Run-Time Subroutine: Sets the value of the floating-point processor control word. This routine is only available on ia32 processors.
Module: USE DFLIB
Syntax
The floating-point control word specifies how various exception conditions are handled by the floating-point math coprocessor, sets the floating-point precision, and specifies the floating-point rounding mechanism used.
The DFLIB.F90 module file (in the \DF98\INCLUDE subdirectory) contains constants defined for the control word as follows:
| Parameter name | Hex value | Description |
|---|---|---|
| FPCW$MCW_IC | #1000 | Infinity control mask |
| FPCW$AFFINE | #1000 | Affine infinity |
| FPCW$PROJECTIVE | #0000 | Projective infinity |
| FPCW$MCW_PC | #0300 | Precision control mask |
| FPCW$64 | #0300 | 64-bit precision |
| FPCW$53 | #0200 | 53-bit precision |
| FPCW$24 | #0000 | 24-bit precision |
| FPCW$MCW_RC | #0C00 | Rounding control mask |
| FPCW$CHOP | #0C00 | Truncate |
| FPCW$UP | #0800 | Round up |
| FPCW$DOWN | #0400 | Round down |
| FPCW$NEAR | #0000 | Round to nearest |
| FPCW$MCW_EM | #003F | Exception mask |
| FPCW$INVALID | #0001 | Allow invalid numbers |
| FPCW$DENORMAL | #0002 | Allow denormals (very small numbers) |
| FPCW$ZERODIVIDE | #0004 | Allow divide by zero |
| FPCW$OVERFLOW | #0008 | Allow overflow |
| FPCW$UNDERFLOW | #0010 | Allow underflow |
| FPCW$INEXACT | #0020 | Allow inexact precision |
The defaults for the floating-point control word are 53-bit precision, round to nearest, and the denormal, underflow and inexact precision exceptions disabled. An exception is disabled if its flag is set to 1 and enabled if its flag is cleared to 0.
Setting the floating-point precision and rounding mechanism can be useful if you are reusing old code that is sensitive to the floating-point precision standard used and you want to get the same results as on the old machine.
You can use GETCONTROLFPQQ to retrieve the current control word and SETCONTROLFPQQ to change the control word. Most users do not need to change the default settings. If you need to change the control word, always use SETCONTROLFPQQ to make sure that special routines handling floating-point stack exceptions and abnormal propagation work correctly.
For a full discussion of the floating-point control word, exceptions, and error handling, see The Floating-Point Environment in the Programmer's Guide.
The Visual Fortran exception handler allows for software masking of invalid operations, but does not allow the math chip to mask them. If you choose to use the software masking, be aware that this can affect program performance if you compile code written for Visual Fortran with another compiler.
Compatibility
CONSOLE STANDARD GRAPHICS QUICKWIN GRAPHICS WINDOWS DLL LIB
See Also: GETCONTROLFPQQ, GETSTATUSFPQQ, LCWRQQ, SCWRQQ, CLEARSTATUSFPQQ
Example
USE DFLIB
INTEGER(2) status, control, controlo
CALL GETCONTROLFPQQ(control)
WRITE (*, 9000) 'Control word: ', control
! Save old control word
controlo = control
! Clear all flags
control = control .AND. #0000
! Set new control to round up
control = control .OR. FPCW$UP
CALL SETCONTROLFPQQ(control)
CALL GETCONTROLFPQQ(control)
WRITE (*, 9000) 'Control word: ', control
9000 FORMAT (1X, A, Z4)
END