Re: Lua string library partially reimplemented in Lua

[Louis Mamakos <louie@...>]

On Sep 22, 2008, at 1:35 AM, Miles Bader wrote:

> Louis Mamakos <louie@transsys.com> writes:
> > Just for completeness, this is only true for CPU architectures that
> > implement 2's complement arithmetic.  While 1's complement CPU
> > architectures [1] are even more rare...
> >
> > [1] UNIVAC/Sperry/Unisys 1100/2200 architectures, DEC-10/20 among
> > them.
>
> The pdp-10 (decsystem 10/20) used 2s-complement integers.

Sorry, it was PDP-1 and IBM 7090/7094's that I was thinking of.

> It seems pretty reasonable to at least _optimize_ for the 2s- 
> complement case...

It's not optimization but rather embedding architecture dependencies
needlessly in a specification.  There ought to be a distinct set
of arithmetic vs. bitwise-logical functions specified.  I suppose
this is similar to not presuming a specific word size for data types
in the specification of the function.

> -Miles
>
> --
> Do not taunt Happy Fun Ball.

Always wise advice to heed.

louie