AiG interviews Brian W. Fuesz, the creator of the Cygnus-1

For those people unfamiliar with the Cygnus-1, visit the AiG page on the subject.

[Brian W. Fuesz]  I developed Cygnus-1 to serve as a multi-processor, parallel computer system. As I designed the board, I included features that would also allow it to function as a single board computer for embedded systems applications as well.

How long did it take to develop?

[Brian W. Fuesz]  I developed the board over a period of about six months, although there was a nine month period during which I developed two wire-wrapped prototype systems.

The 68000 is considered by many to be outdated. Why did you choose this as the basis of the simulator?

[Brian W. Fuesz]  I used the 68k because I preferred the 68k-series architecture and assembly language over the Intel x86 family, which is more common because of its use in PCs.  I also selected the 68k because I had several Amiga systems and associated development software, which was useful during development of a 68k-based system.

What factors led you to use an Amiga as the development system?

[Brian W. Fuesz]  At the time, I did not have a PC, so I used the Amiga.  Plus, I was able to leverage off of my 68k assemblers and C compilers for the Amiga.

What were the Amigas system specification that the software was developed on?

[Brian W. Fuesz]  I used an Amiga 1000 with 2MB RAM and two floppy drives. I did not have a hard disk drive.

What software did you use on the Amiga?

[Brian W. Fuesz]  I used Amiga assemblers and C compilers. I also designed and built a wire-wrapped EPROM burner that utilized the Amiga's parallel port.  Code that was developed on the Amiga was burned onto EPROMs using this device.

Did you discover any problems caused by using the Amiga as a development system?

[Brian W. Fuesz]  No, it actually worked quite well.  I wrote a loader program to take programs written on the Amiga and transfer them to a fixed memory location in the EPROMs for Cygnus-1.

How easy was it to get developer information and software on the Amiga?

[Brian W. Fuesz]  I used commonly available Amiga systems-level documentation.

What lessons did you learn from creating the Cygnus-1?

[Brian W. Fuesz]  I was able to develop a fully functioning computer system.  However, the goal of making a parallel computer was overtaken by microprocessor technology.  When I first started on the project, a PC cost about $2000. I figured that I could produce a Cygnus-1 board in quanities of four boards for about $130 each (materials cost only, not including the labor to assemble and test the boards or the non-recurring engineering development costs). Thus, I could put together a twelve processor Cygnus-1 system for about $1560, which would be cheaper than a PC and would be faster, too.  While this was true when I started the project, by the time I had finished, a PC with equivalent compute power to the twelve-processor Cygnus-1 system, cost about $1200.  So Cygnus-1 lost its price appeal.  In addition, a single processor PC was much easier to develop software for than the multi-processor Cygnus-1 system.

The project used the 68008 and 68000. Were there ever any plans to using a faster version of the 68k family?

[Brian W. Fuesz]  I had considered using a 68020 or 68030, but the economics at project completion made such a new development very uneconomical.

Can you break down the total cost of producing the unit?

[Brian W. Fuesz]  As I said earlier, it costs about $130 to build a board.  I don't recall details at the component level.

Finally, what tips would you give for anyone wishing to develop a similar system on the Amiga?

[Brian W. Fuesz]  Well, if anyone wants to do something similar, I would encourage them to use a Motorola 68HC11 microcontroller instead of a 68k.  However, if they want the leverage off of their Amiga software tools, then a 68k is fine. I'd recommend the EC version of the 68k. As I recall, this processor can be wired to use an 8-bit databus, which I would recommend -- it requires much less board space and you really don't typically need the gain in speed form a 16-bit databus.

Oh, and they had better be prepared to consume alot of hours in design, development, and test. :-)

BACK TO INTERVIEWS PAGE