelizing	O
compiler	O
for	O
Unix	B-Application
shell	I-Application
scripts	O
.	O
</s>
