high precision mathematics

[Richard Brodie]

"François Pinard" <pinard at iro.umontreal.ca> wrote in message
news:mailman.1013991930.24000.python-list at python.org...

> > > [...] a requirement for high precision floats is usually a sign that
> > > the requirer does not understand his problem space.
>
> > Incredibly untrue.
>
> Science people often need more precision.  DOUBLE existed in early FORTRAN.
> C has `float' and `double' as well.  It would be rather surprising if all
> involved people were that dumb! :-)

Well probably not all. However many more scientists use mathematical tools
than have a deep understanding of them. To take the example of fluid
dynamics: some problems are inherently chaotic (e.g. famously, weather
prediction). You will get different results if you ran the simulation at
different precisions. That doesn't mean that the higher precision results
are any more meaningful.

There is a set of problems where single precision isn't good enough
and double precision is. However, that surely is smaller than the set of
problems which are so ill-conditioned that double precision is insufficient.
I strongly suspect the OP was correct in implying most people can't tell
the difference: it's a corollary of Kaa's law.