Robert,
I read what you wrote. I think I can say, "Yes" to all that you wrote. But,
perhaps I was unclear in what I wrote let me expand the model a little.
Consider the PDP-11, first had the basic CPU, next stage add memory
management, then add Floating Point. Now take a look at Intel or Motorola
CISC chips roughly the same model as PDP-11 as die size increases and line
width decreases. Now follow another path PhotoShop is introduced - all
software. PhotoShop hardware acceleration is introduced - some functions
done in silicon. 'Nother example, Intel recently added "instructions" to
improve graphics speed. When they create special "instructions" they have
basically added special hardware (done the job in silicon) once the "need"
for this set of functions was established.
What is my point? Simply if my model above is viable, and you believe you
want to control the entire process (and you are an Apple, Sun, IBM,...)
then you must "own" the development of both silicon and code and this leads
to the problems you so aptly describe.
Now you went and touched on the RISC vs. CISC controversy. First let me say
that the simple answer is if you can make the RISC clock rate high enough
you are going to improve throughput and more easily than you can with CISC.
I think that part of the problem is that digital engineers are the wrong
guys to do the design. You want analog engineers to do it. Analog guys move
Sine waves all over the place really fast and really nicely. You see those
rascally digital guys want to push square waves here and there but if you
want to reduce bandwidth, and power, and a lot of other stuff you really
want to move Sine waves. So for the industry to move in to RISC (and
simplify more than just the instruction set) and increase the clock rate
then there needs to be some more blending of these two engineering
disciplines. Probably the easiest example is EtherNet. Why do those 10 (or
100 Mhz) signals have to be digital and not analog signals? It's
transitions per second not the path (square or sine) you take on either
side of the transition that's important. IMHO.
END
**************************************************
Douglas W. St.Clair
(in New Hampshire) (in Massachusetts)
Tir Na Nog 30 George St.
400 Burton Highway Newton, MA 02158
Wilton. NH 03086-5022
PH: 603-654-9321 PH: 617-233-3387 or
FAX: 603-654-5440 PH: 617-BEEF-FUR
EMAIL: [EMAIL PROTECTED]
WebSite: http://www.tellink.net/~dwstclair
**************************************************
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
