User and Reference Guide for the Intel® C++ Compiler 14.0
Rounds floating-point results after floating-point operations.
Linux OS and OS X: | -fp-port -no-fp-port |
Windows OS: | /Qfp-port /Qfp-port- |
None
-no-fp-port |
The default rounding behavior depends on the compiler's code generation decisions and the precision parameters of the operating system. |
This option rounds floating-point results after floating-point operations.
This option is designed to be used with the -mia32 (Linux* OS and OS X*) or /arch:IA32 (Windows* OS ) option on a 32-bit compiler. Under those conditions, the compiler implements floating-point calculations using the x87 instruction set, which uses an internal precision that may be higher than the precision specified in the program.
By default, the compiler may keep results of floating-point operations in this higher internal precision. Rounding to program precision occurs at unspecified points. This provides better performance, but the floating-point results are less deterministic. The [Q]fp-port option rounds floating-point results to user-specified precision at assignments and type conversions. This has some impact on speed.
When compiling for newer architectures, the compiler implements floating-point calculations with different instructions, such as Intel® SSE and SSE2. These Intel® Streaming SIMD Extensions round directly to single precision or double precision at every instruction. In these cases, option [Q]fp-port has no effect.
Windows: Optimization > Floating-point Precision Improvements
Linux: Floating Point > Round Floating-Point Results
OS X: Floating Point > Round Floating-Point Results
None