COURSE

003
ADVANCED PROGRAMMING
prepared by the Training Section,
Electronic Computer Department
aJTLINE FOR COORSE 003
ADVANCED PROGRAMMING
The purpose of Advanced Programming Cours.e 003 is to complete the basic ,:d$scription
of the Univac System and describe the advanced programming techniques
for the prepatation of efficient data processing routines. Each student
qui red to have satisfactorily completed Course 002 which is concerned with ,a de-
scription of the Univac central computer, and basic programming.
The subject matter of Course 003 consists of four major sections, each of which
is briefly described below3
1. Advanced Programming - This area of study covers the programming
commonly employed in efffcient use of input-output buffers, problem reconsti-
tution, item layouts, relative coding, reduction of instruction execution
times, obtaining increases in available memory space, use of service routines
for detection and correction of programmfng errors.
2. Programming for Heavy Auxiliaries - This area covers by description and illus-
trative problems, programming for the heavy auxiliaries of the Univac System,
such as the Card-to-Tape Converter and the High-Speed Printer.
3. Central Computer Logic - A detailed description of the central computer's
logical construction is the subject matter of this study area.
attention is given to the detailed analysis of how each instruction is exe"!"
cuted. The checking circuits of the Univac and the method whereby the super-
visory control may be used as a programmer aid in debugging and efficiently
running data processing routines. '
4. Thesis Problem - This area consists of workshop sessions wherein each student
selects a moderately complex problem for his thesis. student will ana-
lyze, flow-chart, code, and debug on the Univac System (Univac #7 - Service'
Bureau, Remington Rand Inc. , New York, New York) a problem involving from 600-
800 instructions. This thesis work is designed to provide further experience
in all phases of computer programming and -is conducted under the supervision
of the instruction staff assigned to his course.
A topical outline is presented belowo The amount of time spent on each subje,ct
may vary somewhat among the different instructors and classes. Items 1 and
above are designated under A in the outline, 3 above is topic B, and 4 is topic C.
-1-
MONDAY:
TUESDAY:
WEDNESDAY:

FRIDAY:
A"
Bo
Ceo
Ao
B ..
C
Q
A.
B ..
Co
A.
Bo
C ..
Ao
WEEK I
ANALYSIS OF COMMON CODING ERRORS
1. Types of errors and their manifestations
Errors due to neglected instruction characteristics
3. Improper ,overflow counts
4. Errors in flow chart logic
. l
REVIEW AND INTRODUCTION TO CIRCUITR Y
10 Memory and word structure
20 Binary and XS-3 notation
3. Four-stage cycle of operation
4. Description of logical circuit
THESIS mOBI:.EM
DESCRIPTION OF BIeR RELATIVE CODING
TIMING AND SYNCHRONIZATION IN UNIVAC
10 Minor cycle
20 Cycling unit
3. p and t notation
elements
4. Synchronization of memory and registers
50 Maintenance of synchronization in memory to regis-
register to register, and register to memory
transfers o
THESIS mOBLEM
BIOR RELATIVE CODING (CON9 T)
FOUR-STAGE CYCLE
10 Description of Chart U 102 and EBU-IOO
2"
(1 time
3., ::.HSB O-E checkerS! di fferences
negative type checkers.
THES IS PROBLEM
BIeR RELATIVE CODING (CONCLUDED)
FOUR=STAGE CYCLE (CON
9
T)
10 time
2. Channel selection circuits
30
Time selection circuits
T HES IS PROBLE M
SPECIAL INPUT-OUTPUT TECHNIQUES
1" Preselector
20 2-way automatic
3. Reversal
40 automatic
between positive and
B. FOUR-STAGE CYCLE (CONUI)
10 Y and D time
2. Start circuit
Co THESIS PROBLEM
-2 ...
MONDAY:
TUESDAY:
WEDNESDAY:
THURSDAY;
FRIDAY:
WEEK :2
A. INPUT-OUTPUT (CONCLUDED)
1. Efficient transfers for items a sub-mul tiple of 60
2. Efficient transfers for. items not a sub-mul tiple of 60
3. Variable length items
Bo TRANSFER ORDERS' - TYPE 1
10 B ~ H, K as typical of 1 word transfers
20 Error indications MDS" and DS'S
c ~ THESIS PROBLEM
Ao IDENTIFICATION BLOCKS' AND RERUNS
1. Standard data tape format
2. Generalized two-way merge
B. TRANSFER ORDERS - TYPE 1 (CON'T)
10 V, W, Y, Z as typical of mUlti-word transfers
20 Interchange in V-W transfers
3" R order
C. THESIS PROBLEM
A. IDENTIFICATION BLOCKS AND RERUNS' (CON'T):
10 Generalized two-way merge (con6luded)
2" Rerun procedure for 1 output
B. TRANS'FER CRDERS' .., TYPE 2
1. on, On as typical of shifts
2. Improper shifts causing a stall
30 Extract order
C. THESIS PROBLEM
A. IDENTIFICATION BLOCKS AND RERUNS (CONCLUDED)
1. Reruns with mult.iple output
B. LOGICAL CHOICE
1. Skip and U instructions
Co THESrS PROBLEM
Ao REDUCING PROBLEM RUNNING TIME
1.
Minimum latency tables
B. LOGICAL CHOICE (CONCLUDED)
10 Q and T instructions
Co THESIS PROBLEM
-3-
MONDAY:
TUESDAY:
WEDNESDAY:
THURSDAY #
FRIDAY:
WEEK 3
A. REDUCING PROBLEM TIME (CON' T)
1. Minimum latency coding
Bo INTERRUPTED OPERATION
1. Qn, Tn9 9 and
9
instructions
2. Interrupted operation switch
Co THESIS PROBLEM
Ao REDUCING PROBLEM RUNNING TIME (CONCLUDE D)
R-U counters
2"
Straight line coding
Bo ARITHMETIC CIRCUITS
1. A and S instructions
Co THESIS PR OBLEM
A. CONSERVATION OF MEMffi Y SPACE
1. Overlays
2. Sub-dividing runs
B. ARITHMETIC CIRCUITS' (CON'T)
10 Overflow
2. A- and instructions
3. l2-place addition in time
Ce THESIS PROBLEM
Ao COLLATION METHOD OF SORTING
Bo ARITHMETIC CIRCUITS (CON'T)
10 Multiplication as repeated
20 P instruction
Co THESIS PROBLEM
addition
A. COLLATION METHOD OF SORTING (CONCLUDED)
B. ARITHMETIC CmCUITS (CON'T)
1. P instruction (concluded)
2. M and N instruction
30 Action of,MQC with multiplier digits not numbers
"
Co THESIS PROBLEM
-4-
MONDAY:
TUESDAY:
WEDNESDAY:
THURSDAY:
FRIDAY;
WEEK 4
A. FUNCTION TABLE SORTING
B. ARITHMETIC CmCUITS (CON'T)
1. Theory of division
2. D instruction
Co THESIS PROBLEM
Ao SERVICE ROUTINES
10 Classifications
2. Locator
30 Mark VIII
B. ARITHMETIC CIRCUITS (CON'T)
Co
Ao
B.
Co
Ao
Bo
1. D instruction (concluded)
20 D- instruction
THES IS PROBLEM
SERVICE ROUTINES (CON' T)
1. Mark VIII (con't)
ARITHMETIC CIRCUITS (CONCLUDED)
1. Periodic memory check
2. Memory clear
THESIS PROBLEM
SERVICE ROUTINES (CON'T)
1. Mark VIII (concluded)
2. AC-3
30 AC-4
INPUT-OUTPUT CIRCUITS
1. Input-output problem
20 Uniservo principles
30 Additional logical elements (thyratrons, relays,
autosyns
9
transformers, motors)
CoT HES IS PR OBLEM
Ao SERVICE ROUTINES (CON'T)
10 (line merge)
2" Herb I
Bo INPUT-OUTPUT CIRCUITS (CON'T)
10 Step-down buffering
2. 50 instruction
Co THESIS PROBLEM
",,5-
WEEK 5
MONDAY: Ae SERVICE ROUTINES (CON'T)
L. Code edit
20 Analyzer
Be INPUT-OUTPUT CIRCUITS (CON'T)
1"
5n instruction (conVt)
Co THESIS PROBLEM
TUESDAY: Ao SERVICE ROUTINES (CONvr)
I"
Auto-monitors
Bo INPUT-OUTPUT CIRCUITS (CONvT)
Ie 5n instruction (concluded)
20 Low-density recording
C., THESIS PR OBLEM
WEDNESDAY:
A"
SERVICE ROUTINES (CONCLUDED)
10 Follower
2 .... Code search
30 Word search
Bo INPUT-OUTPUT CIRCUITS (CON'T)
I"
Step-up buffering
2. In instruction
C"
THESIS PROBLEM
THURSDAY: Ao HEAVY AUXILIARIES
le"
Card-to-Tape Converter
B. CIRCUITS (CON'T)
10 In instruction (con' t)
C. THESIS PROBLEM
FRIDAY: AI) HEAVY AUXILIARIES
1. Card ... to=Tape Converter (concluded)
20 Tape-to""Card Converter
BI) INPUT-OUTPUT CIRCUITS
10 In instruction (concl uded)
20 2n instruction
Co THESIS PROBLEM
-6-
MONDAY:
TUESDAY:
WEDNESDAY:
THURSDAY:
WEEK 6
A. HEAVY AUXILIARIES (CON'T)
1. High-speed printer
Bo INPUT-OUTPUT CIRCUITS (CON'T)
1Q 30 instruction
20 30 instruction
Co THESIS PROBLEM
1. Unityping, code-edit and analysis
Ao HEAVY AUXILIARIES (CON'T)
1. High-speed printer (con't)
B. INPUT-OUTPUT CIRCUITS (CON'T)
10 4n instruction
20 6n and 8n instructions
C. THESIS PROBLEM
10 Proof-reading
2. Computer testing
A. HEAVY AUXILIARIES (CON'T)
1. High-speed printer (con't)
B. INPUT-OUTPUT CIRCUITS (CONCLUDED)
10 50 instruction
2Q 10 instruction
C. THESIS PROBLEM
1. Computer testing
Ae HEAVY AUXILIARIES (CONCLUDED)
10 High-speed printer (concluded)
Be SUPERVISCRY CONTROL OPERATIONS
1 '" SCI' FILL
2. EMPTY
3. Clear C
40 General clear
C. THESIS PROBLEM
10 Computer testing
FRIDAY: Ao
Be SUPERVISCRY CONTROL OPERATIONS (CONCLUDED)
I. Starting and stopping Univac
20 Simple error analysis
Co THESIS PROBLEM
10 Post Mortem
-7-
ANALYSIS OF COMMON CODING ERRORS
Practical problems prepared for digital computers involve upwards of several
thousand instructions. While the UNIVAC computer is self-checking, it basi-
cally detects errors in performing the indicated operation and does not
judge whether that operation, or series of operations, is correct in the
context of the complete problem. Errors may occur in these categories:
1. Improper execution of an instruction due to mechanical or elec-
tronic failure of the computer.
2. Failure of the to properly initiate the problem or in
performing rerun procedures.
3. Failure of the programmer to provide coding to carry out the
assigned task.
As mentioned in the first paragraph, errors of type 1- are checked in UNIVAC
by duplication of circuits or by redundant codes (odd-even check, etc). We
need not consider them further.
Errors of type 2 are not inherently checked by the computer, and can cause
disastrous consequences if allowed to perpetuate themselves. To a large
extent these errors may be avoided by requiring the program to check opera-
tor intervention. Thus, the rewinding of multiple output or input tapes with
8n orders insures that these tapes will be removed before those servos are
again called for further operations. Use of rings in those reels that con-
tain permanent information during a prevents the accidental destruction
of information by Supervisory Control action.
Positive checks can be incorporated in the program to insure that the correct
tapes are mounted, and in their correct order. This is the purpose of iden-
tification blocks and sequence checks. Since Supervisory Control operations
are infrequent and the operator often unfamilar with the problem and its
coding, programs should be prepared in such a way as to require little opera-
tor intervention; and when this is necessary, it should be of extreme sim-
plicity. This is especially true in setup- and rerun procedures.
Errors of type 3 are the main concern of this sectiono The first concern is
how do we recognize that a coding error exists? Assuming that a normal pro-
oedure has been followed, the prepared coding will have been desk-checked
(i.e., reviewed by one or more different coders), Unityped, and the Unityped
tape printed and checked against the original copy. Thus, we can assume that
the information on tape agrees with the coding sheets and that a preliminary
cheek has been performed.
A coding error may manifest itself in one or more of the following ways after
the program has been tried on the computer, or by none of them:
A. Overflow where it has not been anticipated (i.e., on A-, S-, X-,
or D- orders)
-1-
B. Closed loops producing a characteristic sound from the HSB speaker.
C. Adder alphabetic errors.
D. Improper sequence of tape movements observable by operator or pro-
grammer. This includes doing 3n orders when rI is empty. In orders,
when rI is filled followed by 3n, reading a rewound tape backwards,
failure to detect end of tape information, and others.
E. Improper commands causing computer to stall.
F. Output results of test data not agreeing with pre-calculated results.
One should not always expect coding errors be easily recognized, and often
the manifestation of an error may occur in a section of the problem remote from
where the error itself was made or even from the nature of the error. For ex-
ample, the following error was abstrccted from a mathematical subroutine de-
Signed to calculate log A. lOa = B.lO. Where is the error and how was it rec-
nized?
000 11 000
001 81 000
006 K 000
007 B 104
008 03 000
009 03 000

