CDC 6600 - Why so awesome?

[Chuck Guzis]

On 06/21/2016 02:06 PM, Swift Griggs wrote:

> - It had some wicked cool "demos", to cop a C64 term. (ADC, PAC, EYE)

Those were mostly toys to amuse the CEs, like the baseball game BAT.

Chess 3.0 was implemented on Northwestern's machine and probably was the
first computer chess program of note.    This was before kids thought
that computer games were *cool*.  I never developed a taste for computer
gaming.

Much of the architectural concept was shared with IBM 7030 STRETCH
(another system worth researching).  Hand-timing instruction sequences
on the 6600 was an art with its own rewards.

> - It wasn't DEC and it wasn't IBM and it was faster than both when it hit 
>   the street? 

With a 10 MHz clock.

> - It has a cool OS? Dunno. Not much info on "SCOPE"

It had several *cool* OSes, but really only two major ones for general
consumption (Special Systems Dvision had several more).  SCOPE (later
NOS/BE), pretty much initially a PP-resident OS based on the old
Chippewa Operating System--and NOS (was KRONOS, originally MACE),
started as a "bootleg" project by Greg Mansfield and (Dr.) Dave
Callender at Arden Hills.  (MACE stood for "(Greg) Mansfield's Answer to
Customer Engineering".  MACE took the then-mostly PP-resident SCOPE and
transfered various non-I/O system management functions to the CPU.
Purdue also contributed quite a bit of code.   Most batch programs
written for SCOPE would run fine on MACE with few, if any,
modifications.  In retrospect, CDC keeping two operating systems (SCOPE
was part of CPD in Sunnyvale, while KRONOS stayed home in Arden Hillls)
was probably a strategic blunder, since much duplicate effort was
wasted.  Eventually, the two were merged into NOS (for Network Operating
System).

> - It used odd sized (by todays standards) register, instruction, and bus 
>   sizes. 60 bit machine with 15/30 bit instructions. But, didn't it cause 
>   a bunch of alignment issues for you ?

There aren't any alignment issues, since the CPU was only
word-addressable.   This was when a character was 6 bits (think IBM
709x, UNIVAC 1100, etc.)  So a word with 10 characters was logical.
Given that PP words 12 bits (5 to a CM word) and there were 10 PPUs,
each executing at a speed 1/10th the CPU, it had a very pleasant sort of
symmetry.

> I dug into the CPU instructions for about 20 minutes and it was actually 
> pretty straightforward. The so-called "COMPASS" ASM code was oh-so-cool. I 
> can't believe they had so many of the features now considered "modern" or 
> "clever" (at least by me) in the 1960s! This code:

COMPASS was indeed advanced for its time, but then so was OS/360
assembly language.  Given that assembly was the lingua franca of system
programming, assemblers had to be good.  Most of the readability was due
to attention to detail by the programmer, not any particular language
feature.

> ... Is super-readable, in fact, probably a bit more than several 
> much-newer dialects on different platforms. There was one instruction 
> "PROTECT" I found pretty interesting, too. 

Where did you find that?  I've never heard of such an instruction.

--Chuck