bear <xxxxxx@sonic.net> writes:

> 1) Represent powers of ten exactly, indicating a numeric format
>    with an exponent-of-five as well as an exponent-of-two.

Why not a single exponent-of-ten instead?

> So, unexectedly, if I can be clever enough with my basic operations
> to avoid cache stalls and conditionals, this isn't even a big
> performance hit.

I believe this would be a serious performance hit. The format is
complex, and it's not supported by hardware like binary FP. And with
a uniform representation you can't use a fast path for integer only
operations.

--
   __("<         Marcin Kowalczyk
   \__/       xxxxxx@knm.org.pl
    ^^     http://qrnik.knm.org.pl/~qrczak/