Time-sharing on the Ferranti-Packard FP6000 computer system
Michael Marcotty, F. M. Longstaff, Audrey P. M. Williams · 1963
A major advance in the design of computers was the provision of buffered or autonomous peripheral transfers which enabled computation to proceed concurrently with a transfer. In many computations, however, the total saving obtained is quite small; for instance, 10 milliseconds of computing overlapped with 100 milliseconds of transfer provides a gain of only 10% and the central processor must remain idle for the remaining 90 milliseconds. An extra burden is also placed on the programmer who must arrange his work so that computation can take place while the transfer is occurring. This is enough of a task when the program is written in machine language but, when the program has to be compiled, the compiler has to be considerably more sophisticated in order to reach the same efficiency and, as a result, compilation will take longer. The problem boils down to an economic one of weighing the cost of program production against the cost of machine time gained by the extra efficiency.