Text preview for : TIE6309-0453_1401coreStrech.pdf part of IBM TIE6309-0453 1401coreStrech IBM 140x TIE6309-0453_1401coreStrech.pdf



Back to : TIE6309-0453_1401coreStre | Home

ITEM NUMOER 6309-0453 (T.I.~.J
14 pages

;OATE August 5, 1963


AUTHOR Frans P. Oudheusden


1rITLE 1401 CORE: STRETCHING FOR TABLE LOOK-UP


SOURCE IBM CORPORA TION
1111 Connecticut Avenue
Washington 6, D. C.




Thla paper 10 In the author'a origlnal form.
The obJectlve In providing thlu copy Is to
keep you Informed In your fIeld of Interest.
Please do not dIstribute thIs paper to persono
outsIde the IBM Company.




---- -------_.... _-----_-...
.IBM .......CONFIDENTIAL




6309 0453
DISTRIBUT:ED BY
THE PROGnAM INFORMA TION DEPAR TMENT (TIE)
I.BM CORP.
11 Z EAST POST ROAD
VVmTE PLAINS, NY
TABLE OF CONTENTS

Page

Acknowledgment of Assistance Letter 3
Abstract 4
Introduction 5
The Single Character Conversion Technique 6-8
The Randomizing and Chaining Approach 9-10
The Multipass Table Look-Up 11-12
Figure 1 - Modulus 16 Arithmetic 13
Figure 2 - Core Storage Layout 14
Figure 3 - Binary to Octal Conversion Routine Flowchart 15
ABSTRACT




1. 1401 Core Str1etching for Table Look-Up

2. Frans P. Oudheusden
Scientific & Special Operations Program
Department 0:36
Data Processing Division
Washington Federal Region
1111 Connecticut Avenue, N. W.
Washington 6" D. C.
333-6700 Ext. 455

3. Unique table look-up techniques are described emphasizing compactness,
speed and flexibility.

The discussions cover a single character conversion technique, a randomizing
and chaining approach and a multipass procedure for tables exceeding core
storage capacity.

The single character conversion uses the four bit arithmetic of the 1401 core
addressing scheme. This is illustrated by an example of binary to octal trans-
formation. The randomizing and chaining technique continues this example
by converting the octal to 7094 language mnemonics.

The multipass approach, illustrated by an editing problem, utilizes the tape
load mode to identify errors with wordmarks.




Page 4
INTRODUCTION




This paper is intended to acquaint the reader with some unusual approaches to
three widely different table look-up requirements. It is improbable that the reader
will be confronted with similar problems on identical hardware, however, he may be
able to use a variation of the techniques discussed here. Therefore, the following
hardware requirements should not be interpreted as limitations to the application of
these techniques. The illustrations used require a 1401 with 8000 positions of storage
and the Advanced Programming Package. In addition, the first two examples need
the Column Binary Feature.




Page 5
THE SINGLE CHARACTER CONVERSION TECHN.IQUE



To illustrate this technique, let us assume a 1401 Utility Program is needed to
print an octal dump of a binary tape written by the IBM 7094. Since the 7094 has a
36-bit binary word, a word is written on tape as six binary characters of six bits each.
The bi nary to octal conversion routine must therefore substitute for every character
read in from tape, two octal digits representing three binary bits each. This conver-
si'on can be accomplished rapidly by emplloying the 1401 core addressing scheme.

A brief review of the makeup of the 1401 core storage address codes is desir-
able. The 1401 has a three character address. The numeric values of the three
characters range from 000 to 999, thereby specifying one out of 1000 locations.
The zone bits over the hundreds' and units' positions of the address indicate by mod-
uJus 16 arithmetic (figure 1) which of the sixteen groups of 1000 core locations is
designated. The zone in the units position of the address specifies which quadrant
of 4000 storage locations is referenced, while the zone in the hundreds position
determines which group of 1000 locations is referenced wthin a quadrant.

The binary character to be converted is used to formulate a unique core
address, which then locates the proper two digit octal representation in a table
(figure 2). Ideally th is character is placed directly in the hundreds position of an
index register used to access the tabl e . This wi II not work in all cases, however,
since only forty of the sixty-four possible character configurations are val id storage
address codes. To clarify this procedure let us first examine how this technique words
for the forty val id address codes.

The table of two digit octal constants for these forty character configurations
is distributed from locations 4087 to 7987, separated by increments of 100 locations
(figure 2) For the conversion, index register 1 is preset to a value of 087 (step 1)
e


and the character to be converted is moved into the hundreds position of the index
register (s