Re: Serious bug

[Timothy Ritchey <tritchey at vne dot com>]

Tim Hollebeek wrote:
 
> A sane floating point model has a well specified floating point
> representation, and well specified rules for how results are
> calculated.  

I am almost positive that the Intel floating point implementation is
well specified. You can find all the literature on it that you want.

> Sane floating point is *approximate* and *deterministic*.

As far as I understand, every time you run the program, you will get the
same results - that is deterministic in my book. no-one is making
guarantees about different platforms, or optimization techniques.
different processors and different instructions will not mean
differences.

> People who understand floating point, and have develop numerical
> programs on a regular basis refer to such architectures using highly
> technical terms.  The behavior is refered to as "stupid", and the
> architectures are considered to be "broken".

Is the Java implementation BROKEN and STUPID? I think you will find that
the scientific community thinks so. Java implements FP just as you
suggest, and it took all of two months before the scientists and
programmers started bitching to SUN to change it. That is why their spec
has been updated recently to allow just the abomination that you
describe. So, were all of these people asking sun to break java and make
it stupid?

Oh, and whoever said they are glad they didn't take CS, I am sure that
the combined knowledge from 30+ years of computing that has created the
entire discipline of Computer Science, is nothing compared to your
intuition on this matter. :)