Re: Strange result with exponents...

["Buck" <buck.calabro@xxxxxxxxxxxx>]

> | Anyone who uses floating point should know that it is an approximation
> | and is only accurate to nearly 16 digits. I understand 16 digits of
> | precision exceeds the experimental accuracy of all known constants of
> | nature but that doesn't help in the business world.
>
> Am I understanding correctly that the 400 doesn't even Need floating point
> to solve all known Scientific math problems??  (Of course, Not suggesting
> it's unnecessary, because of "requirements" of backwards compatibility.)

Floating point exists so that scientific-types can represent very large (the
number of atoms in the universe) and very small (the distance between
neutrons in an atomic nucleus) numbers in a convenient format.  IEEE is one
of the most popular 'convenient format' for computers to use, and is pretty
much available on all modern platforms.  For iSeries to NOT have floating
point would surely be seen as a weakness.  Anyway, because the intent of
FLOAT is to be able to represent  very large and very small numbers in the
same format, there is a trade-off between the ultimate number of digits of
precision and the ability to handle the extremes well.

The reference to "...exceeds the experimental accuracy..." means basically
that there is no known sensor which can measure that precisely.  No
thermometer can measure temperature differences of .0000000000000001 degree;
no micrometer can measure that small a distance, etc.  16 digits provide a
ridiculously large degree of precision, so we can't really critisise the
IEEE on that score.  There may be grounds for critisising software that
insists on _displaying_ those un-precise digits though...

If you'd like to play with floating point fields, here's some quickie SQL
that will let you see how things look:

CREATE TABLE BUCK/FLOAT (CHARS CHAR (10) NOT NULL WITH DEFAULT,
FLOAT4 REAL NOT NULL WITH DEFAULT, FLOAT8 DOUBLE NOT NULL WITH
DEFAULT)

insert into float values ('low',1.23456789E-38,1.23456789012345e-307)

insert into float values ('hi',1.23456789E38,1.23456789012345e308)
select * from float
....+....1....+....2....+....3....+....4....+....5..
CHARS                FLOAT4                   FLOAT8
low          1.2345679E-038   1.234567890123450E-307
hi           1.2345679E+038   1.234567890123450E+308

It has been an interesting discussion.  I say we take up logarithms like the
deciBel next.  :-)  (Just kidding)
  --buck