026 K 000
30 060
F 119
T 036
T 037
T 038
F 118
instructions -+ memory
a--+rL
+.03000
..... 18835 453000 ---+ r A
if a < .18835 453000 transfer control to
line 036.
SL3 (rA) = +.35453
if a < .35453 000000 transfer control to
line 037.
SL3 = + .53000 000000 --+r A
if a < .53000 000000 transfer control to
line 038
zero-+rL
E = 010000 000000 --+ rF
-2-
027 E 100
Q 031
m.s.d. of {log
if m.s.d. = 0 transfer control to line 031
.
028 B 125
! .ee OOOOOO--+rA
.9 000 SR9 (rA) = 0.00000 ooot ee
029 F 102
El = 111111 111000 ---:)orF
E 100
log lOx --+rA
030 00 000 Skip
031 5 118
032 A 125
033 K 000
034 01 000
035 C 100
036 B 112
037 B 113
038 B 114
039 H 100
040 A 101
041 B 100
U (R+ 1) Unconditional transfer of control to main
routine.
-1 000
C 125
B 100
Q 030
U 027
U 039
U 039
U 039
-3 000
C 101
08 000
-. 10000 000000 ---.r A
SRI (rA) =-.01000
+.02000
+.02000 000000 -+125, zero ---+rA

(log lOx)/lOOO
SLI (rA) = (log 10x)/100-+rA
if log 10x/100 = zero, transfer control to
line 030.
(log 10")/100
Unconditional transfer of control to line
027.
- log lOkI +
Unconditional transfer of control to line
039.
- log 10k2 + k2 ---.rA
Unconditional transfer of control to line
039.
- log 10k3 + k3
Unconditional transfer of control to line
039.
- log 10k + k --+100, rA
SR3 (rA) = -(log 10k)/1000--+rA
(b - log 10k)/1000 ---+rA
(b - log 10k)/1000-+lOl
log 10k + k --:JI> rA
SL8 (rA) = -k/lO -+rA
-3-
042 C 100
N 100 k
i
a/10
043 01 000 SL
1
(rA) = rA
H 100
rA
044 K 000 k
i
a---+rL
G 125
+003000 000000 --+ 125
045 00 000
U 010 To continue computation
The error made itself known by a second attempt to read the instruction tape.
Since this tape had been rewound with interlock, the computer stalled. How
was the coding error located? The reasoning went something like this: The
control was somehow transferred to line 000. Since this was not intentional,
it might been due to an unexpected overflow. As the problem analysis
seemed to be correct, at least as an initial assumption, the actual calcula-
tion of the log in steps 010 through 025 probably did not give the overflow.
The next pOint examined was the normalization subroutine designed to calcu-
the zeros lying to the left of the most significant digit. The sub-
routine is in lines 026 to 035. The add order in line 032 was suspect.
This order adds 1 to a counter for each zero found to the left of MSD in
log lOX. The largest number of zeros would, of course, be 11. A check show-
ed that (118) was indeed (-10 000 000 000); thus, (125) must not have
been set correctly. On line 006 is an F lL9 instruction placing the constant
003 000 000 000 in rF, and the G 125 on line 044 is to place this constant in
cell setting (ll5) to its proper initial value. The odd splitting
of the F and G was due to the original omission on the coder's part of the
resetting of 125. After recognizing the omission, he went back to make the
transfer and filled in the F-G in whatever skips were available. The error,
of course, is that the multiplication of line 042 destroys the 003 000 000 000
placed in rF, and places in its stead 31rL. Fortunately, this number caused
an overflow on line 032! The error was made on line 044 but manifested itself
by a read instruction on line 000.
Many coding errors occur because a progr.ammer has ignored or the
peculiarities of certain UNIVAC orders. This subroutine is designed to trans-
fer 30 two-word items from memory locations 301 to 360 inclusive, to memory
locations 451 to 510 inclusive. What is the error?
000 [V 301
W 45lJ
001 B 000
A 005
002 L 006
Q
004
003 C 000
U 000
004 00 000
90 000
005 000 002
000 002
006 VOO 361
WOO 511
-4-
The obvious error, of course, is that a consecutive series of 2-word items
starting at an odd memory location cannot be transferred by the V register.
In spite of the fact that the routine avoids the interchange in rV by start-
ing from an odd location and transferring to an odd location, the item in
309, 310 splits across two channels and so 309 and 300 go into 459 and 450
rather than 309, 310 into 459, 460.
The following error may be more difficult to recognize, and its variants
are among the most frequent errors in editing and mathematical routines:
Memory locations 900-939 contain 40 quantities Yi (i = 0, I, 2, , 39).
A quantity Zi = .0065i is computed and added to Yi (i.e., Yi + Zi --;.. Vi).
000 L 008
M 009
001 [A- 939
C 939 ]
002 B 001
S 010
003 C 001
K 000
004 B 009
Q
007
005 S 011
C 009
006 00 000
U 000
007 00 000
90 000
008 000 650
000 000
009 [039 000
000 000 ]
010 000 001
000 001
011 001 000
000 000
The error in this subroutine lies in reducing the instruction word A-O
9 ~ 6 00939.
It would seem to be perfectly permissible to subtract one t s from the
memory location digits successively reducing the instructions to
A-O 938 COO 938
A-O 937 COO 937
A-O 900 COO 900
But note that when the adder performs algebraic subtraction it complements
the smaller (in absolute value) and adds. Thus
-5-
A-O 939 COO 939
-00 001 000 001
A09 062' C99 062
The presence of the minus sign following the ftAtt forces complementation on
A-O 939 COO 939. A method of doing this problem involves adding the comple-
ment of 1:
A-Z 939 COO 939
000 999 000 999
A-Z 938 COl 938
The 1 ~ t t prevents a carry from adding to the minus sign. The correct program
is as follows:
000 L 008
M 009
001 A-Z 939
C 939]
002 8 001
A 010
003 C 001
K 000
004 8 009
Q
007
005 5 011
C 009
006 00 000
U 000
007 00 000
90 000
008 000 650
000 000
009 039 000
000 000]
010 O ~ ~ ' 999
000 999
011 001 000
000 000
Use of overflow for control purposes has produced great simplification in
coding and decreased the running time of routines. It has also brought a
certain headache with it - improper counting. The example shown below is
a quite common error among beginning progrannners. This fragment is from a
routine that is processing 10-word items. The item counter is increased
after transferring the new item to working storage. Unfortunately. the
counter is initially set too high,. When the last item is transferred. the
counter overflows causing the next block of 6 items to be transferred into
the current block. Item 6 is thus lost in each block.
-6-
050 [Y5 (610)

