Before computers had boot roms they had switches, sometimes lots of switches, that entered the first few instructions to get them going. Here I remember the Control Data CDC 6500 I programmed at Purdue.
YOUTUBE 8zk1zByXpBk Published on Dec 15, 2013. The Living Computer Museum now has possession of my CDC 6500. It was turned off in 1989 after two decades of service at Purdue University.
The machine ran batch and remote batch operations 23-hours a day. One hour each work day, four pm I remember, systems staff took over operations for testing new software. I tested fixes to the paper-tape reader device driver during this "systems" time.
Part of the ritual during systems time was getting to push the small red dead-start button located under and between the two radar tubes that made up the operator's console. Dead-start loaded and ran the boot program from twelve rows of twelve toggle switches located on the back of the cpu cabinet across the room.
Twelve instructions were enough to address a mass-storage device, read a block and then run that. About the only switch manipulation was to select a different device, a tape for testing or disk drive for the rare production restart.
The 6500 had two model 6400 cpus, a cost reduced version of the brilliant 6600. A third small cpu was multiplexed so as to appear as 10 slower peripheral processors that could be dedicated to real-time or dynamically loaded for input-output chores.
A quirk of the machine was that if both cpus concentrated on erasing the dead-start program then the boot instructions to halt the cpus could be rendered useless and the machine could not be stopped.
CDC offered a high-speed paper tape reader that was built like a magnetic tape reader but with an optical read head instead. Without record gaps once the paper started moving a peripheral processor dedicated to reading until the tape was done. This meant 10% of the supercomputer's I/O bandwidth was consumed reading characters at a few hundred a second.
The peripheral processor instructions were much like the minicomputers of the day. I was messing with paper tape because I was cross-assembling for a friend's PDP-8 on the CDC-6500. Paper was the only common medium. My fix to the driver was a couple of instructions. Senior staff guided me through testing it.
See IBM 7094 Console for more switch register lore.
See PDP-8 Simulator for running PDP code on the CDC.