My experience with computers began back in 1970, when I travelled from Southern California to Tucson to attend the University of Arizona. I came to study Geological Engineering, not Computer Science (back in those days the computer folks studied Systems Engineering, or Math it seems). At any rate, I was registering for courses and pondered a possibility to round out my schedule, something called SIE 78, Fortran. I asked someone what in the world Fortran was, and soon I met my first computer.

The Control Data Cyber 6400

The 6400 in those days was the only computer on campus apart from some kind of Univac in the Math Department. The 6400 was a big mainframe in its own special room. It ran the SCOPE operating system ("Supervisory Control of Program Execution"). It was an interesting machine, even by todays standards. Here are some highlights: It was word addressed with 60bit words. Instructions were 15, 30, or 60 bits and got packed as efficiently as possible into the words. Characters were 6 bit display code, and you could cram 10 of them into a word by shifting and masking.

IO was done by a group of 10 peripheral processors that monitored some dedicated bit of shared memory watching for requests to be posted. The assembly language for the beast was known as "compass", and looks amazingly like some of todays load/store RISC instruction sets. I would lay odds this thing was designed by Seymour Cray before CDC ran him off and then ran the company into the ground.

The Digital DEC-10

Around about 1972 or so, the University acquired its second computer, which was the amazing DEC-10. This was an introduction to a whole new way of life, with editing sessions using the SOS editor (TECO was also available for the rugged of heart and mind), replacing keypunch sessions. It ran the TOPS-10 operating system and was really a fun machine to work with (which really couldn't be said of the 6400).

IBM's big iron

After I graduated school, I made a short sojourn to the University of Chicago, where at the time computing was done on IBM 370 mainframes. After TOPS-10, IBM JCL was more than any reasonable soul could stomach, and I pretty much went into computer hibernation and kept my mind on geology. When I did need to use these monsters (you had your choice of fortran compilers, the H-compiler, as well as other letter designations.), I borrowed prepackaged JCL decks from other veteran of battles with big blues big iron.

The Digital DEC-20

Events brought me back to Tucson and in the employ of "Computing Associates" developing software to serve the then prospering mining industry. We ran most of our jobs on CDC Cyber 6600 and 7600 machines in Houston, managing a public access terminal with a high-speed link to the machines in Texas. After a time CAI purchased a DEC-20 machine which was frankly one of the nicest machines I have ever worked on. It ran TOPS-20 (of course!) which had features that are only today being rediscovered. It had virtual memory. The DDT debugger was set up to load into a virtual address dedicated to it, which meant it could be invoked and attached to an already running task in a clean and straightforward way. The DEC-20 was again word addressed, with 36 bit words now, holding 5 characters of 7 bits (leaving an "extra" bit, the observant reader will note -- the editor used this to tag line numbers in files). The DEC-20 did have instructions that could handle strings and magically incremented right across that nasty extra bit, so you didn't really mind it after all. But we still weren't talking bytes yet, and octal dumps were the order of the day, as with the Cyber machines.

The Interdata 8/32

Back at the University of Arizona things weren't standing still. The folks at the computer center had managed to get a second DEC-10 and got the two DEC-10's talking to each other in some way that slowed the two of them down to where it was better to just have had one or the other of them. They also got one or both of them talking to the 6400, or whatever Cyber mainframe they had then. This was a good thing, since jobs could be prepared on the DEC-10 with an interactive editor and submitted over whatever link they had between the two machines. (This was back when men were men and networks weren't invented).

