Re: reproducible format

[Nick Maclaren <nmm1@xxxxxxxxxxxxx>]

Ivan Godard <igodard@xxxxxxxxxxx> wrote:
> I'd call such hardware buggy, but again this is a general system 
> architecture problem that occurs even in integer, and so it is not 
> floating point's problem to solve.

Why do you call hardware that is working to specification buggy?
You may not like the specification, but that isn't a reason.

I agree it is not floating-point's problem to solve, but equally
floating-point should not require solutions that are not provided
in general system architectures.  Whether you like it or not, the
precise semantics of floating-point are of no interest to 99% of
computer users, and of small interest to another 0.99%.  Demand
too much and you may get nothing.

> Let the market deal with it - don't buy defective hardware, even if it 
> can produce wrong results quickly. And, to get portable results, write a 
> portable software emulator to define the floating point semantics.

And the market has said loudly and clearly that it is happy to
accept such behaviour, if it enables more performance, as it does.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1@xxxxxxxxx
Tel.:  +44 1223 334761    Fax:  +44 1223 334679