[Shootout-list] What people say about shootout

[skaller]

On Sun, 2005-05-01 at 20:31, Bengt Kleberg wrote:
> Jon Harrop wrote:
> ...deleted
> > approximately the same but not exactly the same. So we'd have to start 
> > allowing some variation in "correct" answers.
> 
> this seems dangerous to me. i would be happier if it did not happen.

A set of correct answers makes perfect sense for floating
point routines. For a matrix problem one often defines

	| A - C | < epsilon

as the requirements, where A is the result, and C is the 
approximate answer. Of course there are other criteria
such as mini-max solution.

In general, requiring exact answers would make
it impossible to compare results across languages
and machines which may use different floating point
operations for the same formula.

The well known but defunct Cray, master of numerical
computing power, did not use IEEE compliant floats.

-- 
John Skaller, mailto:skaller@users.sf.net
voice: 061-2-9660-0850, 
snail: PO BOX 401 Glebe NSW 2037 Australia
Checkout the Felix programming language http://felix.sf.net