g++ optimization issue??

[Robert Dewar]

ColinLaney wrote:
> I am not asking for a compiler that magically fixes unwanted properties
> of the IEEE representation. I just wondered why it wasn't possible to be
> consistent.

Consistent with what? Your expectations? The standard? Some fundamental
laws?

For some consistency means that optimization never affects the results,
but language standards generally allow non-deterministic results when
it comes to fpt, so two results can be formally equivalent (they are
both arbitrary selections from a common set), even though they are
different.