Opened 5 years ago
Last modified 5 years ago
#291 new defect
LimitedPrecisionReal fmin subclassResponsibility not implemented
Reported by: | Owned by: | ||
---|---|---|---|
Priority: | major | Milestone: | |
Component: | default | Keywords: | Float fmin |
Cc: | Also affects CVS HEAD (eXept version): | no |
Description (last modified by )
I expect Float fmin to work.
Should it answer fminNormalized or fminDenormalized (gradual underflow)?
fminDenormalized is preferred for VW and Squeak compatibility, but LongFloat might not have denormalized numbers...
fminNormalized should be (self unity timesTwoPower: self emin).
fminDenormalized should be (self unity timesTwoPower: self emin - self precision + 1).
I would test for denormalized support and answer fminDenormalized if true, this way no need for subclassResponsibility (or more exactly, subclassResponsibility would be defered to denormalized support testing)