z 600 ]
051 B 050
A* 055 i
052 C 050
U Transfer control to process Ar
053 B 056 Return here if from 051
C 050
054 31 600
U
Transfer control to process Ai
055 010 010
000 000
056 Y5 610
ZOO 600
There is a simple scheme for correctly determining the initial value of the
counter. Suppose we use the above example. After having processed the last
item of block. we will transfer a useless 10-word item into working
storaye and augment the counter. which should then overflow to indicate that
a new block is to be read. The counter reading just before overflow should
then be: Y9 -660 Z 600. Initially, the counter must be Yx 610 Z 600, set to
transfer the second item into working storage. If we add once to x each time
we add 1 to the channel selection digits of the Y order, then 9-x 6-1 = 5;
therefore, x = 4 is the initial setting. This method can be easily extended
to any size item, for the case of 2-word items V99 660 W 600 is the reading
after proeessing the 30th or last item. If we use 002 002 000 000 85 the
augmenting constant, 99-x =58 and x = 41. The initial setting belng V4l
602 W 600.
What is the possible indication of the coding error in the following routine?
It is supposed to count the number of nonsignificant zeros of the number X.
(Nonsignificant zeros would be those to the left of the 6 in the example:
000 006 350 010 except the sign).
000 K 000
K 000
001 F 008
E 009
002 00 000
Q 004
003 00 000
U
004 B 010
A 011
005 C 010
B 008
006 -1 000
C 008
007 00 000
U 001
008 [010 000
000 000]
-7-
009 Quantity X
010 [000 000
011 000 000
000 000]
000 001
In this example the extractor is shifted right to examine each column success-
ively. If the quantity X is zero, however, the routine will never transfer out
and becomes a closed loop.
There is yet another class of errors belonging to section 3. These are errors
Qf program logic and can be detected by an analysis of the flow chart. The
illustration is that of a two-way merge with multiple data reels for each in-
put. The standard sentinel convention is assumed for each input. The two
standby block input procedure is used.
Get New A
Block
x. A
to process Al
t - - ~ end of A input fi Ie rout ine
to process Al
-8-
The error in this example is in the two steps following the servo switch symbol.
As two sentinel blocks are at the end of each input tape, rI must have contained
the second sentinel block and thus not a valid set of A items. The correct ver-
sion would be:
$--1 H Ta--+rIH rI--+A process
STUDENT EXERCISES.
Find the coding or logical errors in the following problems:
This is intended to be an ending routine that will read into the memory the
instructions for the next run. Assume the reading head on tape #1 is in the
correct position and fI is erased.
049 ......
. .....
050 [B 940
L 056 ]
051 00 000
Q
053 ending test
052 00 000
U XXX continue witb Run 1
053 11 000
30 000
054 50 055
U 000 go to Run 2
055 Begin
Run 2
056 ZZZZZl
ZZlZZZ Sentinels
057 ......
. .....
This is intended to be an input routine that will park the current 2-word
input item in a working storage and read in a new input block when the
current block is exhausted. It is to be entered from the main routine by
R 305 U 300. Assume generalized overflow control with an increment of
000000 000001. Also rI contains the next input block from tape 2.
300 V60 925
W 925
301 B 300
A* 306
302 C 300
U 305
-9-
303 32 925
B 307
304 C 300
00 000
305 anput
return]
306 002 002
000000
307 V60 925
W 925
This output routine i. supposed to transfer a lO-word item from an output
working storage to the current position in the output block; then write the
output block on tape after it is filled with 6 items. Assume generalized
overflow control with an increment of 000000 00000"1. Output operation is
to be accomplished in the main routine by R 430 U 426. Working storage is
600 609.
425 000000
010 010
426 B 430
A 425
427 00 000
U 429
428 72 940
B 432
429 H 430
00 000
430
Ii'
600
Z40 940]
431 (9utput
return]
.
432
y
600
Z40 940 Reset constant.
-10-
A SURVEY OF SPECIAL INPUT-OUTPUT TECHNIQUES
1. Additional methods for of !hi input buffer.
Chapter 10, Section 8, of the Univac Programming Manual describes a simple yet
completely general method of keeping rI filled for multiple input tape prob-
lems. In this method each input tape requires the reserving of two blocks
within the memory and the execution of six Y-Z transfers to put the standby"
block into working storage position. This paper assumes the reader is com-
pletely familiar with this method as described in the which is here-
inafter referred to as the "two-standby" technique.
Because of its Simplicity and complete generalitY8 the two-standby technique
was described in the elementary programming manual. However, there are other
techniques, Preselector, Reversal, and Automatic method which' are of greater
efficiency and in more general use
o
The Preselector Method
This method is Simple in concept and extremely conservative of memory requIre-
ments and time o It is the preferred method for all problems where the selec-
tion of the next item for 'processing is based solely upon the relative magni-
tudes of the sequencing fields within the items. The principle will be illus-
trated by the following
Suppose we have two files of input items to a processing run. Let us call the
files A and B. Suppose further, that these are ten-word items, six to a block:
that the items within each file are arranged in ascending sequence by a key
within each item and, that the processing is always done on the item
from either file which has the smallest (in magnitude) key_
At the start of our problem we will have the first A and B block in the memory
and are concerned with what tape to read into rIo Let us write down the keys
of the twelve items q mighl
A Item Keys B Item Keys
1136 1000
1137 1010
2100 1011
2501 2050
2502 2161
2600 2163
In comparing the keys of the last A and B item of the block we note that the
A item has 8 smaller key than Bo Since the smallest item of A or B is processed
first by our routine, it is evident that A6 will be used before B6- But since
the A and B items are in ascending sequence within each file, we then know that
the A block will be exhausted before B; therefore
o
we should order the next A
block now
o
That is
o
fill rI with Ao We can then begin processing A2, ---,
B
l
, 82,--- with the full confidence that when A6 has been processed the new A
block is in rIo If B6 had been smallest we w,ould fi II rI wi th B items II Of
course, after rI is transferred to the appropriate empty block, the last items
must be compared again to determine what now should fill rIo
The method is simplicity itself as shown in the following flow chart:
Process Bj
l ~ j
Process Ai
i + l ~ i
l ~ i
The reader will note in the flow chart that if the keys of the current A and
B items are identical, an A item is selected for processing first. Then, if
A6and D6 have like keys, and A block must be read into rI since it is the A
block which is exhausted first. This can be a subtle point in some routines
and care should be taken in the treatment of the case of equal keys.
Since no standby blocks nor block transfer instructions are required, the
preselector is the best method for filling rI,when it is applicable. There
is
o
however, a class of problems with complex processing rules that do not
admit the preselector method in these problems because the order of reading .
tape is either not dependent solely on the original contents of the tapes, or
the order of processing the items cannot be easily determined as simply the
smallest (or largest) item key first. In these cases one of the standby
block methods, which are, completely general, must be used.
-2-
The Automatic Method EQr Two-way Input
A study of the two-standby block method previously mentioned will make evident
to the reader that one of the two standby blocks is always emptyo
The Automatic Method uses only one standby block
o
the other block in rI is con-
sidered as the second. This method is thus somewhat more economical of memory
space than the two-standby block methode The following notation is used in the
flow chart and coding:
Ta an input file on UNISERVO a
A a block from tape a
rI register I
X designation of a standby block
Ta .... rI a block from tape a is read into rI
Initially, the first block from Ta is read into A and the first block of Tb is
read into B. The standby block X is filled with the second block from Til and
rI is fi lIed wi th the second block from 1b" Variable connectors a 2 and ~ l are
set.
Get new
,A Block
Get new
B Block'
rI-4A
-3-
To process
new A Block
In coding the
automatic method the input
blocks assigned were:
A -- 940 - 999
B -- 880 - 939
X -- 820 - 879
If a new A block is needed we go to
line
100, and for a new B block to line
109.
100
[y
820
940J Z
Variable connector a ( a 2 shown)
101
y
830
Z 950
102 Y 840
Z 960
103 Y 850
x-+A
'Z 970
104 Y 860
Z 980
105 Y 870
Z 990
106 3a 820
T ---+rI
B 118
J .al
a
107 C 100
B 121
} 2
108 C 109
(i)
U
109 [3b
880
0
J
u
Variable connector
shown)
110
y
830
Z 890
III Y
840
Z 900
112 Y
850

Z 910
113
y
860
Z 920
114 Y 870
Z 930
3b 820
T
b
--,) rI
B 119
.a2
116 C 100
B 120

117 C 109
CD
U
118 3a 940
CD
U
119
y
820
Z 940
-4-
120 3b 880
121 Y 820
U0
Z 880
The automatic method requires 22 words for; instruct ions and constants, pI us
180 words for the At B, and X input blocks. The obtaining of the next A or B
block will require either 3785 J.JiS if the block is in rI
f
or 13,365 Jls if the
block is in Xo If we assume a completely random reading sequence, then the
average will be 8575 Compare this to the two-standby block method de-
scribed in Chapter 10, Section 8, UNIVAC PROGRAMMING MANUAL which requires 25
words of instructions and constants plus 240 words for the A and B blocks and
their standbyso The time to obtain the next block is 15,395 J.ls. Obviously,
the automatic method is superior to the method described in the programming
manual.
Reversal Method for Two-way Input
The reversal method is feasible for those cases where the volumes of the two
inputs are known to be very dissimilar. Assume the A input on tape a is very
much larger than the B input on tape b
o
The initial setup is to fill an A
block and a B block with the first blocks from their respective tapes, placing
the second A block in rI. The following flow ehart indicates the method of ob-
taining the next A and B blocks. (The symbol is a shorthand notation
for the statement 1tfill rI with a block from tape at reading this tape back-
wards".)
Get new
A Block
Get new
B Block



.
-5-
To
new
rI---+ B
In coding the reversal method
Q
the A block is assumed to be in 940-999 the B
block in 880-939& If a new A block is needed, we go to line 100; and if a new
B block is needed, we go to line 101
0
100 3a 940
rI--:lo Ao Ta
U
To process"A block
101 4a 880

3b 880
rI-70Bo
102 3a 880
Ta-+rI
U
.,..-
To process B block
The coding is small; 3 and an additional 120 words for the A and B block
storage. The time to get the next A block is 3780 while the time to get the
next B block is 10 53 seconds due to the tape reversals necessary to reposition
Tao A simple calculation shows that if the ratio of the size of the two files
is : 1300 the reversal method is faster than the two-standby block method and
requires less space than the automatic
o
Although the automatic method is always
faster than the reversal for a very sm'all B file" the few extra seconds required
by the reversal method may be tolerable to gain the extra 79 memory cells o
The Automatic Method for
This method is an extension of the automatic method described for two-way input.
and uses two standby blocks in the memory with rI considered as the thirdo The
following brief annalysis will show that the information in the standby blocks
varies throughout the process.
Assume that the A, B, and C input blocks are filled from tapes a,
and ceo respectivelYq and that the two standby blocks labeled X and Yare filled
with the second blocks from tapes a and bl} respectivelyo The second C block is
in
A Input B Input
BQ
x Standby

A
Y Standby
D
C Input
k
c
Suppose the A input block is exhausted first: (A shaded block indicates the
portion of that block processed)
n
t=J
x
rI
tJ

'-
C
y
[j
the contents of X are A so they are transferred to the now empty A
input block
o
rI is transferred to X so as to be able to order the next A
block from tape a.
X
G
Q
rI
--7-
If we again exhaust the A block, we can replace it directly from rI without
disturbing the two standby blocks
o
If, however! the B block is exhausted
next, we fill B from the Y standby, transfer rI to Y! and order the third
B block from tape:

.. -
C-=;
x
[J
rI
[J.
Thus, in this example, the standby blocks which start with A and B items soon
are changed to C and A items
o
Unlike the simple procedure described in the
Programming Manual where each standby block always contains only one type of
item. this method permits "floating" information in the,standbyso The attached
flow chart indicates the general solution for this methodo At the beginning,
connectors a 3. 5, Yl, are set for the initial conditions already described.
Let us investigate the logic behind the flowchart by first making a table.
showing the possible configurations of Xe and
Configuration
Contents
Number
rI X Y
1 A B C
2 A C B
3 B A C
4 B C A
5 C B A
6 C A B
Now suppose an A block is needed; this block will always be available in either
rI, X, or Y. After placing this A block in the A input position, we might ask
ourselves: What is the new configuration? The following table shows the
figuration numbers resulting from the act of obtaining an A block, B block, or
C block from each of the six possible
-8-
-----I'::)
To pro-
cess new
A Block i
To pro-
n e ~ l o ~ ~ W
To pro-
cess new
C Block
0'

. ,, __ "" ......... .. ,..... d, "" ........ " ...... v ....
Configuration Number Transformation Table
Original Configuration after obtaining an:
A Block B Block C Block
0; I
1 1 3 S
2 2 4 6
3 1 3 6
4 2 4 " S
S 1 4 S
6 2 3 6
--
Note that obtaining an A block from configurations 1 or 2 leaves these configu-
rations unaltered while configurations 3
Q
and 6 are altered by obtaining
an A block
o
By alteration is meant a change in position of either an At B, or
C block
o
The connector aI, in the flow chart suffices -to obtain the next A
block when either configuration '1 or 2 exists o a
2
and a3 are the connectors
used where an A block is in X: a
2
for configuration 3, a3 for configuration 6.
Separate connectors are needed even though the A block is in X for each con-
figuration since the act of moving the A block alters the position of the Band
C blocks in different ways 0 a 4 and as are connectors used when is
in Y: U
4
for configuration 4 and as for So A similar ass'ignment of connectOrs
to configurations applies for obtaining Band C These are summariized
in the following table:
Configuration
Connectors
Number A B C

