Floating point non-exactness

[Andrew Koenig]

In article <2646 at dataio.Data-IO.COM>, bright at Data-IO.COM (Walter Bright) writes:
> In article <6855 at ozdaltx.UUCP> doug at ozdaltx.UUCP (Doug Matlock) writes:

> <In article <11160 at alice.UUCP>, ark at alice.UUCP (Andrew Koenig) writes:

> << It also guarantees exact results
> << for primitive operations including addition, subtraction, multiplication,
> << division, and square root.

> <What if these primitive operators generates a result that cannot be
> <represented exactly in IEEE floating point notation?

> IEEE enables you to specify that an exception is generated if the result
> is not exact.

Yes.  Also, and this is the point I was trying to make before, for the
primitive operations I mentioned, IEEE requires that the result of
each operation be exactly what would be obtained by doing the operation
in infinite precision and then rounding it.  In other words, for
those operations, all IEEE implementations must give bit-for-bit
identical results.
-- 
				--Andrew Koenig
				  ark at europa.att.com