Re: OT: Number bases

On Thu, Feb 12, 2009 at 6:57 AM, Henrik Theiling <theiling@...> wrote:
> Ah, now I understand what you mean.  You're right.  So you'd expect a
> ternary CPU'd still have both signed and unsigned arithmetics even if
> the CPU was build with balanced ternary logic in mind?
No, I wouldn't. While, as with two's complement, the basic truth
tables at the single-trit level is the same (given states 0, 1, and X,
the result of adding two trits is the same whether X is treated as -1
or 2),  the carry out is different, so hardware adders would have to
be designed explicitly one way or the other.  So the choice has to be
made sooner, at a more fundamental level.  I was just pointing out
that it was still a choice. :)

--
Mark J. Reed <markjreed@...>

Reply