previous | contents | next

Chapter 11

The Pilot ACE1

I. H. Wilkinson

Introduction

A machine which was almost identical with the Pilot ACE was first designed by the staff of the Mathematics Division at the suggestion of Dr. H. D. Huskey during his stay at the National Physical Laboratory in 1947. It was based on an earlier design by Dr. A. M. Turing and its principal object was to provide experience in the construction of equipment of this type. It was not intended that it would be used on an extensive programme of computation, but it was hoped that it would give practical experience in the production of subroutines which would serve as a useful guide to the design of a full scale machine. An attempt to build the Pilot Model, during Dr. Huskey's stay, was unsuccessful, but a year later after the formation of an Electronics Section at the NPL a combined team consisting of this section and four members of the Mathematics Division started on the construction of a Pilot Model, the design of which was taken over almost unchanged from the earlier version. The machine first worked, in the sense that it carried out automatically a simple sequence of operations, in May 1950 and by the end of that year it had reached the stage at which a successful Press Demonstration was held. The successful application of the machine to the solution of a number of problems made it apparent that, in spite of its obvious shortcomings, it was capable of being converted into a powerful computer comparable with any then in existence and much faster than most. Accordingly a small programme of modifications was embarked upon early in 1951, but the machine was not functioning satisfactorily again until November of that year. After a month of continuous operation it was transferred from the Electronics Section to Mathematics Division where it has since been in use on a 13-hour day. During its first year of full scale operation it achieved a 65% serviceability figure based on a very strict criterion. Its performance during its second year has so far been considerably better than this.

General description

The Pilot ACE is a serial machine using mercury delay line storage and working at a pulse repetition rate of 1 megacycle/sec. Its high speed store consists of 11 long delay lines each of which stores 32 words of 32 binary digits each, with a corresponding circulation period of 1024 microseconds, 5 short lines storing one word each with a circulation period of 32 microseconds and two delay lines storing two words each. It was inevitable that in the design of a machine originally intended for experimental purposes, overriding consideration should be given to the minimization of equipment rather than to making the machine logically satisfying as a whole. This is reflected to a certain extent in the code adopted for the machine and in its arithmetic facilities, which are in general fairly rudimentary. The design of the machine was also decisively influenced by the attempt to overcome the loss of speed due to the high access time of the long storage units. The machine in fact uses what is usually known as a system of "optimum coding."

Code of Pilot ACE

The Pilot ACE may be said to have a "three-address code" though this form of classification is not particularly appropriate. Each instruction calls for the transfer of information from one of 32 "sources" to one of 32 "destinations" and selects which of eight long delay lines will provide the next instruction. This third address is necessary because consecutive instructions do not occupy consecutive positions but are placed in such relative positions that, in so far as is possible, each instruction emerges during the minor cycle in which the current instruction is completed. An unusual feature of the instructions is that the transfers they describe may last for any number of consecutive minor cycles from one to thirty-two. The instruction word contains three other main elements which are known as the wait number, the timing number and the characteristic which together determine when the transfer starts, when it stops and which instruction in the selected instruction

1Automatic Digital Computation, National Physical Laboratory, Teddington, England, pp. 5-14, March, 1953.


193

previous | contents | next