Before I get flamed, I want to back away from this suggestion a bit.  I
went further back in the archives, and read Dave N. Smith's posts about
doubles, and the many posts about immediate small floats.

In the particular application for which I need FloatArrays, I only need a
few digits of precision, and the exponential range is not at all extreme.
I wonder how often this is the case.  I started thinking about this when I
checked all of the instance of Float in the system.  My guess is that most
of the values in the virgin system are a result of windowing ratios, etc.
None of these floats seem to require more than single precision.  Would it
make sense to have a SmallFloat as an immediate that would convert itself
to a normal double precision (or more) when necessary?

I know my application would also work with fixed points as well.  Perhaps
Andreas 3D graphics application could use them as well.

At 7:01 PM 9/2/98, Christopher Oliver wrote:
>To be blunt, why should one CPU receive preferential attention?
>Maybe this should be a compile time issue for the VM.

I should have said both single and double precision arrays.

--Maurice

BTW, I see that it was only Hans-Martin Mosner's experimental system that
uses more than a single tag bit.  Hans, is your website temporarily down?
I haven't been able to get on for the past few days?



---------------------------------------------------------------------------
  Maurice Rabb    773.281.6003    Stono Technologies, LLC    Chicago, USA