Browse Prior Art Database

PRINTING AND CHECKING FOR LINEAR PROGRAMMING CODES

IP.com Disclosure Number: IPCOM000128843D
Original Publication Date: 1956-Aug-23
Included in the Prior Art Database: 2005-Sep-19
Document File: 5 page(s) / 92K

Publishing Venue

Software Patent Institute

Related People

Judd, H.A.: AUTHOR [+3]

Abstract

The print subroutine in the LP code is quite dependent on the master code both for instruction and for data. It is never called into high speed storage unless the data is ripe for printing. It does not have to know whether the printed output will prove a good or bad omen for the operator.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 26% of the total text.

Page 1 of 5

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

PRINTING AND CHECKING FOR LINEAR PROGRAMMING CODES1

H. A. Judd

International Business Machines Corporation

P-925

August 23, 1956

The RAND Corporation 1700 MAIN ST., SANTA MONICA, CALIFORNIA

PRINTING AND CHECKING FOR LINEAR PROGRAMMING CODES

H. A. Judd

The print subroutine in the LP code is quite dependent on the master code both for instruction and for data. It is never called into high speed storage unless the data is ripe for printing. It does not have to know whether the printed output will prove a good or bad omen for the operator.

An individual print-out consists of m + 4 lines of information. The first line is the problem identification that the operator assembled when the data was loaded originally. The second line specifies the current iteration number, the stage number, the form number (i.e., the row index of the current objective form being optimized) and the caption describing the type of print-out. There are 14 different captions, each of which is intended to be descriptive of the reason for printing. On five of the print-outs, the program prints S IS XXXXX on the second line so the operator will know that the activity XXXXX came into the basis on the completed iteration or cycle.

The third line is a set of headings which identify the different columns of printing on succeeding lines. The first column is labeled J and denotes the activity names for those columns which are currently in the basis. The second column is called BETA, indicating the current solution. The third column is I for the row index i which runs from 0 to m. The fourth column is variable, depending on the type of print-out. It may be headed by one of the following: B for bi (or bi(T) in PLP)

E for ηri(T)

G for γi(T-1)

A for ais(T).

The fifth may be

ER for εi (error)
PI for πi(T)

1 Prepared for The RAND Corporation Short Course in Computational Aspects of Linear Programming, September 4 - 13, 1956.

Rand Corporation Page 1 Aug 23, 1956

Page 2 of 5

PRINTING AND CHECKING FOR LINEAR PROGRAMMING CODES

or, for one type of print-out, blank.

The fourteen captions with the corresponding heads for the fourth and fifth columns are:

CYCLE PRINT S IS XXXXX E PI NEW SOLUTION S IS XXXXX B PI CHECK SOLUTION B ER END OF PHASE ONE B ER END OF STAGE B ER NO FEASIBLE SOLUTION B ER FEASIBLE SOLUTION B ER OPTIMAL SOLUTION B ER PRIMAL-DUAL SOLUTIONS B PI UNBOUNDED SOLUTION S IS XXXXX A PI RIGHT HAND SIDE OPEN S IS XXXXX G PI MATRIX SINGULAR S IS XXXXX A

BASIS INVERTED B ER THETA AT MAXIMUM B ER Before describing each of these types of print-outs, it seems necessary to point out that on-line printing is not the only conceivable way of getting information out of the machine. Printing is very time consuming or, stated in simpler economic terms, expensive, as compared with computing time. If too much information is printed, it becomes very repetitious and boring. If too little is printed, a part of the problem s...