On Thu, May 20, 1999 at 12:26:13 +0200, kim@pvv.ntnu.no wrote:
> As a programmer of numerical math, I say that programs depending on
> the format of floating point numbers, are bad  And yes, there are a
> lot of bad programs out there. I needed it for my own numerical work,
> where speed is indeed much more important than compliance to some
> irrelevant standard. Besides, I had planned on just letting floats
> be fast, while doubles was to be IEEE.

I partly disagree. Concerning the formats, some programs may need to
use a fixed format (e.g. the IEEE-754 double precision). However, I
agree that most programs shouldn't depend on the format. But this
isn't the problem. What is more important, is some properties of
the operations and functions. For instance, programs may need exact
rounding for the +, -, *, / operations, as it may be used to ensure
coherency (e.g. to have properties like -1 <= x/sqrt(x^2+y^2) <= 1).
And at least, some properties such as the monotonicity are quite
important. If you use non-IEEE fast floats, things may go wrong.
Well, a compiler+libraries with such fast floats could be written,
but it should only be an option.

-- 
Vincent Lefèvre <vincent@vinc17.org> - PhD student in Computer Science
Web: <http://www.vinc17.org/> or <http://www.ens-lyon.fr/~vlefevre/> - 100%
validated HTML - Acorn Risc PC, Yellow Pig 17, Championnat International
des Jeux Mathématiques et Logiques, TETRHEX, Faits divers insolites, etc.