RE: Strange result with exponents...

[Scott Klement <klemscot@xxxxxxxxxxxx>]

> In fact, the only saving grace for this is that there is a note in the
> documentation that says floating point is used.  It would be nice to see
> a warning that the use of this opcode effectively limits your precision;
> it's way beyond the pale to expect an RPG programmer to know the
> precision limits of FP arithmetic.  Despite the fact that it was
> immediately clear to you, it was clearly not apparent to Scott, and
> Scott is one of the brighter people I know.

Wow.. thanks for the compliment...

However, I'm quite familiar with the limitations of floating point, I just
didn't know (until Rob or someone posted it) that the exponent operator in
ILE RPG outputs floating point.

And then, by the time I had learned that, I was confused because it used
to work in V5R1, and that did not make sense.  I'm pretty confident that
IEEE hasn't changed the format of 64-bit floats, so if the limitation is
in the number of digits of precision, why would the same exact number
change from release to release -- and I'm still not clear on exactly what
changed or why, but it doesn't sound like even Hans knows.

What we do know is that the user shouldn't use the exponentiation function
for large numbers, even if it works today, it might not work tomorrow
because of the fact that it uses floating point.

It hasn't made much difference to me -- I've never needed a number that
large, and even if I ever did, it would take me 2 minutes to write a
subprocedure that used the multiply operator in a loop.   I only posted
this question here to help out someone else.


> If he (and several other very bright people on this list) didn't get it,
> it might need a little explaining.

Yes, but understanding is the start point for explaining.   First you need
to understand what's going on, and likewise, IBM needs to understand where
we get confused.

Then comes documentation to save others from the same fate.