1
al

')'4
2
0:1
Y2
3 (12
Ys
4
a
4
1 Y3
5
as
3
Y
1
6 a
3
S
Y
l
-10-
Now that we have assigned the variable connector numbers for each configuration,
we note in the transformation table that obtaining an A block from configuration
6, for instance, will produce configuration 20 Now configuration 6 requires that
ex 3. 5- and YI be and the resulting configuration 20 will require connec-
tors a It' and Y 2 to be seto Thus, the connector ex 3 must set 4 and Y2
as well as ClIO This kind of reasoning makes evident the rationale behind the
flow charto
The average time required to obtain an A, or C block is 11,622 387 words
of storage are required for the three input blocks, the two standby blocks, and
the coding and constants
o
Summary
While it is evident from the foregoing discussion that the so-called preselection
method of keeping rI filled is superior to the standby block schemes in both
time and space requirements! there is occasionally a class of tape problems en-
countered for which preselection is not suitable
o
For these problems, one of the
standby block methods is preferable
o
For two-way input the automatic method is
preferred, the reversal method being used when the second input is of very small
volume and memory space is at a premium
o
For three-way input the automatic method is preferred over the two-standby block
method (Chapter 10. Sec. 8) 0 For more than three-way input wi th no great' dis-
parity in the relative volumes of the inputs an extension of the two-standby
block method is superior in space and time to similar extensions in the methods
described above"
20 Arrangement of information for e'ffieient t'ransfer wi th in the computer ..
As a direct consequence of reading information in units of a block rather than
an it is necessary to consider the most efficient arrangement of the items
within the block. Where the item size is one, two, or ten words in the
usual sequential arrangement of an entire item followed by the next one in suc-
cession clearly does not offer much room for improvement 0 In the case of one-
or two-word items, minimum latency considerations might require that the order
of items in the block be rearranged from the normal one for a particular
but such a situation would be unusual o
The above statements follow directly from the existence of two-, and ten-
word registers information from an input block to an output
block or to working',storage as the case may beG
In dealing with twenty- and thirty-w9rd items, or four- and six-word items
o
the
same remarks apply with several transfers taking the place of the one required
previously@ There are cases
o
however
o
where the normal sequence of items can
be vastly improved upono Consider the twelve-word item for example o In general,
to transfer a twelve-word item from an arbitrary position in the input block to
-11 ..
working storage or to an arbitrary in the output block requires six
V-W instruction pairs o
If this is accomplished with iterative coding, the following routine will have
to be executed six times for each item transferred:
100 V40200
101 B 100
102 C 100
103 00 000
104 010002
W 300
A* 104
U 100
U XXX to processing routine
000002
Straight line coding would offer some improvement in speed but would require 35
lines of coding per input to transfer the items to working storage and 35 lines
of coding to transfer from working storage to the output block o
The following arrangement of the items would be far more efficient,
both with regard to memory space and execution time: (j, k is word k of item j)
Cell
200 1, 1 11/ 2 1,3 1,4 1,5
l1l6 10 7 1,8 1,9 1,10
210 2,1 20
2 2
11
3
29 4 2,5 2,6 2,7 2,8 2,9 2,10
220 3,1 3,2 3
11
3 3,,4 3,,5 3,6 3,7 3
0
8 3,9 3,10
230 4,1 4
e
2 4,3 4,4 4,5 4,6 4,7 4,8 4,9 4,10
240 5,1 5,2 5,3 5,4 5(f5 5,6 5,7 5,8 5,10
250 19 11 1,12 2.,11 2,12 3
11
11 31)12 4,11 4,12 5,11 5,12
For this configuration
a
an item transfer would require only a Y-Z and a V-W.
Using straight line coding 19 lines per input will"suffice-to transfer the
item to working storage and similarly 19 lines for the output block o The
coding necessary to accomplish the transfer of items to working storage is
illustrated below.
100 00 000
U 101
101 R 100
U 110
102 R 100
U 113
103 R 100
U 116
104 B 108
C 100
105 V 258
y
240
106 W' 250
Z' 2.00
107 00 000
U XXX to processing routine
-12-
108 R 100
U 109 constant
..
109 30 200
U XXX to preselector read
110 V 252
y
210
III W 250
Z 200
112 00 000
U XXX to processing routine
113 V 254
y
220
114 W 250
Z 200
115 00 000
U XXX to processing routine
116 V 256
y
230
117 W 250
Z 200
118 00 000
U XXX to processing routine

Use of minimum latency coding could further improve this routine. The following
examples will illustrate further applications of this technique ..
15-Word Item
1,1 1,2 1,3 1,4 1,5 1,6 1,7 1,8 1,9 1,10
2,1 2,2 2,3 2,4 2,5 2,6 2,7 2,8 2,9 2,10
3,1 3,2 3,3 3,4 3f)5 3,6 3.,7 3,8 3
e
9 3,10
4,1 4,2 4,3 4,4 4,,5 4,6 4,7 4,8 4,9 4,10
I,ll 1,12 1,13 1,14 2,11 2,12
20 13 2\1 14 3
11
11 3,12
3,13 3,14 4,11 4,12 4
11
13 4,14
10 15 2
11
15 3
0
15 4,15
-13-
5-Word Item
1,1 1,2 1,3 1,4 2,1 2,2 2,3 2,4 3,1 3,2
3,3 3,4 4,1 4,2 4,3 4,4 5,1 5,2 5,3 5,4
8,3 8,4 9,1 9,2 9,3 9,4 10,1 10
0
2 10,3 10,4
11,1 11,2 11,3 11,4 12,1 12,2 12,3 12,4 1,5 2,5
3-Word Item
1,1 1,2 2,1 2,2 3,1 3,2 4,1 4,2 5,1 5,2
11,1 11,2 12,1 12,2 13,1 13,2 1401 14,2 15,1 15,2
16,1 16,2 17,1 17,2 18,1 18,2 19,1 19,2 20,1 20,2
11,3 12,3 13,3 14,3 15
i
3 16,3 17,3 18
0
3 19,3 20,3
Thus we see that a IS-word item can always be transferred using one Y-Z, two
V-W's and one B-C. Similarlyw a five-word item can be transferred by using
two V_WIg and a B-C.
It is to be noted that input data from in cannot be
obtained in the efficient configurations illustrated above. Therefore, the
rearrangement of the information will have to be executed by the first Univac
run on the data.
Item sizes multiple sub-multiple of 60
Occasionally, an item size which cannot be contained an intergral number of
times within a block is desirable. Generally, this would be true where the
processing is done at less than tape time. Then, any reduction of the total
amount of tape required will reduce the overall processing time.
An obvious solution to this problem is to consider the item as being composed
of a series of sub-itemsuthe length of the sub-items being a sub-multiple of
a block. In the practical case the sub-item will be ten words in length as the
time required to move many one- or two-word sub-items negates any advantage
gained in reducing the amount of tape. For example, if the number of digits
required in an item is 420 (35 words) an item size of 40 words could be used.
-14-
Each item is then composed of four ten-word sUb-items. If it were not possible
to handle items multiples or sub-multiples of a block, the item size would
have to be 60 words in length. The item thus saves 1/3 of the tape re-
quired and reduces the processing time up to 1/3 depending on how much the item
processing time is under the tape time:
The following flow chart is an iterative solution to the selection of an N-word
item, W, composed of Pksub-itemsAio Each input block, At containing i' such
sub-items. The k in W is shown as a superscript since W , W
k
are, in
fields of the item Wo
Get New
VI Item
Note: initially i = 0
=
-15-
i +
k+ l----+k
To Process
New W Item
The symbol
represents an appropriate subroutine which will obtain the next A block. The
reader should note that while it i ~ not possible to use all of the input block
A for the working storage W, it is possible to use a portion of the block for
some item sizes and thereby conserving memory space. For example, consider
the possible configurations of blocks containing a forty-word item:
Block 1 Block 2 Block 3
1-1 2-3 4-1
1-----.---
1-2 2-4 4-2
1-3 3-1 4-3
1-4 3-2 '4-4
2 ... 1 3-3 5-1
2-2 3-4
5-2
(The notation j-k means the kth sub-item of the item j). Thus, only two block
configurations are possible. Now, if the working storage WI, W2, W
3
, and W4
is positioned as follows with respect to the input block AI, A2, A3, A4, A5
and A6:
working storage
input block
Only twenty extra positions are necessary for the working storage W.
Although it would be desirable to have the read routine be the preselector, it
has not been found possible! up to this time, to suitably modify this technique
to make it work properly for the ttodd
1t
item sizes. This means that one of the
standby block methods described earlier must be employed in ordering blocks
from tape o
-16-
Item sizes ]&! of fixed length
o
In some large volume tape problems the variation in mInlmum required digits for
the items of a file may be very largeo This case may occur when a master infor-
mation file is laid outo If we allot to each 'item of the file the tape space
required for the largest possible item that can we may find that a very
large portion of the file contains areas causing inefficient use of the
tape and could greatly increase the processing time o
Hence. it is desirable to let each item use as little tape space as is feasible o
If this is done, then the items on the tape may be of varying lengths o This
variable item size can be easily handled by the computer by considering each
item to be composed of a variable number of sub-items. each sub-item being of
a fixed size
o
All that is then necessary is to read into the memory all of the
sub-items comprising an item
o
The routine that selects the next sub-item from
the block (or the next block from tape) is designed in the usual manner since
the sub-item is of fixed size
o
The problem next to be solved is how do we know when all of the sub=items for
a given item have been assembled? This may be done in several ways, the one
to be selected depending on the particular situationo
The first method to be described assumes that each sub-item carries the item
key word e Then, as each sub-item is selected its key word is checked for
equality with the current key word
a
The flow chart for this method is shown
below. A is the input block containing i' Ai e W is the working
storage which will contain the current tape a item; this item is composed of
fields Wko K stores the key word for the current input itemo K is initially
set to the value of the key word of the first input itemo The key word of
each sub-item is assumed to be field Ayo
Get Next
W Item
k +

-17-
To Process
New W
This method assumes that two items with the same key word is not possible. It
is also inherently assumed that the space required for repeating the key word
is not unduly excessive
o
Perhaps an inherently better method is to consider the sub-items of an item to
be of two kinds" The first sub-item may be thought of as a t@header
tl
sul) ... item
which contains the key word for the item" The remaining can be con-
sidered as "trailer" sub-items
o
Nowo instead of requiring all sub-items to
carry the item key word, we can simply require that each sub-item have a fixed
digit position which will contain
ll
sayv the symbolttH
t
if the sub-item is a
header and ffr
t
if the sub-item is a trailer. The flowchart for selecting the
N-word item W composed of a variable number of sub-items Aie with key as Ai.
Get Next
W Item
k +
-18-

