Re: Chucks address

[Jaap van Ganswijk <ganswijk@xxxxxxxxx>]

At 23:10 20000613 -0600, Mark Walker wrote:
>Jaap van Ganswijk wrote:
>> Computer architecture shouldn't depend on what
>> is doable in hardware at the moment. The folly of the
>
>You miss the point of MISC completely.  MISC is about
>maximum performance from the essense of computing devices
>and systems.

Maximum performance is reached by using
hard coded hardware designs. Being able to
control the part using a program is already a
severe speed-compromise. I'm just saying that
because software writing is expensive the
hardware should facilitate it at the cost of
some speed even. Especially since the hardware
techniques may evolve, but the architecture
must stay compatible in future.

Consider for example that writing a compiler
to optimally use a certain generation of the
CPU takes one year instead of one month
because of the complexity of the chip. It's
better to use a simpler chip design and the
simple easy to make compiler then to wait
a year. After a year the simple chip in a
new technological process will also be
say twice as fast, and the design of the
chip and the compiler will still be cheaper
and less prone to errors.

The design goals for MISC were if I recall correctly:
- small
- quick
- cheap

But implicit goals are always:
- easy to use
- available
- well documented

Just focusing on speed isn't wise. Anyway, with
a low budget you can not hope to have a production
run every say 6 months, to keep upto hi-tech speed.

When you manage to reach the last four goals
it may happen that people are actually going to
use the part and you can produce it more often.
That will make the actual mean speed of the parts
in the field much higher than optimizing the hardware
completely on speed disregarding other aspects.

>Also, *your* point is completely wrong.  Realizable
>computer architecture depends completely on what
>is doable in hardware at the moment.

At the moment of the birth of the MISC the 486 was
already around so there was already a lot doable.

>You compound
>your error when you then use this fact (which you've
>denied) to take issue with perceived deficiences in the
>internal implementation methods used on the *21 chips.

I understood from postings here that people were
confronted with this problem in software. I wouldn't
mind if the hardware were completely transparent.

I worked on an OS once for a computer that had
some of it's I/O-ports inverted for (perceived) speed
reasons. It's really very error prone to have to program
counter's, address registers and enable bits etc. in
DMA controllers etc. To keep the software more or
less readable you can translate everything after taking
it out of the chip and before putting it in, but it really
messes up your program. Doing a block move from
the registers to normal memory to save the complete
state of the device introduces tainted data in the
data structures of the OS. Hardware idiots should
really oversee problems like these before basing
there architecture-designs on tricks like these.
The error-chance and speed-loss in software may
be much more costly than even a slightly slower
clock speed of the device.

>Do you believe a logical one has to be represented
>with a particular voltage everywhere in a circuit?

Not FULLY internally.

>Are you boycotting machines that store text strings
>"backwards" in memory because memory dumps are hard
>to read?

Yes!

It is also one of the main reasons why I prefer the
Big-endian way of number representation.

>No?  Good!  Patterns in *21 are no different.
>
>> day. Technologies will progress and lame choices made
>> in the past because of hardware technological reasons
>> will hinder the future software developments.
>
>Excellent point!
>
>As bad as x86 was in the beginning, your collection
>of hardware, software, marketing and finance men had
>no choice but to make it a memory to register
>architecture.
>
>It's not that memory to register is a bad thing in and
>of itself, it just that the cost of memory would not
>support any other choice.

I don't understand you. Do you compare 'memory to
register' to RISC or against a stack-machine?

>> Computer architecture should always be a compromise
>> between hard and software. And therefore a compromise
>> between hardware men and software men.
>> 
>> And also a compromise with marketing men and
>> finance men etc.
>
>Yes, yes, yes, conventional wisdom, Ra, ra, ra ...
>
>On the other hand, you have all of these compromise
>elements in the MISC development environment.  Do
>you insist these elements be separate people or
>departments of a company?

I think that, as the RISC philosophy acknowledges,
new CPU architectures should be jointly designed
by hardware and software people. As regards the
software people, compiler builders, OS writers
and application writers should have their say or
at least the interests of the later two should should
be given consideration.
The AMD 29000 series for example has an array
of 128 registers, which are a software regulated
cache on the stack. However since it has to be
arranged in software it's very expensive and task
switches become very expensive. That's probably
why the Am29000 flopped as a general processor.
It later also flopped as an embedded processor and
AMD stopped supporting it in favor of their embedded
186 family.
 
>Your compromise team made other choices on the x86 too,
>some of which were real boat anchors for software
>development.  Great job guys!

Haha, you'll never hear me defend the x86. The NS320xx
is from around the same time and it's completely build
on orthogonality and has no segmentation etc. Intel
has never built a nice processor except perhaps the i960.

>I'll side with the philosophical rather than the
>committee approach, thanks just the same.

I agree, but all too often the hardware designers
ignore practical software demands.

>> Let's give up the MISC dream! It's over, get real!
>
>Now, there you've gone and burst my bubble!
>
>Lets say MISC in the *21 implementations are never
>commercial sucesses (whatever that means).  Was the
>MISC effort a waste?  I think not.  MISC is definitely
>thinking outside the box.

Yes, it was nice and interesting for a couple of years.


Groeten/Greetings,
Jaap

-- Chip Directory
-- http://www.chipdir.com/chipdir/
-- http://www.fh-kl.de/~rscherer/chipdir/ - New in Germany
-- And about 30 other mirror sites world-wide...
--
-- To subscribe to a free 'chip issues, questions and answers'
-- mailing list send a message to listguru@fatcity.com with
-- in the body subscribe chipdir-L