[Simh] VAX vectors

[Clem Cole]

On Thu, Jul 9, 2015 at 4:53 PM, Richard <legalize at xmission.com> wrote:

> VAX was pretty much the last CISC-y architecture that sold in any
> sizable units
>

​68K family was after the VAX (  VAX in mid-70s vs. 68K is late 70s).  Both
pre-John Cox or Patterson's papers.   In fact, as Les Crudele [HW lead of
the 68k] once said to me,  the ISA and HW interfaces were picked to try to
emulate​ PDP-11 without tripping over the DEC IP (DEC had recently put
Cal-Data out of business). [ Many of the 68K team was ex-Schlumberger folks
who had been using PDP-11's for process control ].

That said - Vax was Dave definition of a CISC system and he completely
missed the point.  The Vax (and 68K) were wonderful systems and super
successful.   Dave also missed Cox's point BTW.  John's paper never asked
to reduce the instruction set, he just said - why not let the compiler
generate directly to the microcode - which seems to be a simpler target?

I used to say, any ISA (particularly based on a Load/Store scheme) that was
designed post Patterson was by definition a RISC for political / marketing
reasons.

BTW: VAX/68K vs Alpha/MIPS or RISC vs CISC is not really the question.
When Patterson et al wrote the RISC papers I do not think he understood the
economics of CPU architecture design.

What you are observing is that in today's world the complexity or not of
the ISA does not matter.  If you peal back the layers of any modern chip,
be it Xeon or Alpha they look surprising similar.  They are data flow
machines under the covers, with lots of different parallel execution units
being scheduled and used as efficiently as possible.   Both are equally
complex and both are trying to do all sort of complicated things to keep
those functional units doing useful things.    Thus the whole RISC vs CISC
debate is actually moot.   FYI:  My friends in gate laying land tell me
that the INTEL*64 as an ISA pays a few % silicon tax for one of the less
than friendly ISA's but ... by paying that tax, brings the economics of the
Intel ecosystem to the party.

I loved the VAX and the 68, and also loved Alpha.  They all had wonderful
technical things they brought to the table.   The fact is that Alpha was a
beautiful and technologically excellent ISA and as a computer scientist, it
has never been matched.  But Alpha (and Vax and 68) lost an economic war
not a technical one.

Clem
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20150710/5e285e0d/attachment.html>