I found myself over at the Lunar and Planetary Lab working on an unusual machine known as the Interdata 8/32. This was what you would call a minicomputer. It didn't cost multiple millions of dollars, so a small department could afford it. It was sold by Perkin-Elmer and ran a nightmare of an operating system known as OS-32. I would ask the questions: Why didn't you just buy a VAX? and later, Why aren't we running UNIX on this thing? But some folks felt hurt by questions like this, so eventually I learned not to ask. We could have been running VMS or TOPS-10, but instead we had OS-32 that didn't even allow typeahead. Of course it was coded in assembly language for speed. :->

The Interdata could crank about 1 MIPS and sported a full complement of 1 megabyte of memory! They told me it was core memory, but I didn't believe it then and still don't. The interdata was byte addressed, and did have a segmented MMU. We had edition 7 unix on a tape, and the grand moment arrived when the previous system manager moved on to another job, and we replaced the pair of 80 megabyte "removable pack" disks (each was a washing machine sized monster), with a pair of 220 megabyte Fujitsu eagles. It turned out that OS-32 couldn't support the eagles so it was a fine chance to dust off the unix tape and make the switch.

Personal computers

While we were busy with unix on the Interdata, other folks were busy with Z80 computers, the S100 bus, and other forms of 8-bit desktop computing. This whole period pretty much just passed me by. But then my boss purchased one of the original IBM PC's and I had a shot at programming it. Once again an IBM product severely abused me, but I really cannot blame IBM, rather Intel and Microsoft conspired to repulse me. I honestly approached the whole business with an open mind, but between the segment register business from Intel and the bondage and discipline assembler MASM from microsoft, I was soon thoroughly disgusted.

Experiences with Data General

About this time, I jumped ship and went across the street to work at Steward Observatory. At the time their computing was primarily services by a Data General MV/10000 running AOS. True, this thing did run 2 or 3 mips, but it had the most disgusting internal architecture I have ever laid eyes on, bar none. Just for starters, it could address memory with byte addresses or word addresses, depending on just what instruction was being executed. This was the first machine I did work on that I did not attempt to program in assembly language. I browsed the manual and got a birds eye view of the architecture and that was plenty. I even managed to mostly ignore the operating system as well by running the unix emulator which had a C shell interface. This stood me in good stead and spared many healthy brain cells a terrible fate.

Water under the bridge

It is perhaps worth just stopping to ponder a valid and wholesome approach to some issues. Some stairs can be climbed two steps at a time. The computer industry moves fast enough that it is reasonable to just let some developments pass by. I pretty much let FORTH pass me by (tho it has tried to grab me on several occasions and I even put together a Threaded Interpretive language that ran on the 6502). Then there are DEC VAXen and the whole world of VMS that I managed to miss. Then there is the whole world of Z80 machines and CP/M - this too tried to grab me, but I soon arrived at a rule of life: Anything that doesn't have a hard drive isn't really a computer. And of course there is the world of Data General machines and AOS that I largely managed to ignore.

Sun workstations

As the MV/10000 passed into its golden years, the observatory began to work with these fine computers. Here is unix at its best. Here are wonderful machines with beautiful interactive graphics. And when we added our first sun4 to our sun3 collection the poor Data General was absolutely outclassed. I compiled my first little C program and couldn't believe the compile had really happened the prompt was back so fast. I was used to starting a compile with a script that would beep when it was done and pass the time reading a few pages in a book.

Intel revisited

I have been forced to reexamine the whole x86 architecture and the discovery is that an amazing piece of work has been done. The 8088 is a reasonable chip if you view it as an 8-bit processor on steroids. It is like a Z80 with a divide instruction and built in bank-switching features to get beyond the 64k barrier. The 8088 was perhaps the finest 8 bit processor ever made, but I have had more fun programming the Z80. The 80286 was brain damage from the word go, what else needs to be said. But the 80386 and up show an incredible tranformation. The giant wart of the segment registers is transformed into a transparent segmentation mechanism and paging and virtual memory capabilites have been added. These things can run a real operating system, so I do, I am running NetBSD on an x86 (where x86 is 486, 5x86, or P5 at different times), and it does a beautiful job.

The future

What lies ahead? Can I ignore C++ and leapfrog to the next language technology? Will java be the way of the future. Will I one day be running Linux instead of NetBSD. Stay tuned for answers to these questions and more!
Have any comments? Questions? Drop me a line!

Adventures in Computing / ttrebisky@as.arizona.edu