acos, sin, tanh, exp, log, sqrt - Floating Point Functions
The following table shows the floating point functions that expect a
number as an unnamed argument. Floating point functions are overloaded so that the return code can have the type
Decimal floating point numbers
are still only possible as arguments of
... func( arg ) ...
The argument of a floating point function must be a single data object outside an arithmetic expression and can be an arithmetic expression itself within an arithmetic expression.
Effect of the floating point functions:
|Function func||Meaning||Definition Range|
||hyperbolic cosine||-, no
||hyperbolic sine||-, no
||hyperbolic tangent||-, no
||Exponential function for base e||[-709, 709] for type
||Natural logarithm||> 0|
||Logarithm to base 10||> 0|
||Square root||>= 0|
Functions that specify "no
decfloat34" cannot currently have the
decfloat34. If one of these functions is specified in an expression with this calculation type, a syntax error occurs or the exception CX_SY_UNSUPPORTED_FUNCTION is raised.
The following applies to the floating point arithmetic in which a floating point function is calculated, and to the data type of the return code:
- If the argument has the type
decfloat34, a floating point function is calculated in decimal floating point arithmetic and the return code has the type
- If a floating point function is used in an arithmetic expression whose
calculation type is
decfloat34, it also calculates a return code with the type
decfloat34and the argument is first converted to the data type
decfloat34, if necessary.
- In all other cases, floating point functions use binary floating point arithmetic to calculate a
return code with type
fand the argument is first converted to the data type
f, if necessary.
Functions with a definition range require the value of
arg to be within the
specified limits. Arguments within the definition ranges are guaranteed to be error-free for the exponential function
exp, since the results are then within the
value ranges for
numbers in accordance with IEEE-754. For arguments less than -709, the result for binary floating
point numbers is (depending on the platform) a subnormal number, 0, or a handleable exception of the class CX_SY_ARITHMETIC_OVERFLOW is raised from a specific value.
The trigonometric functions
tan are defined for any arguments but the results become imprecise if the argument is greater than approximately 100,000,000.
atan function is undefined for odd-number multiples of pi/2, but the
definition range of
atan is nevertheless restricted since an argument of this function can never contain the precise value of pi/2.