Numerics question: reading floating point constants

Milan Zimmermann milan.zimmermann at sympatico.ca
Tue Apr 25 05:19:15 UTC 2006


On 2006 April 9 12:23, stéphane ducasse wrote:
> Do you know a good book or lectures on the problems related to number
> representation in our nice
> computers?

Perhaps the best resource is the web site of William Kahan, "father" of IEEE 
754 here:

http://www.cs.berkeley.edu/~wkahan/

It is a fascinating read about the IEEE-754 battle. On the bottom of his site, 
he has recent (extremely interesting) articles, this is just one example:

http://www.cs.berkeley.edu/~wkahan/ARITH_17.pdf

In all those recent articles, he is describing the efforts to maintain 
hardware manufacturer's interest to support IEEE 754 while software support 
is lagging badly, even 20 years later, apparently only C99 and Fortran 
(2000something) are compliant, Cell chips have already sacrificed their 
hardware support, and other interesting stuff. It really feels sad for this 
industry ... but anyway, that seems the best place to look. 

Also this site is interesting:

http://babbage.cs.qc.edu/

In terms of books, perhaps this:

http://www.cs.nyu.edu/overton/book/

(but I did not read it)

I was looking for software that would test how much a language (libraries) 
adhere to IEEE754, thinking it would be interesting to translate to Squeak, 
there is stuff like:

http://www.tybor.com/

But I run out of understanding (and time)

Sorry to flood with links, I was just recently reading about this stuff and 
saved it all on Flock :)

Milan


>
> Stef
>
> On 9 avr. 06, at 18:18, nicolas cellier wrote:
> > 1 point for Andres, conversion from base 10 to base 2 is inexact.
> >
> > 1 point for Boris : successive inexact operations are not only
> > costly, they
> > can degrade accuracy.
> >
> > I wonder if the conversion shouldn't be made with respect to IEEE
> > rounding
> > mode, and if i remember, this rounding mode is a parameter of
> > underlying
> > hardware.
> >
> > In the past, i tried several algo. and finally, i lazily decided to
> > use a call
> > to sprintf/sscan from within VW.
> > But with a good reference such as Tony's (1 point) we should make
> > it in
> > Smalltalk.
> >
> > Nicolas



More information about the Squeak-dev mailing list