|
Floating-point Precision
In most cases, as floating-point precision increases, so does processor
execution time. The following table shows the floating-point precision
optimization options along with a description of each option's precision
vs. speed.
Windows* |
Linux* |
Description
|
-Op |
-mp |
Strict ANSI C and IEEE 754 Floating
Point (subset of -Za or -ansi)
very precise; slows your application down |
-Za |
-ansi |
Strict ANSI C and IEEE 754. Slower and
more precise than -Op. |
-Qlong_double |
-long_double |
Changes the defaultsize of the long
double type from 64 to 80 bits. This option introduces
a number of incompatibilities with other files compiled without
this option and with calls to library routines. Therefore,
Intel recommends that the use of long double
variables be local to a single file when you compile with
this option. |
-Qprec |
-mp1 |
IA-32 only. Almost, but not quite, as
precise as ANSI; faster than ANSI. All but the most costly
operations are done according to ANSI. |
-Qprec_div |
-prec_div |
IA-32 only. By default, the Intel compiler
converts division operations into reciprocal multiply operations,
which is less precise but much faster. This option disables
the floating point division-to-multiplication optimization. |
-Qpcn |
-pcn |
IA-32 only. Specifies required number
of bits of precision: 32, 64, or 80. Rounds to n
precision, where n=[32|64|80].
|
-Qrcd |
-rcd |
IA-32 only. Removes code that truncates
during float to integer conversions.
Improves the performance of code that requires floating-point-to-integer
conversions. |
|
|
|