=

To Process
New W
- . .. --
IDENTIFICATION BLOCKS AND RERUNS
1. Introductory Conside!2tions
The governing characteristic of commercial data processing problems is the
very large amounts of information (records or items) to be fed into the com-
puter. Even some of the simpler problems may involve tape files of 20 to 25
reels per file, while large problems can demand nearly 300 reels per file.
Most often the processing of these tape files requires that the items com-
prising the file be processed in a definite sequential manner. If this is
true, then certain precautions must be exercised to insure that these items
processed in sequence. Two teChniques in common use to guarantee this
sequential processing are identification blocks and block counts.
The reader should bear in mind the situation encountered by an operator
during the running of a large tape problem involving hundred reels.
Since the contents of a reel of tape are not directly visible to the oper--
ator, a gummed label is usually attached to each reel. Now, although each
reel has a visible identification label, what assurance have we that when
the computer calls for the fifth reel of file A to be mounted on a Uniservo
that the operator does indeed mount reel #5 of file A on that Uniservo?
Or, granting that a reel bearing a visible label "Reel #5 - File Att is
mounted on the desired Uniservo, what guarantee is there that this label is
a correct description of the contents of the reel? These questions arise
because of the bias of hmnan operations toward Since the computer
is self-checked in its operation, it is desirable that human intervention
in the proceSSing chain be limited as much as possible and that such inter-
vention be done in a Simple fashion, subject to check.
Assuming, now, the correct data reel has been mounted on the desired Uniservo,
can we rest assured that the jtems on that reel will be processed in their
correct sequence by the computer? We can provided, again, that no operator
intervention is aecessary. Occasionally, it is necessary for the operator
to affect the position of the tapes mounted on the Uniservos; for example,
if a block on Uniservo #2 was read into rI incorrectlYf the operator will
set into SR from SCP instructions causing the block to be re-read.* If the
number of data blocks recorded on the tapes is placed on the tape itself,
then the computer can check that the operator did not misposition the tape.
With this check made possible, a block of items cannot get lost in the pro-
ceSSing run, nor oan a block be prooessed twioe,
To illustrate the use of identification blocks and block counts in checking
the human interventions mentioned above, a general two-way merge for ten-
word items will be described. First however, we shall describe the stand-
ard data tape format.
* This condition will not often arise if the computer is equipped with the
automatic re-read device.
..1-
2. The Standard Data
The first block recorded on a data tape will be the identification block of
which the first word only will be of interest to us now.. This word is the
reel identification number, designating the file of which this reel is a part,
and the sequence number of the reel within the file. For our two-way merge
the identification word has the following appearance (for the 13th reel of
the file A):
Following this ID block are the data blocks. If the reel is not the last
reel of the file, there may be up to 1981 tull data blociti. Following tbe 1a.t
data block there will be recorded a sentinel block having the following
composition: word 00 (the first word of the block) consists of twelve Z's
(the sentinel); word 50 also consists of twelve Z's; word 51 contains the
number of data blocks recorded on tape plus this sentinel block. For ex-
ample, if this tape contains 1981 data blocks, word 51 of the sentinel
block looks like:
0000000001982
Word 59 (the last word of the block) will not consist of twelve P's.
Following this block will be other blocks which are provided for the pos-
sibility of the reruns to be described later. If the reel is the last reel
of a file, its identification block is in the same format noted. However,
being the last reel of a file. it may have less than 1981 full data blocks
recorded on it. Following the ID block WIll be as many data blocks as
needed ( 1981) to record the remaining items. If the items on the last
reel are a multiple of six, they will exactly fill the last data block. In
this case, the sentinel block described will follow the last data
block, except that word 59 of the sentinel block consists of twelve "s
which serve to identify the last reel of a file. But if the data items are
not a multiple of six in number, the last data block will be only partial-
ly filled. In this case, a Z sentinel will be placed in the next word
position of the partial data block, following the last word of the last
data item. For example, if the partial block contains only three ten-
word items, word 30 will contain all Z's. In addition, the 50th word of
the block will also be a Z sentinel; amd word 51 will contain the number
of full data blocks on this tape, plus the partial data block. Word 59
of this block will consist of twelve "s. In either case, following the
partial or full sentinel block, is a second full sentinel block.
3. Description of the Generalized Two-Way Merge
Briefly, the generalized two-way merge problem is this:
Other information in the ID block is for use in re-runs.
There are two multi-reel files, labeled for convenience, file A and B.. Each
file consists of a series of ten-word items. the first word of each item be-
ing a key word or serial number. The items iT! each file are recorded on tape
in an ascending sequence by their key words. The reels comprising each file
are recorded in the standard data tape format just described. The problem is
to produce a new file, labeled C, also in standard data tape format, which
contains every item on the A and B f i l e s ~ arranged in ascending sequence by
their key words. Two Uniservos will be allotted for each file:
Unisel"v..Q!,
A
B
C
In this way, after reel I of file At which is mounted on Uniservo 2, has been
processed, it is given a rewind instruction. When this tape is rewound, the
operator will replace it with reel 3 of file A. While reel I is being rewound,
reel 2( which has been mounted on servo 3, is processed by the computer; thus,
the computer need not wait for the rewinding of a reel and the mounting of a
new reel of tape. This same procedure is used for the Band C files as well.
The complete flow chart and coding of this problem is attached.
The following list is an explanation of the flow chart symbols.
Ta =
A
=
Ai =
AU
==
1
la =
lla
=
--.. - - ~
=====*,
)
Tape on Uniservo a
A block of 6 items from Ts
ith item (i::: 1.2, .... ,6) of A
nth word (n = 0,1,2, ,9) of Ai
File and reel label for Ta
Block counter for Ta
A block from Ta is read into rI (forward read)
Rewind Ta with interlock set
The block C is written on T
c
Servo interchange symbol, a, which was Uniservo
2, now becomes 3 .. Circle rotates 180
0
so that
next trip through a
v
which was servo 3, becomes
2 again.
Flow line followed by individual items
Flow line followed for end (or start) of block
Flow line followed for end (or start) of tape
Flow line followed for end (or start) of file
-3-
A brief description of the flow chart follows:
At the start of the problem the flag indicates that the input and output Block
counters are set to zero, and the input and output tape labels are set to one.
At connector CI) the input file labels with their appropriateUniservo numbers
are printed so that the operator can mount the first reel of the A file on the
Uniservo 2, the second reel on 3, the first reel of the B file on Uniservo 4,
the second on 5, and blanks on Uniservos 6 and 1. xt step, l2iJ, is to
write the identification block on the output tape. At 2 is irans-
fer breakpoint 10 If breakpoint 1 has been depressed, e computer will stop

point. If the operator then forces transfer., the computer will go to

6 for the re-run procedure. The normal path through this breakpoint is to@.
were the tape label for tape b is tested. If the wrong Tb has been mounted,
the tape will be rewound with interlock, the correct label printed on Sept and
the computer will stall on the next read until the new tape is mounted. When
Tb is correct (or the first data block is read into the memory and
the computer goes to to test the Ta label. When the first data block of
Ta and Tb has been reJUl. to @ where the prese lector method
is used to fill rIo the merge proper. As each output
block C is filled, the computer goes to where the output item counter is
reset, the block written on Tc. and the output bloclc counter increased. When
the data block has been written on the tape, a Z sentinel block is written
and (jV set to indicate the contents of rIo The output tape label, Le, is
printed on SCP to enable the operator to label the current output tape correct-
ly. Lc is then augmented for the next output reel, the output block counter,
is reset, and the output tape ,rewound with interlock. The output servos
are then switched and the augmented tape label is written on the new output
tape. The output operations concerned with re-runs will be discussed later.
As each A block is exhausted, the computer goes to @ where the A input item
counter is reset and the next A block obtained from rIo The input block
counter, at is augmented and the new input block is examined for a sentinel
in the key word pOSition of the last item of the block. Should this be a sen-
tinel block, the computer data block count, at is tested against block
count on the tape. If they do not agree, the next operation is at where
all tapes are rewound and the error noted on SCP. If the counts are correct,
Ta is rewound wi th interlock' and a test is made t'o determine whether the last
Ta has been processed. If it has not, the A input block counter the
Ta input servos interchanged, and the Ta tape label augmented. new Ta
label is examined as noted before. If this was the last A tape
v
is set to
send us later into the ending routine at @. A simi lar procedure holds for
exhausting a B input block.
4. Re-run Procedure
Next, let us examine ihe procedure planned for a possible re-run. A procedure
must be prepared by the programmer for the continuation of computation on .his
problem in the event of a computer breakdown. It is always preferable to have
a general re-run procedure which will be used for aD1 re-run to avoid any pos-
sibility of errors on the part of the operator. ThuI, the re-run (to be ex-
plained) is designed for even the extreme cases where nothing in the computer
can be used.
-4-
The theory of this re-run method is quite simple: We Uphotograph" the memory
and register setup at the conclusion of each output tape. Thus, in the event
of a computer breakdown while processing output tape k we need only restore
the memory and reposition the tapes as they existed at the beginning of out-
put tape k. At the end of output after we have written the sentin-
el block, we var .. iable connector in accordance with the contents of
rI. Then, at 20 we augment the outpu tape label, L
c
, and reset the output
block counter, c. At this point, we write the entire contents of the mem-
ory onto the output tape and rewind it.'
Let us now suppose a computer breakdown or an operator error occurs, or even
in some succeeding one of the output tapes of this proves unreadable.
In all of these cases, we need only locate the memory dump recorded on the
previous output tape. By reading this memory dump back into the computer, we
will restore the memory to the state which existed at the beginning of the
output tape which later proved unreadable, or during the writing of which the
computer or operator error occurred. The input tape labels and block counters
are used to reposition the input tapes. Hence, we have reconstructed the
problem at the beginning of an output tape.
If a re-run is to be done, breakpoint 1 is depressed and the last good output
tape is read down past the sentinel blocks (a search routine is placed jn the
output tape 10 block for this purpose).* The memory dumped on this output
tape is then read into the computer and control transferred to the start of
the merge routine. After writing the output tape 10 block, stops
on Ql where we release the breakpoint and force transfer. At the input
tape label for Ta is and this tape in accordance with the
reading of a obtained from the memory dump. the Tb label is checked
and this tape positioned. At rI is filled from the appropriate tape,
control then being to begin the reprocessing of this new
output tape.
* As indicated in the coding for this example, the operator will mount on
servo 1 the last completed output tape. This tape is easily identified by
the Lc last printed on SCP. He then does Initial Read on this tape. The
instructions in cell 000 which set up in SR are ignored through manual
clearing of SR. The computer then executes the instructions in cell 001
of the 10 block which locates the memory dump at the end of this tape.
-5-

a=2,b=4,c=6
La=Lb=Lc=l


0-0G-0

NORMAL
G)-{;J-) 8fi10
I
..0
I
01 La-+LKa-+ HI b-7
t

0
rI-+C

@-{0
.

I
I
L --{0
l-G
d

I
f-
I
G H B
I I
0r[
--l tf-----
I
15 :.

r .

"Block Count Off- l 1

RWD
*T ,Tb,T. 1
II SCP I a . i __ .----J . ..


