②98% PROGRAMMING CONSIDERATIONSA st

②98% PROGRAMMING CONSIDERATIONSA study was made to determine how Multiple- UDOFT with three-address code will compare with the one-address code UDOFT. For this study, it was as- sumed that the Multiple-UDOFT will have a minor cycle equal to that of UDOFT, and that the new in- structions will require the same number of minor cycles for execution as do the instructions for UDOFT. It was found that there will be no saving of time with the three-address code for simple sums of variables be- cause only one variable can be obtained from the mag- netic-core memory during a given reference cycle. The three-address computer will be at least as fast as UDOFT for polynomial evaluations. For sums of squared variables and for O33 quadrature formula evalu- ations, it provides a saving on the order of 30 per cent of the time required by one-address UDOFT; and for updating set of three time-different values of the same variable, the saving is on the order of 25-40 per cent. These comparisons were made on the basis of a UDOFT without subroutines, since the Multiple-UDOFT cannot have a subroutine structure (because of the drum). Under true UDOFT conditions, subroutines would be used. This represents a considerable saving in pro- gramming and instruction storage space at the expense of computation time. Thus, from the more realistic comparison of Multiple-UDOFT without subroutines vs UDOFT with subroutines, if was found that Multiple- UDOFT saves on the order of 15 per cent of the time required by UDOFt for a typical "Aight"path com- putation.