[ < ][]   [Contents][Index][ ? ]

3.4 Double precision

The default for real variables is the real Fortran type. It is possible to use double precision instead. In that case all the occurences of ‘real ’ in mortran code is substituted with ‘double precision ’ at precompilation stage, and the Lapack subroutine names are replaced by the double precision names. Eventual users’declaration of complex  Fortran variables is also changed to double complex .

This feature is turned on by sel double in ‘selseq.kumac’ with cmz and double = 1 in the ‘Makefile’ with make.

In order for the model to run as well in double as in simple precision, some care should be taken to use the generic intrinsic functions, like sin and not dsin. No numerical constant should be passed directly to subroutines or functions, but instead a variable with the right type should be used to hold the constant value, taking advantage of the implicit casts to the variable type.


[Contents][Index][ ? ]