On 12/30/08, Dieter Kaiser <drdieterkaiser at web.de> wrote:
> I will have a look at the failures concerning the accuracy of some special
> functions and adjust them accordingly.
Well, I 've noticed that there are failures of special function tests
for other Lisp implementations as well (all for numerical evaluation
if I recall correctly). I don't know what we can do about this in
general. If all implementations use the same floating point
precision, shouldn't they obtain the same result? Or do the results
depend on stuff like rounding mode? or maybe library functions like
SIN, EXP, etc?
I wonder how we can invent a test which is less dependent on
implementation details, or else arrange all implementations to have
the same details.
best
Robert Dodier