Re: float decimal


In article <c-udnVbk2ct0iO7YnZ2dnUVZ_qGdnZ2d@xxxxxxxxxxx>,
glen herrmannsfeldt <gah@xxxxxxxxxxxxxxxx> writes:
|>
|> In terms of range and precision, I would say it is reasonably equal to
|> IEEE binary. The cost in bits for a base 2 exponent balances the loss
|> for small most significant digit in a decimal representation. In
|> addition, that loss is already considered in most calculations
|> (most people still write answers in decimal and compute significant
|> digits from the decimal representation.)

The latter is slightly incorrect; error buildup is slightly worse for
a larger base, and the potential for biassing is higher. But your
summary of "reasonably equal" is unaffected :-)


Regards,
Nick Maclaren.
.