FORTRAN floating point accuracy problems

[Dieter]

(04-04-2016 09:02 AM)Werner Wrote:  Can I chime in?

Sure. 8-)

(04-04-2016 09:02 AM)Werner Wrote:  The following formulas, courtesy of John H. Meyers on comp.sys.hp48 do not require any fiddling at all:
...
T := HMS + FP(HMS*100)/150;
HR := T + FP(T)/1.5;

Ah, great. Yes, now I remember these formulas, too. Although I cannot definitely say (yet) if this solves all problems due to roundoff by non-decimal number representations, I think this should work very well.

So a Fortran version of this sounds like a good idea. I would give it a try. Maybe (or even "probably") this will solve the whole issue very elegantly.

Dieter