> -----Original Message-----
> From: Brian Hurt [mailto:bhurt at spnz.org]
> Sent: Saturday, December 08, 2007 11:42 AM
> To: Chuck Cole
> Cc: tclug-list at mn-linux.org
> Subject: RE: [tclug-list] the 2038 bug already bit me!
>
>
>
>
> On Thu, 6 Dec 2007, Chuck Cole wrote:
>
> >
> >
> >> -----Original Message-----
> >> From: tclug-list-bounces at mn-linux.org
> >> [mailto:tclug-list-bounces at mn-linux.org]On Behalf Of Brian Hurt
> >> Sent: Thursday, December 06, 2007 7:16 PM
> >>
> >>
> >> In my opinion, we should have switched to 64-bit circa 1997.
> >
> > That's both profoundly naive of how important having different CPU type
> > really is for differing applications
>
> I thought I had explicitly made clear that this was for desktops.
>  Servers
> had already gone 64-bit by 1997.  And it was circa 1997 that I read an
> article in EEtimes that indicated a new era has dawned- for the
> first time
> in history, Moto had sold more 68000's than 6502's.

Your world is commercial IT by that description.  The universe of Linux is
bigger.

When designing a new embedded flight system, basic models of dataflows, etc
are benchmarked in order to determine whether the memory usage of desktop
crud doubles the power and weight or not.  The criteria and engineering
efforts applied are considerably more involved than what you seem aware of.




>
> >
> > The "excess bits" in data flows and control flows are points of
> potential
> > failure, and consume power, and waste time, and cost more.
> >
> > In spacecraft, extra pounds cost millions, extra watts do also.
>  Cheaper in
> > fighter aircraft, etc, but still a great concern and expense.
>
> Unfortunately for this argument, the number of data flows, control flows,
> etc., a CPU has is not strongly correlated with the number of
> bits in it's
> word size- as proof positive of this, compare the complexity and
> size of a
> earlt Dec Alpha 21061 (a 64-bit chip, 1.68 million transistors, 21W power
> dissaption) with a later Intel Pentium-4 (a 32-bit chip, 178 million
> transitors 110W power dispation).  The pentium has 10 times the
> transistor
> count, five times the power dissapation, and half the bits.
>
> Also, remember what my real argument was- "If you can't mmap your whole
> hard disk, you don't have enough bits in your address space".  If you're
> working in an embedded project that has a few 10's of meg of ram, and few
> 100's of meg of rom ("hard disk"), then 32 bits is fine (16 bits is
> small).
>
> Brian
>


Your comments are irrelevant or wrong outside the small "snow dome" of
architecture you are addressing.


Chuck