> On Sat, 11 Sep 1999, J.S. Havard wrote:
> >You know, if there is one good the about Pee Cees, it is that whatever can
> >run on a 12MHz 386 can run on a PIII Xeon 500MHz box.  It may not be
> >optimized (at that speed, who could tell?), but atleast it will work!  As
> >great as the vax is, this is something that could have been better.  Low
> >level stuff.  There is so much work that must be put into getting it to
> >run on the next model, it just isn't even funny.
> 
>   Umm...that's because there's actually some *advancement* in the VAX family. 
> Intel's processor design isn't all that different from the '386 to the PIII,
> with the possible exception of the '586->PPro step.  And for all their marketing
> that says the PPro->PII or the PII->PIII were such revolutionary steps...bull. 
> Look at the design.  It's the same damn processor...the PPro went to the PII
> because they couldn't get their fab yields up with big on-chip caches, and from
> there on up to the PIII it's primarily a matter of increasing clock rates on
> the same tired design.  We're talking about the folks who integrated their FPU,
> added two simple instructions, and called it a "new processor" (386->486) and
> then, off the hook, rested on their marketing department for a couple of
> years...Sure, later they did a few really cool risc and post-risc things in the
> PPro (out-of-order execution in the core, instruction decomposition into u-ops,
> etc) but can that really hold up nearly fifteen years of processor evolution?
> 
>   Sure the VAX-11/780 executes pretty much the same instruction set as my zippy
> little VAX-4000/400...but compare the designs of the two processors.  Look at
> all the architectural variations...different I/O buses, different memory buses,
> different kinds of caches...sure, maybe they could have done better with
> compatibility, but at what expense?
> 
They should have done more with compatability.  Exactly my point!  And
yes, the x86 family is just ca-ca, crap, poo, <insert your favorite
epletive here>.  And from an architectural stand point, it is even worse.
Hopefully the IA64 won't be so bad.

What they should have done is created an abstract hardware interface,
where it would be easy to interface with everything.  Just a generic
interface.  Only speeds, I/O bandwidth and performance would differ.

What would be nice would be to create a generic hardware interface for
some new architecture, say 512 bits wide, in 32bit blocks or so.  In the
first generation of the architecture, you only use either the first 32 or
64 bits.  Later on, you can expand.  Also, you might be able to say, make
some I/O channels and use the whole 512 bits, although relying on that
would be risky, if future generations use wider data paths, up to the 512
bits.  Maybe use the channelizing only for real time and embedded systems.
Once you reach the 512 bit limit, just make a new architecture.  That much
advancement should require it.  Disk controllers should have a generic
interface, memory, serial ports, networking, etc also.  I should be able
to hardwarily acceses a scsi or ide or even mfm disk the same way.  Think
sort of like the Emulex qbus scsi host adapter.  It emulates MSCP.  Oh
well, I think my fantasy architecture would be nice, but nobody in there
right mind would make it. It is just too good of an idea to actually be
put in use.  It would be truly plug and play.  Cards plugged into the bus
would send some sort of identifier stating their type.  Based on slot
number, they would get priority (disks 0-7 for scsi controller in slot 1,
disks 8-11 for ide controller in slot 4, etc).  Blech!  I'm falling in
love with that thing, and it isn't even real.  :-)

>   [Not flaming you John, you make an excellent point, just a little friendly
> debate...I got out of bed in a weird mood today :-)]
> 
That isn't a flame.  Flames are "you fscking poopoo head."  This is
debate.  Debate is good.  Debate causes progression, but can develop into
mondo flamage if people take things personally, or get in a bigottess
mood.  That said, forget I ever said it, unless you want to quote me.  :-P
 
> -Dave McGuire > 

Regards,
John Havard