[Info-vax] RealWorldTech on Poulson

[glen herrmannsfeldt]

Bob Koehler <koehler at eisner.nospam.encompasserve.org> wrote:
> In article <cca3eb67-0d1d-458f-b633-e7cd144d6a76 at j15g2000yqf.googlegroups.com>, Neil Rieck <n.rieck at sympatico.ca> writes:
>> 
>> Personal Comment: We all know that CISC architectures were too
>> complicated to allow further development of computer technology.

>   Which is why x86 from Intel and AMD are now so slow even desktops
>   don't use them, right?

They have done some amazing things with x86, but I believe the
command was regarding VAX.

DEC had plans to continue in the IBM tradition of an "Architecture"
that would allow for growth for many years.

Note that current z/Architecture machines will still run OS/360
code assembled almost 50 years ago.  (The PL/I (F) compiler has
been run on current machines/OS.)   In some sense, IBM was lucky,
but it does seem that the VAX architecture had some mistakes that
it couldn't grow out of.   The 512 byte page was too small very
early, but seemed right at the time.  S/370 started allowing
both 2K and 4K pages, and later went to 4K only.  

Many of the fancy CISCy VAX instructions were already known to
be a problem in 1977.  I remember stories about POLY and subscript
bounds checking that far back.  While S/360 has some pretty CISCy
instructions (EDMK for one), it is not so far from RISC.
Only a few, fairly simple, addressing modes, and only three
instruction lengths, with the instruction length determined from
the first byte.   One can do out-of-order with S/360, as IBM
did with the 360/91 in 1967.  

The VAX instruction set was designed around assembly programming,
just at the time that people were giving up on assembler, except
for the lowest level device drivers.

-- glen