<s>
strictfp	B-Language
is	O
an	O
obsolete	O
and	O
unused	O
reserved	O
word	O
in	O
the	O
Java	B-Language
programming	I-Language
language	I-Language
.	O
</s>
<s>
Previously	O
,	O
this	O
keyword	O
was	O
used	O
as	O
a	O
modifier	O
that	O
restricted	O
floating-point	B-Algorithm
calculations	O
to	O
IEEE	O
754	O
semantics	O
in	O
order	O
to	O
ensure	O
portability	O
.	O
</s>
<s>
The	O
strictfp	B-Language
keyword	O
was	O
introduced	O
into	O
Java	B-Language
with	O
the	O
Java	B-Language
virtual	I-Language
machine	I-Language
(	O
JVM	B-Language
)	O
version	O
1.2	O
and	O
its	O
functionality	O
was	O
removed	O
in	O
JVM	B-Language
version	O
17	O
.	O
</s>
<s>
As	O
of	O
Java	B-Language
17	O
,	O
IEEE	O
754	O
semantics	O
is	O
required	O
,	O
thus	O
using	O
this	O
keyword	O
has	O
no	O
effect	O
.	O
</s>
<s>
The	O
IEEE	O
standard	O
IEEE	O
754	O
specifies	O
a	O
standard	O
method	O
for	O
both	O
floating-point	B-Algorithm
calculations	O
and	O
storage	O
of	O
floating-point	B-Algorithm
values	O
in	O
various	O
formats	O
,	O
</s>
<s>
including	O
single	O
(	O
32-bit	O
,	O
used	O
in	O
Java	B-Language
's	O
float	B-Algorithm
)	O
or	O
double	O
(	O
64-bit	O
,	O
used	O
in	O
Java	B-Language
's	O
double	O
)	O
precision	O
.	O
</s>
<s>
Some	O
hardware	O
also	O
provides	O
extended	B-Algorithm
precision	I-Algorithm
formats	I-Algorithm
that	O
provide	O
higher	O
precision	O
and/or	O
a	O
larger	O
exponent	O
range	O
.	O
</s>
<s>
This	O
may	O
avoid	O
round-off	B-Algorithm
errors	I-Algorithm
,	O
overflows	B-Algorithm
and	O
underflows	B-Algorithm
that	O
would	O
otherwise	O
occur	O
,	O
but	O
can	O
cause	O
programs	O
to	O
produce	O
different	O
output	O
on	O
such	O
architectures	O
.	O
</s>
<s>
It	O
was	O
particularly	O
expensive	O
to	O
avoid	O
the	O
use	O
of	O
extended	B-Algorithm
precision	I-Algorithm
on	O
x86	O
machines	O
with	O
the	O
traditional	O
x87	B-Application
floating-point	B-Algorithm
architecture	O
.	O
</s>
<s>
Prior	O
to	O
JVM	B-Language
1.2	O
,	O
floating-point	B-Algorithm
calculations	O
were	O
required	O
to	O
be	O
strict	O
;	O
that	O
is	O
,	O
all	O
intermediate	O
floating-point	B-Algorithm
results	O
were	O
required	O
to	O
behave	O
as	O
if	O
represented	O
using	O
IEEE	O
single	O
or	O
double	O
precisions	O
.	O
</s>
<s>
This	O
made	O
it	O
expensive	O
on	O
common	O
x87-based	O
hardware	O
to	O
ensure	O
that	O
overflows	B-Algorithm
would	O
occur	O
where	O
required	O
.	O
</s>
<s>
Since	O
JVM	B-Language
1.2	O
,	O
intermediate	O
computations	O
are	O
,	O
by	O
default	O
,	O
allowed	O
to	O
exceed	O
the	O
standard	O
exponent	O
ranges	O
associated	O
with	O
IEEE	O
32	O
bit	O
and	O
64	O
bit	O
formats	O
.	O
</s>
<s>
On	O
platforms	O
like	O
x87	B-Application
,	O
overflows	B-Algorithm
and	O
underflows	B-Algorithm
may	O
not	O
occur	O
where	O
expected	O
,	O
producing	O
possibly	O
more	O
meaningful	O
,	O
but	O
less	O
repeatable	O
,	O
results	O
instead	O
.	O
</s>
<s>
Since	O
x87	B-Application
floating	B-Algorithm
point	I-Algorithm
is	O
no	O
longer	O
necessary	O
on	O
x86	O
processors	O
supporting	O
SSE2	B-General_Concept
,	O
Java	B-Language
17	O
again	O
made	O
all	O
floating-point	B-Algorithm
operations	O
strict	O
,	O
effectively	O
restoring	O
the	O
pre-1.2	O
semantics	O
.	O
</s>
<s>
In	O
the	O
absence	O
of	O
overflow	O
or	O
underflow	B-Algorithm
,	O
there	O
is	O
no	O
difference	O
in	O
results	O
with	O
or	O
without	O
strictfp	B-Language
.	O
</s>
<s>
If	O
repeatability	O
is	O
essential	O
,	O
the	O
strictfp	B-Language
modifier	O
can	O
be	O
used	O
to	O
ensure	O
that	O
overflow	O
and	O
underflow	B-Algorithm
occurs	O
in	O
the	O
same	O
places	O
on	O
all	O
platforms	O
.	O
</s>
<s>
Without	O
the	O
strictfp	B-Language
modifier	O
,	O
intermediate	O
results	O
may	O
use	O
a	O
larger	O
exponent	O
range	O
.	O
</s>
<s>
The	O
strictfp	B-Language
modifier	O
accomplishes	O
this	O
by	O
representing	O
all	O
intermediate	O
values	O
as	O
IEEE	O
single	O
precision	O
and	O
double	O
precision	O
values	O
,	O
as	O
occurred	O
in	O
earlier	O
versions	O
of	O
the	O
JVM	B-Language
.	O
</s>
<s>
Programmers	O
can	O
use	O
the	O
modifier	O
strictfp	B-Language
to	O
ensure	O
that	O
calculations	O
are	O
performed	O
as	O
in	O
the	O
earlier	O
versions	O
;	O
that	O
is	O
,	O
only	O
with	O
IEEE	O
single	O
and	O
double	O
precision	O
types	O
used	O
.	O
</s>
<s>
Using	O
strictfp	B-Language
guarantees	O
that	O
results	O
of	O
floating-point	B-Algorithm
calculations	O
are	O
identical	O
on	O
all	O
platforms	O
.	O
</s>
<s>
It	O
can	O
be	O
used	O
on	O
classes	O
,	O
interfaces	B-Application
and	O
non-abstract	O
methods	B-Language
.	O
</s>
<s>
When	O
applied	O
to	O
a	O
method	O
,	O
it	O
causes	O
all	O
calculations	O
inside	O
the	O
method	O
to	O
use	O
strict	O
floating-point	B-Algorithm
math	I-Algorithm
.	O
</s>
<s>
When	O
applied	O
to	O
a	O
class	O
,	O
all	O
calculations	O
inside	O
the	O
class	O
use	O
strict	O
floating-point	B-Algorithm
math	I-Algorithm
.	O
</s>
<s>
Compile-time	O
constant	O
expressions	O
must	O
always	O
use	O
strict	O
floating-point	B-Algorithm
behavior	O
.	O
</s>
