|
Authored by: DieterWasDriving on Friday, March 29 2013 @ 11:04 PM EDT |
The inventor certain would understand the Apple 6502 implementation.
And understand at least some of the flaws in the as-simple-as-possible
implementation. The kind of flaws that motivated the IEEE standards not too
much later.
The pre-biasing (not pre-rounding) described in the patent is a workable
approach if you are working at the logic gate level. Without reading it too
deeply, it's likely simpler to implement when using pre-designed logic modules,
results in fewer gate delay than a IEEE-compliant implementation, and gets
compound operations wrong more often. It very likely changes previously exact
results into inexact results e.g. 1+1 is now 2.00000999 (but with numbers where
the flaw isn't quite as obvious. BTW, the correct answer here is '2'.)[ Reply to This | Parent | # ]
|
|
|
|
|