I
co
I
k
II
If U
Ii @ II
-,;
l!:.=- --
II
It

@-0

-G
I
cr-

5. Coding for Generalized Merge
000 R2W 007
UMG 005 transfer control to generalized overflow
_. ____ IJAIM'I_
.......,....,---_. __ ._' ____ 1-
001 000 000
000 001
________ .. _""._M_"' __ '_ ........ A .. _ ..."'_ ...__ ................. ____ .. ,_""j .._OW_____________
002 11 000
31 060
003 31 120 ins t ruc"t ions 000-239
30 180
004 81 000
U 008
------_ ........-----,"'._---
005 B 007
A-
001
006 C 007 generalized overflow subroutine
00 000
OO? [ ece eee
/
eee eeeJ
-----_ ....--".. .. _--_--. ..... -
__ wo__
(0008
F 193
B 189
009 E 194
F 195
010 E 183
e 820
011 B 190
E 185
012 F 193
E 194
013 C 821
F 082
014 B 196
E 184
015 H 822
E 186
016 e 823
50 197 merge
017 50 198 JfMOUNT TAPES
50 820 Jtt-:. Ll
018 50 822
ALT W/xJII SCP
50 821 R M Bxxx x
019 50 823
W/xJJJ
50 199 Jt M ON
-10-
020 50 217
00 000
1. JtJtJt
021 R 142
h..l
U 141 transfer control to connector 21
0.
22 L 000
01 156 force transfer for rerun
023 R 026 .al
U 098 transfer control to connector 16
024 R 026 .a2
0
025
U 065 transfer control to connector 18
B 218
e 026 .a3
0-0
026 reee eee
eecJ
027
eee
00 000
T 029 transfer control if AR >Bg
028 [12 000

U 030] transfer control to connector 11
029 [1-4 000

00 000]
0)030
B 880
11 L 940
031 00 000
T 073 transfer control
if A9 > B9
-
(;)032
[y
880
Ai --+ ws
00 000]
033 R 108

U 106 transfer control to connector 12
834
[eee eee
}i+l---+i
02 eee eee]
(overflow if i = 6)
035 B 034
A* 219
036 e 034
U 030 transfer control to connector 11
037 B 220
10
38
C 034
30 880

B 191
039 AOI 19S
JAa
ell 191
It-
040 B 930
L 230
041 00 000
Q 043 transfer control
if Ag = Z
042 00 000
U11 026 transfer control to connector 5
043 B 931
L 191
044 00 000
Q 054 transfer control if A ~
=
Ila
G
045 50 221
JlBLOCK Ii COUNT }
50 222
Il OFF-RERUNJUl ~ SCP
046 F 157
B 050
047 E 183
H 050
048 E 185
H 051
049 E 187
e 052
050 [00
000
000
O:j
} -rld with ilrterlock Ta'
051 ece eee
cee cee T
b
, T
052 eee eee
c
eee eee
053 90 000 stop
00 000
054 [82 000 rewind with interlock Ta
8 939J
055 L 223
if Al=
Q 144 transfer control
P
056 K 000
e 191 O - + ~ a
057 F 183
B 184
058 G 184
H 183
059 F 224
E 028 interchange Ta uniservos
-12-
060 H 028
E 054
061 H 054
E 065
062 H 065
E 068
063 e 068
B 189
064 A- 225 La +
e 189
-.---' .. --...--.--..--A .. - .... -----.--.-----r---,.----
0065 [12
000

18 32 880J
At rI
066 B 880
L 189
067 00 000
Q 037
t:ransfer control
Of A
O
-
1 1 - La
_ .. -..._ ...... _ ..._ ...._..,_ ....._"".-.........N ......-......___ ..... ________ __ ....._w .. ... ___
068 [82 000
069 F 216
070 e 881
071 50 227
072 30 880
073 Y 940
074 R 108
G
075 [eee eee
bl
076 B 075
077 e 075
8
078 B
15
079 30
228
940
B 226J
E 183
50 881
50 189
U 065
00 000
U 106
eee cce]
A* 219
u 030
e 075
B 192
rewind Ta with interlock
jtTa 11 LAB:rI
OULDt. BE--JJJ SCP
RMAxxx
rI--+ A

J
.bl
transfer control to connector 12
j + (overflow if j = 6)
transfer control to connector 11


-13-
080
A- 195
fb+
1--..,. Ll b
C 192
081 B 990
L 230
082 00 000
QOl 084
transfer control
if Bg = Z
__ ,. ....... ,.. .. ______ .. 'J ....... .., ... _. __ _ .. , ....... ..... _., .. __ __ ,,_ ...... ,. ....... _ ..... __ .. ___ "' .... ___ .. _
_.
083 00 000
U, 026
transfer control to connector 5
--..--.. ....--.. ..-... ____ ...__ "_ .. h'_. ____ ._ ... -.-,..-_ ..._ .._ ......._ ..._,_ .. __ .,.,_ ... __ ...... _. ______
084 B 991
L 192
085 00 000
if = L).b
Q 087
transfer control
__ .___ .."'_ ......"'..__ ._ .........___ ..... __ ... ";0.__ .... __ .......... ___ ... _"._. __ ,.,___ .........."'...
086 00 000
U 045
transfer control
to connector 14
087 [84 000
rewind Tb with interlock
B 999J
088 L 223
if B9 = r
Q 026
transfer
control
6
089 K 000
C 192

090 F 185
B 186
091 G 186
H 185
092 F 224
E 029
093 H 029
E 087
interchange Tb uniservos
094 H 087
E 098
095 H 098
E 101
096 C 101
B 190
097
A- 225
Lb +
C 190
G0
98 [14 000
16
34 940J
099 B 940
L 190
-14-
100 000 000
Q
078
transfer control if B? = Lb
101 [84 000
B 226 ]
rewind Tb with interlock
102 F 216
E 185
103 e 991
50 991

104 50 227
Out.DA BE--Jll sep
50 190
JtllBxx
105 30 940

U 098
0106 [Z4
820

B 106 ]
107 A* 219
.k + 1--7k (overflow if k = 6)
C 106
108[ eee
eee
eee eee]
8109
B 229
e 106
l--+k
110 [56 820

B 23il
J f:J, c +
III
A- 195
H 231
112 L 232
Q 114 transfer control if c = 1981
113 00 000
U 108
transfer control to 0
114 F 230
G 820

115 G 870 Z-).cg
A- 195
116 e 871 fj. c + I--=. c
1
0
117
56 820 e ---+T 6
B 930
c
L 990
118 F 042
if > B2
T 120 transfer control
119 B 028
U 121
1
120 B 029

00 000
_.'"_ ................. __ " ..... " .. "." .. _ .... _ ..... _ ... .......... ...... _ .. _ ... .. _ ... A _ _ .. "" ...... _ ........... _ .... ___ ___ .. _. ___ _
121 .6 000
e 162
122 B 235
E 187
123 H 128
F 210
124 E 233
e 821
125 50 821
50 200
126 B 200
A- 225
127 e 200
} Lc + l_Lc
._' . ".M"W." _.""" .. " ........ _ ........ "." ...... ............ ?;}l.., .... .. .. " .............. ,Q ........... .. c .... . " ........ _ .... .. .. " ... ... .. _._ ..... ........ ,P".,,_"_'_
128 [eee eee .
eee eee J
129 A* 234
e 128
f 3(f"'()'6-_'A'(f6'o .. .. .... " ......................... , .............. " .. .... .,,,. .... .... H..-... .. . _ ......... r _ .... --..... ... --.-.... ---
U 128
_ .. ..... ..._ _"._ .. 'N' ....... "". . ' ......... ............. , ........ _ ............ , ........... . . '" .. "" ' .. _ .... " ............ " .,_ ___ ."_" _____ __
131 L86 000 Rewind Tc with interlock
B 188J
132 F 187
G 188
133 H 187
F 224
134 E 110
H 110
Interchange Tc servos
135 E 116
H 116
136 E 131
H 131
137 E 141
H 141
138 E 152
H 152
139 E 153
e 153
140 e 142
00 000
G 141 ..
142 [ece eee
\!.V ...... .. _""'._ .. _ .. _ ...... .. , .. ............ " .. ....... .. ____ ....... , ... .... _ .... _." .... _._ . H._ .. _ . __ . __ .. __
Q... 143 00 000
U 108 transfer control to b
---.... ,..-_ ... -_ .. _--_._-_.---------------------
-16-
144 R 032
Y2
U 026 transfer control to 5
.--.-. ..................... _ .......... , ............ .......,.".. ...... .,,, I _ ................. _ .... ..., .......... .........,.. ..... _ ... "' ... ........................ Jt ....... .......... ____
145 L 230
Q
147 transfer control if A
O
= Z
1
146 Y 880
Ai--+ ws
U 033
147 Y 230 Z-+rY
00 000
148 R 107
U 106 tItans fer control
to do
149 Z 870
0
Z-+C
6
B 231
150 A- 195
e 871 D,.c + eJ
151 B 223
e 879

152 [ 56 820 C---+Tc
Z 820J
Z---+cP
153 [ 56 820
B 187J
154 R 126
U 123 transfer control to do
..
155 50 236
JtEND MERGE. g -+ sep
U 046 trancfer control to RWO* tapes
____ 'M ............ ......... ..,. ................................ -., ............... ............ JII..... - ......,. ...........
156 B 189
L 191
157 FlO 183 a.--+t
00 000
158 R 182 .8
1
U 164 transfer control to 7
. , .. ................ N_ ................................. " ....... JIIIUt ..........................
.. ............................ -----
159 B 190 Lb ........ L
L 192
160 F 185 b--+t
00 000
9
2
161 R 182
U 164 transfer control to 7
162 [Cee eec

00 000 la 000
ece eec]
2:
00 000 Ib 000
163 00 000
U 108 transfer to control to b
-
.............. --............ .............. ................... ....,.,.....,.. .....
/'
-17-
0
164 C 237
J 238
165 G 239
F 039
166 B 239
E 169
167 H 169
H 178
168 E 172
C 172
169 [1 t 000
30 820J
Tt----+rI
170 B 820

L 237
171 00 000
Q 177 transfer control if CO = L
1
.......... .,
172 [8t 000
RWD T
t
with interlock
B 226J
173 F 216
E 239
174 C. 821
50 821
JtTt fl LABEL fl SH l
175 50 227
-4 SCP
50 237 .MLxxxl\l\l\AI\ __
176 00 000
U 169
177 K 000
C 237
....
178 [1 t 000

30 820 J
rI--+C
179 B 237
}g + A- 195
180 H 237
K 000
181 B 238
T 178
transfer control if g= D.
...
". """
,
182 ecce eee
eee eecJ
--
}a
183
6
22 222
222
222]
184 333 333
} a
333 333
185
r
44 444
}b
444 44
J
186 555 555
Jalternate b
555 555
187 e66
666
}c
666 66
J
188 III III
J al ternate c
III III
. -18-
189 [JtM
190 [JtM
191 [000
192 [000
193 000
194 000
195 000
196 .6A
197 Jt2-
198 JtMO

AOO
BOO
000
000
000
000
000
LT6
WAY
UNT
BLA
000
000
011
oooJ
oooJ
110

000
WID
AME

001
RGE
. PES
NKS AON
199
200
COO

201 11 000
30 100
202 8 100
L 009
203 00 000
005
204 00 000
U 001

205 21
206 Y
207 L
208
209 R2W
210 III
211
212 A-
213 C
000
010
008
960
007
III
000
996
991
30
Z
U
800
Cll
8
Q
100
990
990
991
005
III
991
994

-19-
214 81 000
00 000
215 30 960
U 008
216 001 060
000 000
217 116tJ. AND
L\ 1. JUtJl
218 BOO 930
LOO 990
219 010 010
000 000
220 Y40 890
ZOO 880
221 JtBL OCK
L\co UNT
222 F-R
FRU NJtJl
223


