Design similarities with CDC mainframe?

rhb1

This is a very interesting design, to me at least.· In many ways it resembles the design of the peripheral processing units of the Control Data 6600/6400 mainframes of the 60's and 70's that I worked on.· I believe that Seymor Cray designed those.· From what I remember, each "processor" (cog) had its own memory and registers, but shared the processor arithmetic logic during its time slot in the "barrel" (hub).· They each could access the I/O channels, but I believe that the software was written so that only one used any given channel.· One processor was dedicated to driving the keyboard and dual vector graphic displays.· They were also able to access the large main memory and communicate with each other and the central processing unit(s).· All of this was done by message passing (no interrupts).

All in all, a very cool machine with many·inovative design aspects and, for a time at least, the fastest computer on the planet.· There is an on-line copy of "Design of a Computer - The Control Data 6600"[size=-1] by J.E. Thornton (another Control Data engineer) at:[/size]
[size=-1][/size]
[size=-1]http://ed-thelen.org/comp-hist/DesignOfAComputer_CDC6600.pdf[/size]
[size=-1][/size]
[size=-1]The peripheral processing unit discussion starts on page 141.[/size]

Would anyone at Parallax like to comment on the similarities?

Oh, and I want one too.

Martin Hebel

I don't know much about this in general, but the cogs are full CPU's in themselves having ALU's. The only thing that is shared is the location of data & code (when using Spin - In assembler it all resides in the cog itself).

So, a little difference I see.

-Martin

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Martin Hebel

Disclaimer: ANY Propeller statements made by me are subject to my inaccurate understanding of my limited time with it!
Southern Illinois University Carbondale -Electronic Systems Technologies
Personal Links with plenty of BASIC Stamp info
and SelmaWare Solutions - StampPlot - Graphical Data Acquisition and Control

Paul Baker

While there are some similarities, it isn't that significant. The similarity is the round-robin abitration thats going on, but that is a very old concept predating even the CD6600. Also the PPUs have dedicated interfaces with external elements, the propeller doesn't. All cogs have access to all external pins simulataneously.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10

Bruce Bates

Paul -

You are 100% correct. The "commutator" method of processing pre-dates computers as we know them today. These same concepts were used in a mechanically timed fashion on the original EAM (electronic accounting machines) and hand wired (actaully plug wired) EAM machines that I originally learned on nearly 1/2 a centruy ago! Ah yes, the mighty IBM 85 collator, and 402 accounting machine - my favorites!

Essentially, even the venerable RPG compilers used a commutator (software timed processing) method of operation. Ever think of that?

Regards,

Bruce Bates

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
<!--StartFragment -->

Post Edited (Bruce Bates) : 2/23/2006 7:32:38 PM GMT

Paul Koning

PaulB,

You're not quite correct on the I/O for the CDC PPUs. Instead, it is just like in the Propeller -- every PPU has access to every I/O channel.

The similarities are pretty neat. Memory capacity is fairly close (4k by 12 bits times 10 vs. 512 by 32 times 8 plus 8k more). I/O fairly similar (12 channels at 12 bits each vs. 32 bits). The Propeller is significantly faster, though (40 MHz vs. 1 to 4 MHz).

By the way, the closest analog is the CDC 6416, which was just the PPUs, no CPU.

Paul

Paul Baker

I stand corrected, Im too young to have worked with mainframes. The closest I came was with the VAX (don't ask which model, I just remember it wasnt a microVAX), and the series of SGIs starting with the Onyx then the Challenge XL and finally the Origin 2000 (sad to see they are heading towards bankruptcy, those machines were phenominal).

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
1+1=10