Against PEP 240

[Darren New]

Greg Ewing wrote:
> But for some reason all the designers of really fast
> FP hardware seem to concentrate on binary.

The Sigma-9 mainframe (old) had a "scientific hardware unit" (floating
point) and a "business hardware unit" (BCD, packed BCD, COBOL-style
"edit" fields, etc). 

BCD is good for high precision small range. Float is good for lower
precision wider range. Science is generally limited by the number of
significant digits in your real measurements. You can't know how close
that star is down to the inch. Hence, using binary for FP gives you more
precision in the same number of bits.

-- 
Darren New / Senior MTS & Free Radical / Invisible Worlds Inc.
       San Diego, CA, USA (PST).  Cryptokeys on demand.
     This is top-quality raw fish, the Rolls-Rice of Sushi!