Page:Alan Turing - Proposed Electronic Calculator (1945).pdf/14

 Now let us consider what is done right at the beginning. Arrangements are made for setting into CI and CD a certain invariable initial order and IN. These state that the card is to be transferred into a particular delay line, and that the next order is to be taken from a particular spot, which will actually be in this same delay line. The information in this delay line can contain sufficient orders to ‘get us started’. The first few orders obeyed will probably be to take in a few more cards. The information on these will later be sorted to its final destination. When the final instructions are in place it will be as well to ‘read them back’.

Actually it has been arranged that the special initial order consists of 0 throughout so that there is no need to set it up.

(iv) Binary-decimal conversion. It is proposed to do binary-decimal and decimal-binary conversion as ITO. This will be appreciably assisted by the fact that short multiplication is a CAO.

(v) Instruction-table cards. It was explained in connection with the input organ that the instructions would be on cards, of whose columns all but 32 were available for external use. A proposed use of the 80 columns is suggested below, without proper explanation; the explanation comes later.

Of these the genuine input has already been spoken of to some extent, and will be spoken of again further. The job number and the spare columns do not require explanation. The popular data describe the instruction in letters and figures in a manner appropriate for the operator to appreciate quickly if for instance the cards are listed. In this respect we might say that the popular data is like a telephone number Mol 1380 whereas the genuine input is like the pulses used in dialling: indeed we shall probably carry the analogy further and really only distinguish 10 different letters, as is done on automatic exchanges. The popular data have also another important function, which only appears when we consider that the same instructions will be used on quite different jobs. If we were just to number the instructions serially throughout all the instructions ever used on any job, then, in the set of instructions actually used in any particular job there would be large gaps in the numbering. Suppose now that these instructions were stored in the DS with positions according to their numbers there would be a lot of wasted space, and we should need elaborate arrangements for making use of this space. Instead, when a new job appears we take the complete set of cards involved and make a new copy of each of them; these we sort into the order of popular group name and detail figure. We then renumber them consecutively in the binary scale. This number goes into the columns described as ‘repeat of destination’. The renumbering may be done either with a relay counter attached to a collator, or by interleaving a set of master cards with the binary numbers in serial order. To complete the process we have to fill in other instruction numbers in binary form into the genuine input, e.g. if an instruction in popular form were/