[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
Motion #17 -- make nextafter() homogeneous
David Hough's comment made me examine current the semi-homogeneous
definition more closely, and something definitely needs to be done
about it -- as written, it is inconsistent unless copySign() is also
made semi-homogeneous (because for x==y nextAfter(x,y) is defined in
terms of copySign(x,y)).
More interestingly, the current definition of semi-homogeneous does not
exclude mixed-radix forms -- not in 7.4.1 ("all supported non-storage
formats") and not in 7.1 where "semi-homogeneous" is defined.
This means that the current draft includes a primitive for mixed-radix
comparisons: (bfp > dfp) is equivalent to (bfp > nextAfter(bfp, dfp))
where the second comparison is same-radix! (Well, almost equivalent,
as nextAfter() could signal Overflow or Underflow as well as Inexact.)
Michel.
Sent: 2006-08-09 18:27:55 UTC