Perspectives on computers: What they are and what they do
To avoid getting lost in chemistry and physics, things "are" what they "do". In this talk I have been depicting the set of tasks which seem evident to me to be general descriptions of what people now want computers to do for them.
Computers, of course, are machines, and the programs that direct the actions of these machines are algorithmic schemata.
But the new idea in the industry is the recognition that machines and programs "are" their representation, and with a common representation for both the machine and its program, we can concentrate our research attention on the objectives desired of this system, analyze the structures needed to accomplish those objectives, and supply only that hardware and software required.
An example of this arises when one asks the question — "What is the purpose of memory in a computer?" If decision-making is what computers do, then the only purposes of memory in that process are to store the permanent data in its data base and to provide only that temporary storage necessary to effect its decision-making stratagems. Thus future, well designed computers, should be expected to use less working storage memory, and this memory is likely to be distributed throughout the decision-making apparatus. The first trend is counter to current operating experience in which we find that with contemporary machines we obtain better performance if we use more main memory concentrated at the center of a system, but the second trend prediction is supported by the growth of cache memories and buffer memories being distributed throughout all new system designs. It will be interesting to see how this prediction of the role of memory will work out in future system designs.
In summary then, it is my belief that the purpose of computers is to help people make decisions, that the computers and their programs achieve this purpose by making a series of interconnected decisions, and that our best understanding of what computers and programs are is achieved by studying what they do.
Unable to display preview. Download preview PDF.
- (1).Bachman, C. W., and Bouvard, J., "Architecture Definition Techique: Its Objectives, Theory, Process, Facilities, and Practice", Proc. ACM SIGFIDET Workshop, Data Description, Access and Control, Nov. 29–Dec. 1, 1972.Google Scholar
- (2).Chu, Y., ACM-IEEE Symposium on High Level Language Computer Architecture, University of Maryland, November 7–8, 1973.Google Scholar
- (3).Computer Architecture Conference, December 9–11, 1973, Gainsville, Florida (IEEE, ACM).Google Scholar
- (4).Dijkstra, E., Dahl, O., Hoare, C., Structured Programming, Academic Press, 1972.Google Scholar
- (5).Johnson, R. R., "Some Steps Toward an Information System Performance Theory," Proc. First USA-Japan Computer Conference, AFIPS, 1972.Google Scholar
- (6).Johnston, J. B., "The Contour Model of Block Structured Processes", DSIPL, (1971).Google Scholar
- (7).Kotov, V. E., "Practical Realization of Asychronous Parallel Computations", Review of Systems and Theoretical Program Siberian Academy of Science, Novosibersk, 1972, page 110 (in Russian).Google Scholar
- (8).Organick, E., Computer System Organization, The B5700/B6700 Series, Academic Press, 1973.Google Scholar