- Fortran 90 has a way to allow
non-technical programmer
to select a data type to
store a given accuracy.
- Recall that:
- INTEGER(4)
defines the integer type
made up of
4 bytes
- INTEGER(8)
defines the integer type
made up of
8 bytes
- REAL(4)
defines the floating point type
made up of
4 bytes
- And so on..
- INTEGER(4)
defines the integer type
made up of
4 bytes
- Suppose your program operates on:
- Integer values that has 12 digits
- Floating point numbers with 13 digits matissa and of the order of 10^6
- Integer values that has 12 digits
- The precision requirement in F90 can be best illustrated with
integers.
- Recall that variables in computers can come in various (but
limited) sizes:
- 1 byte (characters or byte integers), range: -128..127
- 2 bytes (short integers), range: -32768..32767
- 4 bytes (ordinary integers)
- 8 bytes (long integers)
- Without knowledge of binary number representation, a non-technical
programmer would not know that a short (2 byte) integer
cannot represent a number such as 60000
- To this purpose, F90 provides the following construct
to allow non-technical programmers to define the
right kind of integers:
INTEGER( SELECTED_INT_KIND(#digits) ) :: VarName or INTEGER( KIND=SELECTED_INT_KIND(#digits) ) :: VarName
defines variables that can represents ALL integers that can be written with "#digit" digitsExamples: INTEGER( SELECTED_INT_KIND(2) ) A !! Large enough to store !! all integers of 2 digits INTEGER( SELECTED_INT_KIND(4) ) A2 !! Large enough to store !! all integers of 4 digits
-
Example Program:
(SELECTED_INT_KIND) ---
click here
- Fortran also provide a way to specify precision for
floating point numbers.
The meaning of floating point precision is different than that of integer numbers.
- Floating point number precision is related to
round off errors in addition to
the magnitude of the value represented
- How to specify the right kind of floating point numbers:
REAL ( SELECTED_REAL_KIND(prec, magn) ) :: VarName or REAL ( KIND=SELECTED_REAL_KIND(prec, magn) ) :: VarName
defines variables that can represents ALL real numbers that has at least "prec" number of digits precision and a magnitude in the range of 10-magn and 10magnExamples: REAL ( SELECTED_REAL_KIND(14, 50) ) A Variable A will have at least 14 digits precision and can represent values between 10-50 and 1050
-
Example Program:
(SELECTED_REAL_KIND) ---
click here