Re: Strange result with exponents...

On Fri, 20 Feb 2004, Buck wrote:
>
> Yes indeed.  It looks as if Simon's conjecture is on the money.  Some sort
> of conversion/rounding error between FLOAT8 and packed.  It looks like C
> will by default only display the 16 digits that double is supposed to hold
> while the RPG conversion from float to packed manages to extract that little
> something extra!  :-)

Yep, in fact if you call the CEESDXPD API instead of using pow() or **,
you get the exact same results.  Even though you call the same API from
both RPG and C, printf() will display less precision than the conversion
to packed and DSPLY.

That, to me, makes it quite clear that the difference isn't the language,
but rather the conversion to packed.

In fact, the following RPG program gives the same (16-digit) result even
though no C code is involved, and again I'm assuming it's because it's
never converted to packed...

     D z               s              8F
     D msg             s             50A
     c                   eval      z = 16**15
     c                   eval      msg = %editflt(z)
     c                   dsply                   msg
     c                   eval      *inlr = *on