Vi'V
224 101 III
101 III
225 000 000
100 000
226
"Ta
flLA
BEL
227 OUL MB
E-- JJJ
228 Y40 950
ZOO 940
229 Z40 820
BOO 106
230 ZZZ ZZZ
ZZZ ZZZ
231 [000 000
232 000 000
000 000 ]
1 flc
001 982
233 aLA BEL
flTc
---
234 001 000
000 060
235 B83 128
500 000
236 JtEN DflM
ERG E.Jt
237
238
\'1orking storage
239
-20-
6. Reruns with Two O u t H u t ~
The rerun procedure for multiple outputs described below is very similar to
the single output rerun previously described. The only complexities intro-
duced are the problems of repositioning one of the output tapes and of dif-
ferentiating between the output controlling the rerun and the tape that must
be repositioned.
The latter problem is dealt with by writing on the identification block of
every output tape the letter identifying the output file. When the rerun is
initiated, this information is made available to the rerun routine by the
memory dump locator (also in the identiftcation block).
The output tape which is not being rerun is treated as an input tape for
repositioning purposes.
During the repositioning, all tapes are read' into the output block of the
file being rerun. The contents of tl!e input and the other output bloek at
the end of the rerun routine are the same as they were at the time the mem-
ory dump was written. rl is filled by the read subroutine of the normal
instructions.
It is to be noted that the rerun must occur at a point where the contents
of the computer registers (except r!) are of no consequence as the rerun
makes no provision for restoring them. (Please refer to the coding in
SectionS.)
Unlike the single output rerun this routine requires no breakpoint option
to rerun. Breakpoint options represent a manual operation and, therefore,
a possible operator error. Since one of the output tapes contains good
information, a manual error can ruin this tape and necessitate rerunning
from an earlier point.
To rerun, the operator merely mounts the last good output tape of the file
and does an initial read. If this is done correctly, no other manual op-
erations are required and, therefore, no further possibility of error exists.
rl---+ A
A
1
La
Explanatio,n of Flow Chart Symbols
Tape mounted on Uniservo a
A block from Ta is read into input buffer (forward)
The block in the input buffer is transferred to A
The ith item comprising the block A (i= 1,2, , i')
The keyword field of item Ai
The labelldeatifyln;g the current A fi le reel being
processed ..
-21-
fj,a

c
cy. C
The number of A blocks processed on Ta.
The C block is written on To.
The symbol C1. is made identical with the symbol C.
Tape C is rewound and an interlock set preventing
further tape motion until the tape reel is removed.
Servo interchange symbol. a which initially read
2 is changed to 3 and the symbol rotated 180.
Do specified by subroutine Ba.
Supervisory Control Printer
;'22- .
i = j = k =.i = 1
La = Lb = Lc = Ld = 1
= = = = 0
a = 2,b = 4,c = 6,d =
1--.... ' .. i a, b, c , d , La, L b -+ SCP
Process Ai and B
j
to form
Deposit
D Item
i)--0

I Yes
I /23)

C'I
,

I

C\J
I

"Block Count SCP RWD* Ta,T
b
, Tc,T
d
I
tn
N
I
r-- -- 0 !
r >1 -----+ C 1 Rerun Locator --....,.. C

t----iliflk + 1 k I ]Il
"
{z- 1 z + _
C + J:---i>-LC.l

[----,
13 T !
r--"""!L c :-
I
1 z .. C-Tc
.,
Tct-0
I
...0
N
I
Rerun Locator I \@

bd + l-+bd
=
@
-t
d
+ 1-7* )

1H 2Kr
Jl..'; P -+D it : -+Td
&B
RWD* Ta,T
b
"End SCP
I
r--
C\I
I

'-----lIiJ1T rI ..... --
a
r = Last Good
Output Tape
Read R. 11'
a-,----b, La, L b SC'PI )I
Memory Dump on Tr
Memory
d, Lc --+ SCP
c, Lc ---+ SCP
I
co
C\I
I
r-e
H M-.d H d-+t I-@
c-+tt)
t-3
<+!
Ii
1-3
c1"
~
Ii 1-3
'i
H
c+ H
i
M 1:"4 Ii
~
H
i CIl
O'Q
0
M
I-d
+
.....
J
O'Q
o
l
O'Q
';'29-
EFFICIENT CODING TECHNIQUES
1. Introduction
Many data processing applications of the UNIVAC System involve routines that are
to be used repeatedly or which must process large volumes of data. It is fre-
quently desirable to reduce the running time of these routines. This can be ac-
complished by:
1. Reducing the number of instructions executed by
the computer. (Linear Coding)
2. Reducing the time spent in execution of the in-
dividual instructions. (Minimum Latency Coding)
In the discussion to follow we will consider techniques to accomplish these sav-
ings in time. It should be noted, however, that maximum success in reducing the
running time of an average routine can be achieved only by increasing the amount
of memory space devoted to the program.
2. Linear Coding
The instructions of a typical program can be considered to belong in one of the
following categories:
1. Instructions for starting and ending the program.
2. Instructions which read and write blocks of data
on tape.
3. Instructions which move items to (or from) work-
ing storage from (or to) data blocks.
4. Instructions which process the
Instructions in the first category are executed only once per run and, there-
fore, no effort need be exerted to optimize them. For a discussion of instruc-
tions in the second category the reader is referred to the paper "A Survey of
Input-Output TechniquesUs The following will, therefore, be restricted to a
discussion of the instructions in categories three and
Let us assume first that a particular sequence of instructions has been devel-
oped Which will process an item. There are then three choices available in the
development of the routines which will process a block of such items. These are:
1. Repetition of the instructions for processing an
item with the addresses modified for each separate
item in the block.
2. An auxiliary routine to modify the addresses of
the processing instructions.
-1-
3. An auxiliary routine to transfer successive items
into a working storage from which the item is pro-
cessed ..
Clearly the first of these alternatives will involve the execution of the fewest
number of instructions since every instruction nair processing
items. Note, however, that it will also require the most memory space.
Where the available memory does not permit using straight line processing
instructions (method 1 above), a choice will have to be made between methods 2
and 3. In general, it has been found more efficient to transfer items to work-
ing storage than to modify processing instructions; there are some exceptions
to this rule. These exceptions Q,oclilr ,generally' where '!odd'9 s izec items are,tc-
ing processed which cannot be transferred very efficiently within the computer
and where the processing instructions have very few references to the memory
location of the item.
The reader should note that straight line techniques are also available for the
movement of items into or out of working storageo The following examples will
illustrate this:
Example 1: The R-U Counter
100 R 100
U xxx To transfer the first item
101 R 100
U xxx To transfer the second item
102 R 100
U xxx To transfer the third item
R 100
U xxx To transfer the last item
B xxx
C 100 Reset line 100
End of block routine
Example 2: The Function Table
100 B 101
A 102
101 C 101
U 104
102 000000
000004
104 x xxx
x xxx Transfer the first item
108 x xxx
x xxx Transfer the second item
-2-
11:2 x xxx
x xxx Transfer the third item
x xxx
x xxx Transfer the last item
B xxx
C 101 Reset line 101
End of block routine
Both of the methods illustrated are perfectly general although the method in
example 1 requires more memory space than does the method of example 2. The
number of lines of instructions e x e c ~ t e d in moving each item is the same* for
both. Example 2 has a disadvantage in some cases (as we shall see in the
discussion on minimum latency coding) in that it requires that each set of
coding handling successive items be stored in the memory with a fixed spacing
between the addresses.
A further example will aid in fixing these principles in the readervs mind.
Example 3
Select from the four-word stock items on Ta all those items for which the ex-
pected requirements is less than the sum of the on hand and on order and write
them on Tb. The item format is:
where:
SSSSSS
RRRRRR
HHHHHH
000000
SSSSSS RRRRRR
HHHHHH 000000
-other
-other
data-
data-
is the stock number
is the expected requirements
is the on hand amount
is the on order amount
The coding which will process one item is in cells 800 - 803. The coding is
shown below.
010 F 100
B 800
011 E 101
K 000
012 B 801
E 101
013 C 102
X 000
*One line less is used in the R-U Counter than in the Function Table method
in transferring the first item only.
-3-
014 .6 000
015 R xxx
098 F 803
099 V 800
100 111111
101 000000
102
A 102
T 098
B 802
U xxx
000000'
000000
Set exit of output routine
Transfer control if OH + 00 > R
Transfer control to output routine
Temporary storage
This routine requires that at least 6 lines of coding be executed for each item.
Two extra lines must be provided for the item in case it is to be placed on the
output tapeo A block of such items might be processed as follows:
If linear coding were to be,used, a total of 122 '
lines of coding two constanta in
cells 100 and 101) would have to be stored in the
memory exclusive of the output routine. The to-
tal instruction pairs executed per block would be
only 90 (6x15) plus 2 for each output item.
If an auxiliary routine were provided to modify
the addresses of the instructions on lines 010,
012, 098, and 099 so as to be able to process
the succeeding items, the routine would begin
on line 016 and might look as follows:
016 B 010
A* 097
017 C 010
B 012
018 A 096
C 012
019 B 098
A 095
020 C 098
B 099
021 A 096
C 099
022 00 000
U 010
023 B 091 From overflow on line 016
C 010
024 B 094
C 012
025 V 092
W 098
-4-
026
End of block routine
.
091 F40100
B SOO
092 F S03
B S02
093 V SOO
U xxx
094 B SOl
E 101
095 000004
000004
096 000004
000000
097 004000
000004
Here the total number of lines "required for storage is 27, but the total in.
struction pairs executed per block of data is (including 4 for generalized
overflow) 196 exclusive of 2 lines for each output item.
A00ther alternative is to provide an iterative routine that will move items
into a working storage, such as:
016 V40So4
W SOO
017 V S06
W S02
olS B 016
A* 09;'
019 C 016
B 017
020 A 095
C 017
021 00 000
u ala
022 v 096 From overflow on line OlS
W 016
023 End of block routine
095 004004
000000
096 V4oS04
w SOo
097 V 806
W S02
-5-
This method requires only 20 lines of storage. A total of 182 lines of coding
are executed for each block of data exclusive of output.
A last possibility is to again make use of the working storage principle, but
uses linear coding for the routine transferring each item to the working stor-
age. The following is an example using the function table technique:
016 B 017
A 096
017 C 017
U 018
018 V 804
W 800
019 V 806
W $02
020 00 000
U 010
021 V 808
W 800
022 V 810
W 802
023 00 000
U 010
024 V 812
W 800
057 V 856
W 800
058 V 858
W 802.
059 00 000
U 010
060 B 097
C 017 Reset line 017
061
End of block routine
096 000000
000003
097 B 017
A 096
This method requires 57 lines of coding to be stored, but only 163 instruction
pairs are executed for each block of items processed (plus 2 lines for each
output item).
These results are summarized in the following t a b l e ~ Only the total lines for
storage or execution required in processing a block of input items are considered.
-6-
Method Total Lines Stored Total Lines Executed
(excluding Gen OF) (No Output Items)
1. Linear Coding 122 90
2. Iterative Alteration of 27 196
Instruction Addresses
3. Iterative Transfer 20 182
to Working Storage
4. Linear Transfer to 57 163
Working Storage
It is generally true, as the previous discussion has pointed out, that the min-
imum number of instructions to be executed per routine can be obtained by ex-
changing memory space for time. The relative merits of each method discussed
depends to a large extent on the amount of processing required for each item
and on the item size. Where memory space is not too critical, the programmer
would do well to investigate linear or semi-linear (methods 1 and 4 of the
above table) techniques for reducing problem running time.
3. Introduction To Minimum Latency
A study of UNIVAC logic reveals that the time required to execute an instruc-
tion pair falls into two categories:
1. Time spent in actual execution of the instructions.
2. Time spent in waiting for information to appear at
the read out or erase gates of the memory.
The first of these is constant and fixed by the logic of the computero The
second of these, called latency time, depends for its duration upon the re-
lation between the Time Selection Counter (TSC) and the Time Selection Digit
(TSD) of the memory address involved. By judicious selection of memory lo-
cations, the latency time or a routine can be materially reduced.
Consider first an, instruction pair that' does 'not involve a transfer of con-
trol. It is desirable to minimize the elapsed time between the selection
of the instruction pair at address m and the selection of the next pair at
atidress m + 1.
If the TSD of m is k, then the ~ Time Selection Minor Cycle occurs when the
TSC is reading k. For the instruction pair in m + 1 this will occur at a
TSC reading of k + 1. An instruction pair cannot be executed in one minor
cycle; therefore, the minimum time between ~ time selection will be 11 minor
cycles: ten minor cycles to step the TSC around to k and one more to k + 1.
If the instruction pair is not executed in this time, additional time, in
units of ten minor cycles (one Major Cycle), will elapse before the ~ Time
Selection Minor Cycle of m + 1.
-7-
~ T S yTO * * * * * * * * * yTS oTO * * * * * * * * * oTS aTO aTon ~ T O * * * * * *
I I I I I I I I I I I
5' 6 7 8 9 o 1 2 3 4 5' 6 7 8 9 o 1 2 3 4 5' 6
~ T S yTO yTS oTO * * * * * * * oTS aTO aTon ~ T O * * * * * * ~ T S
I I I I I I I I I
5' 6 7 8 9 o 1 2 3 4 5' 6 7 8 9 o 1 2 3 4 5' 6
~ S yTO yTS Mo MS aTO a T ~ ~ O *" * * ~ S
I I I I I I I I I
5' 6 7 8 9 o 1 2 3 4 5' 6
*Latency Time
7 8 9 o 1
Total Time 31 :r.c
Latency Time 23"M::
2
Ratio L.T./Err.T. 2.9:1
Total Time 21 M:::
Latency Time 13 M:::
Ratio L.T./Err.T. 1.6:1
Total Time 11 Me
Latency Time 3 :r.c
Ratio L.T./Err.T .38,1
3 4 5' 6
0)
The following examples will illustrate these concepts: ~ s e e Figure 1)
I. 025 B 056 L 057 31 MC
2. 025 L 057 B 056 21 MC
3. 025' L 057 B 059 11 Me
Example 2 interchanges the instructions in example 1 which results in sub-
stantial time savings. In example 3 a further increase in speed comes from
changing the address of the B instruction.
The critical parameters ,involved in a calculation of the latency time are:
1. The difference between TSDvs of the address
of the LH instruction and the instruction
line number (n).
2. The difference between TSDus of the address
of the RH instruction and the instruction
line number (m).
3. The amount of time taken up in the actual
execution of the Left and Right Hand In-
structions individually. .
4. Construction of the Tables
--- ---
The following will refer to a limited number of instructions which will act
as prototypes. For each UNIVAC instruction (except tape orders) there is a
prototype which has the same execution time.
TABLE OF PROTOTYPE INSTRUCTIONS
Prototype Instructions
B B C F G H J L R E
A A S
K K X 00
U U
T Q T
Ox Ox -x .x ;x
W V W
Y Y Z
M M N P
D D
-9'-
Thus"the previous examples would be written:
1. BIB 2
2. ,B 2 B 1
3. B 2 B 4
To simplify tables of latency times, references to prototype instruction
pairs such as BB, AT, WU,etc., will be made as illustrated above.
EXAMPLE PROBLEM:
Pte pare a table of latency times for the instruction pairs K Band B K.
KB

