Portability Function: Gets, sets or clears IEEE flags for rounding direction and precision as well as queries or controls exception status. This function provides easy access to the modes and status required to use the features of IEEE Standard 754-1985 arithmetic in a Fortran program.
Module: USE IFPORT
Syntax
result = IEEE_FLAGS (action, mode, in, out)
action
(Input) Character*(*). One of the following literal values: 'GET', 'SET', 'CLEAR', or 'CLEARALL'.
mode
(Input) Character*(*). One of the following literal values: 'direction', 'precision', or 'exception'. The value 'precision' is only allowed on Intel® 64 and IA-32 architectures.
in
(Input) Character*(*). One of the following literal values: 'inexact', 'division', 'underflow','overflow', 'invalid', 'all', 'common', 'nearest', 'tozero', 'negative', 'positive','extended', 'double', 'single', or ' ', which represents an unused (null) value.
out
(Output) Must be at least CHARACTER*9. One of the literal values listed for in.
The descriptions for the values allowed for in and out can be summarized as follows:
| Value | Description |
|---|---|
| 'nearest' 'tozero' 'negative' 'positive' |
Rounding direction flags |
| 'single' 'double' 'extended' |
Rounding precision flags |
| 'inexact' 'underflow' 'overflow' 'division' 'invalid' |
Math exception flags |
| 'all' | All math exception flags above |
| 'common' | The math exception flags: 'invalid', 'division', 'overflow', and 'underflow' |
The values for in and out depend on the action and mode they are used with. The interaction of the parameters can be summarized as follows:
| Value of action | Value of mode | Value of in | Value of out | Functionality and return value |
|---|---|---|---|---|
| GET | 'direction' | Null (' ') | One of 'nearest', 'tozero', 'negative', or 'positive' | Tests rounding direction settings. Returns the current setting, or 'not available'. |
| 'exception' | Null (' ') | One of 'inexact', 'division', 'underflow' , 'overflow', 'invalid', 'all', or 'common' | Tests math exception settings. Returns the current setting, or 0. |
|
| 'precision' | Null (' ') | One of 'single ', 'double ', or 'extended' | Tests rounding precision settings. Returns the current setting, or 'not available'. |
|
| SET | 'direction' | One of 'nearest', 'tozero', 'negative', or 'positive' | Null (' ') | Sets a rounding direction. |
| 'exception' | One of 'inexact', 'division', 'underflow' , 'overflow', 'invalid', 'all', or 'common' | Null (' ') | Sets a floating-point math exception. | |
| 'precision' | One of 'single ', 'double ', or 'extended' | Null (' ') | Sets a rounding precision. | |
| CLEAR | 'direction' | Null (' ') | Null (' ') | Clears the mode. Sets rounding to 'nearest'. Returns 0 if successful. |
| 'exception' | One of 'inexact', 'division', 'underflow','overflow', 'invalid', 'all', or 'common' | Null (' ') | Clears the mode. Returns 0 if successful. |
|
| 'precision' | Null (' ') | Null (' ') | Clears the mode. Sets precision to 'double' (W*32, W*64) or 'extended' (L*X, M*X). Returns 0 if successful. |
|
| CLEARALL | Null (' ') | Null (' ') | Null (' ') | Clears all flags. Sets rounding to 'nearest', sets precision to 'double' (W*32, W*64) or 'extended' (L*X, M*X), and sets all exception flags to 0. Returns 0 if successful. |
Results
IEEE_FLAGS is an elemental, integer-valued function that sets IEEE flags for GET, SET, CLEAR, or CLEARALL procedures. It lets you control rounding direction and rounding precision, query exception status, and control exception enabling or disabling by using the SET or CLEAR procedures, respectively.
The flags information is returned as a set of 1-bit flags.
Examples
The following example gets the highest priority exception that has a flag raised. It passes the input argument in as a null string:
USE IFPORT
INTEGER*4 iflag
CHARACTER*9 out
iflag = ieee_flags('get', 'exception', '', out)
PRINT *, out, ' flag raised'
The following example sets the rounding direction to round toward zero, unless the hardware does not support directed rounding modes:
USE IFPORT
INTEGER*4 iflag
CHARACTER*10 mode, out, in
iflag = ieee_flags('set', 'direction', 'tozero', out)
The following example sets the rounding direction to the default ('nearest'):
USE IFPORT
INTEGER*4 iflag
CHARACTER*10 out, in
iflag = ieee_flags('clear','direction', '', '' )
The following example clears all exceptions:
USE IFPORT
INTEGER*4 iflag
CHARACTER*10 out
iflag = ieee_flags('clear','exception', 'all', '' )
The following example restores default direction and precision settings, and sets all exception flags to 0:
USE IFPORT
INTEGER*4 iflag
CHARACTER*10 mode, out, in
iflag = ieee_flags('clearall', '', '', '')
The following example detects an underflow exception:
USE IFPORT
CHARACTER*20 out, in
excep_detect = ieee_flags('get', 'exception', 'underflow', out)
if (out .eq.'underflow') stop 'underflow'