I have met Fortran (77 and 90) compilers that compute ANINT(X) by using
REAL(NINT(X)), and AINT(X) by REAL(INT(X)).
Was it the intent of the standard for AINT(X) to be more accurate, for
|X| larger than the largest integer?
E.g. (for positive X) to compute AINT(X) = X - MOD(X,1.0)?
Or was it the intent of the standard to be silent here?
To me it appears quite clear that the former was desired.
Best regards,
Van Snyder
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|