Fortran output



>>>>> "Judah" == J Milgram <milgramjh@nswccd.navy.mil> writes:

    >> If you really want single precision, can't you always do this..
    >> 
    >> single precision   s,t,u,v
    >> s= 3.1234567890123456d0
    >> s is single precision even though the constant is double.

    Judah> (actually, I really  want  double :)...
    Judah> but if "d0" were inflicted on someone who wanted single, yes, by the
    Judah> time "s" were assigned, the value would be truncated to single. The
    Judah> only issue would be whether the processor somehow wasted time doing
    Judah> some stuff in double precision along the way but I can't think of an
    Judah> example. So maybe there's no compelling reason not to make "d0" the
    Judah> default output format.

What if the translated code was

        f(5d0)

when f really wanted a single-float?  Your translated Fortran would be
wrong.

I don't really know what the solution would be, but somehow, I think
that there will always be some hand editing.

Ray