[daip] AIPS / POPS numerical precision
Eric Greisen
egreisen at nrao.edu
Mon Nov 1 11:39:38 EST 2004
Walter Brisken writes:
>
> After spending some time trying to understand why two of my coordinates
> didn't match after moving one of them (CLCOR), I came to the conclusion
> that the aips command interpreter is likely using low precision (32 bit
> floats) in its calculations. I discovered it when calculating:
>
> >print 51.052246-51.052300
> AIPS 1: -0.00005341
>
> whereas my handy python interpreter gives the more correct value
>
> >>> 51.052246-51.052300
> -5.4000000005771653e-05
>
>
> Would it be too much trouble to promote the internals of POPS to
> double-precision? Until then I will have to do these simple calculations
> elsehow.
>
Yes - it would be too much trouble - every task would have to change
and the whole structure of the POPS interpreter.
Eric Greisen
More information about the Daip
mailing list