The World’s First Computer Operating System Implemented at General Motors Research Labs in Warren, Michigan in 1955
If you are looking for the first operating system, please read this (small) article firstly.
Steve Holland from General Motors Research Laboratories sent to me an interesting text about two of the three earliest operating systems: General Motors OS and GM-NAA I/O. I am publishing it here because when I tried to find some resources about this operating system, I realized that there are no a lot of sources on the Internet. It may be a good reference for writing relevant articles on Wikipedia. (I added links to Wikipedia articles.)
Citation: “FORTRAN Anecdotes,” IEEE Annals of the History of Computing, vol. 06, no. 1, pp. 59-64, Jan-Mar, 1984 [a portion of pp. 60-61 copied here]
There is an interesting, but little known, relationship between the first FORTRAN compiler on the IBM 704 and the early development of operating systems. In 1955, what was probably the first operating system had been developed and was in use at the General Motors Research Laboratories. Called the I/O system, it was a three-phase
system in that a batch of jobs was processed entirely with regard to input formats, then the entire batch was processed for execution, and finally the entire batch was processed for output conversion and printing. There were no higher-level languages in that system.
Shortly thereafter, IBM released the first version of the FORTRAN compiler as a set of programs on a magnetic tape, but without any listings of the source programs. The compiler itself consisted of a short bootstrap record on the tape, followed by many short records, each fully aware of the position of every other record on the tape, so that when a new procedure was to be brought in by loading another record from the tape, it was usually found by rewinding the tape and spacing out the appropriate number of records. Even this much organization had to be deduced without any documentation or listings.
Jim Fishman at General Motors did this analysis by obtaining an octal dump of the entire tape. He very carefully analyzed each instruction until he understood how the system worked. Without disturbing the order of the records, he then expanded some of them to include other components of what later became the excellent General Motors Research operating system. Thus the bootstrap loader grew from about 20 instructions to several thousand when it became the assembler preceded by the bootstrap loader in the same record. Later a relocatable loader was added, as well as other functions that we would now recognize as appropriate to an operating system. All this while FORTRAN never knew what happened!
A couple of years later, we adapted this operating system to our needs, beginning a long history of operating-system development at the University of Michigan.
Bernard A. Galler
University of Michigan
Our records corroborate the essence of Bernie’s anecdote. However, there was no need for an assembler in the first version of the F system since FORTRAN I object code could not be linked with other object code (except that object cards punched by FORTRAN I when edited onto the IBM FORTRAN could be accessed by subsequent FORTRAN programs as a function). It was the entire F system itself which replaced the first record of IBM’s FORTRAN tape. Note that FORTRAN issued a load-card sequence following each compile. This happy feature plus the “continuation cards” we placed in the card reader made this FORTRAN operating system possible. Also note that computer job billing was a fair accomplishment in 1957 at GMR, using a computer-readable real-time clock. I believe the first such clocks in the industry were built by GMR and Westinghouse, Pittsburgh. They used the echo entry hubs of the printer, and really were better than IBM’s first clock, on the 360 eight years later. It may be observed that GMR was not only the first site with a computer operating system (I/O system) but also the first site with more than one operating system (I/O system and F system), now apparently a perpetual encumbrance.
I also recall the first University of Michigan modification to GMR’s operating system. It consisted of replacing the “GMR” in columns 73-75 of all source cards with “UOM.”
J. J. Fishman
General Motors Research Laboratories
Footnote: Details of the first operating system are documented in a GM Research Staff internal memo #34-1028 “The IF System” written by Floyd G. Livermore and Jim Fishman of the Data Processing group in the “Special Projects” Department, dated November 5, 1958. (from Steve Holland, GMR)
~ by millosh on September 7, 2007.
Comments are closed.