On Thu, Dec 29, 2016 at 9:54 PM, Alex Shinn <xxxxxx@gmail.com> wrote:

`decode-float', which breaks an IEEE-float into its constituent
sign, exponent and mantissa, is very useful and worth specifying.
See the portable call-with-mantissa&exponent implementation
from SRFI 56.

The existing accessors plus flcopysign achieve this, except on a (very
improbable) system in which not all flonums have the same radix.
 
What's the rationale for the constants instead of the expressions?
If we keep them it's worth noting that e is (flexp 1.0).

I'm going to get rid of the Scheme expressions in favor of mathematical
definitions.
 
  Also I think
tau is more useful than pi (http://tauday.com/) and worth defining.

Meh.  http://www.thepimanifesto.com/
 
Also consider defining the machine epsilon (DBL_EPSILON).

I never thought of looking in float.h.  I'll add its contents to the next draft. 

Could you provide an example for:

  Flonum operations must be at least as accurate as their generic
  counterparts applied to flonum arguments. In some cases,
  operations should be more accurate than their naive generic
  expansions because they have a smaller total roundoff error.

flexp-1 and fllog1+ are the obvious examples.

-- 
John Cowan          http://vrici.lojban.org/~cowan        xxxxxx@ccil.org
Be yourself.  Especially do not feign a working knowledge of RDF where
no such knowledge exists.  Neither be cynical about RELAX NG; for in
the face of all aridity and disenchantment in the world of markup,
James Clark is as perennial as the grass.  --DeXiderata, Sean McGrath