[Openmcl-devel] Re: single-float support in OpenMCL (MCL actually)

[Dan Corkill]

> In MCL, a lisp SHORT-FLOAT denotes an IEEE single float, and the
> other CL float types all denote an IEEE double.

I've found MCL's use of only SHORT-FLOAT for IEEE single float to
be problematic for portable code. When IEEE formats are supported,
having the two-representation separation occur between SINGLE-FLOAT
and DOUBLE-FLOAT is the better choice (based on the minimum precision
and exponent size requirements).  IMHO, SHORT-FLOAT expresses the
desire to use as compact a representation (potentially immediate)
as possible and should only be split from SINGLE-FLOAT when there
is a representation that does not satisfy the minimums required by
SINGLE-FLOAT.  I'm happy that OpenMCL implements the more desirable
floating-representations split!