y y 0 OTS for 11 Me a a

a. a

TS TO Ton
TO .
A
"'
TO Ton TO TS TO Ton TO TS
r
I
, ,
I

I I II
0 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1

oTS for 21 Me
The time at which TS could oCcur is. at TSC 4. The last point at
which it could occur and still leave enough time (3 Me for TO Ton and
TO) for the instruction pair to be executed. in 11 Me is 7. If m is' 8, 9,
0, 1, 2, or 3, then the second TS is the one that occurs and the operation
takes 21 MC.
Y TS for 11 Me a a B B
'rS
a a
TO Ton
" , I I ,I --t--'-I'!---t-I - ..... ,
o 1 2 3 4 5 6 7 8 9 02 3 4 5 6 7 8 9 0 1
__ .J
yTS for 21 Me
For n between 2 and 7,the instruction takes 11 MC; and for n between 8, 9,
o and 1, the instruction takes 21 Me. This problem is relatively simple
since the latency time is independent of the address of the K instruction.
STUDENT EXERCISES:
1. An A order takes 3 Me for execution, a B order takes 1 and an X order
takes 1. How long does it take to execute the following instruction pairs?
(Do not forget the TO minor cycles which follows each instruction.)
025 B 030
026 H 058
027 C 056
A 056
X 000
B 025
2. Prepare a table of instruction times for the prototype pair B n B m.
-10-
PART I
MINIMUM LATENCY
TABLES
-11-
Line L = [ B L + n B L + m ]
n '2 3 Lj. 5 6 7 8 9 0 I
8
9
o
1
2
3
ij.
5
6
7
8
9
0
'I
2
3
ij.
5




Li ne L == [A L + n A L + m ]
21
- 12 -
Av = 21.5

Av = 25.5
L i np. L = [B L + n A L + m ]
3
4-
5
Line L = [ A L + n B L + m ]
- 13 -
Av = 23.5
Av = 23.5
Line L = [K B L + m ]
m =
Av = 17
Line L = [8 L + n K ]
n =
[!!]
Av = 17
Line L = [K A L + m ]
m
Av = 19
Line L = [A L + n K ]
n =
Av = 19
Line L = [K U L + m ]
Av = 12.5
- 14- -
Linp. L = [ K K
Av :::: II
Line L = [B L + " U L + m ]
n
2 3 4- 5 6 7 8 9 0 I
m
8 8 18 18 18 18 18 18 18 18 18
9 9 9 19 19 19 19 19 19 19 19
0 10 10 10 20 20 20 2.0 20 20 20
1 II II
"
II 21 21 21 21 21 21
2
12 12 12 12 12 22 22 22 22 22
3 13 13 Ij
I)
OJ
1 i",
v 13 23 23 23 23
4- 14- 14- 14- 14- .14 lit 14 24- 24- 24-
5 15 15 15 15 15 15 15 15 25 25
6 16 16 16 16 16 16 16 16 16 26
7
17 17 17 17 17 17 17 17 17 17
Line L = [A
L +" U
L + m ]
2 3 4- 5 6 7 8 9 0 I
U 10 20 20 20 20 20 .20 20 20 20
I
II II 21 21 21 21 21 21 21 21
2 12 12 12 22 22 22 22 22 22 22
3 13 13 13 13 23 23 23 23 23
'" ~ .
L';)
4- lit 14- 14- 14- 14- 24- 21t 21t 24- 24-
5 15 15 15 15 15 15 25 25 25 25
6 16 16 16 16 16 16 16 26 26 26
Av ~ ~ . I 9
7
17 17 17 17 17 17 17 17 27 27
8 18 18 18 18 18 i8 18 18 18 2B
9 I ~ 19 19 19 19 Id 19 19 19 19
- 15 -
a) TRANSFER
L i ole L = [B L + n T L + m ]
n
2 3 4- 5 6 7 8 9 0 I
m
0
10 20 20 20 20 20 20 20 20 20
I
II II 21 21 21 21 21 21 21 21
2
12 12 12 22 22 22 22 22 22 22
3 13 13 13 13 23 23 23 23 23 23
4- I 4- 14- 14- 14- 14- 24- 24- 24- 24- 24-
5 15 15 15 15 15 15 25 25 25 25
6 16 16 16 16 16 16 16 26 26'
26
7
17 17 17 17 17 17 17 17 27 27
8 18 18 18 18 18 13 18 18 18 28
9 19 19 19 19 19 19 19 19 19 19
Av = 19
b) NO TRANSFER
n ==
- 16 -
a) Transff3t"
Line L = [A L+ n T L + m ]
n
2 3
q
5 6 7 8 9 0 I
m
2 12 22 22 22 22 22 22 22 22 22
3
13 13 23 23 23 23 23 23 23 23
q
Iq Iq Iq 2q
24-
2q 2q 2q 2q 2q
5
15 15 15 15 25 25 25 25 25 25
6 16 16 16 16 16 26 26 26 26 26
7 17 17 17 17 17 17 27 27 27 27
8 18 18 18 18 18 18 18 28 28 28
9 19 19 19 19 19 19 19 19 29 29
0 20 20 20 20 20 20 20 20 20 30
I 21 21 21 21 21 21 21 21 21 21
Av = 21
b) NO TRANSFER
n = 2 3 ql 51 til7 8 I 9 o
@
- 17 -
a) TRANSFER
Line L = [K T L + m ]
m =
5' 17 8 9
,51,: : ,71, 81, 9 I
Av = 12.75
b) NO TRANSFER
- 18 -
Line L = [B L + n 0 x
6 7 890
6
~ r - ~ ~ - + ~ ~ ~ ~ ~ ~
Av = 16 + X
7
r - + - ~ - r - + ~ ~ ~ ~ ~
8
~ ~ + - ~ ~ - 4 - 4 ~
9
Line L = [0 x B L + m ]
x
8 9 o I 2 3 q 567
I
2
3
4
5
6
Av = 16+x
7
8
9
- 19 -
Line L = [ A L + n 0 x
] .
6 7 8 9 0
,)
V
4
~
5
6
Av = 18+ X
7
8
9
Line L = [0 x A L + m ]
234- 5
Av :;: 18+x
- 20 -
x =
[ Ox Line L =
Line L = [ K
L" (:l L=[O xI In.
K ]
Ox]
- 21 -
Line L = [0 x U L + rn ]
x
I 2
"I
4 5 6 7 8 8
m
~
8
d 18 18 18 18 18 18 18 18
3 9 9 13 19 l ~ 19 19 19 13
0 10 10 10 20 20 20 20 20 20
I II II II II 21 II 21 21 21
2 12 12 12 12 12 22 22 22 22
3 13 13 13 13 13 13 23
f) .,
L ~ 2'-' .:J
4 14 14 (4 14 14 Itt 14 24- Zit
AVI = 11.5 + X
5 15 15 15 15 15 15 15 15 25
6 16 16 16 16 16 16 16 16 16
7 17 17 17 17 17 17 17 17 17
a) TRANSFER
L i I'lP' L = [0 x T L + m ]
x
I 2 3 4. 5 6 7 8 9
m
0 10 20 20 20 20 20 20 20 20
I II II 21 21 21 21 21 21 21
2
12 12 12 22 22 22 22 22 22
3 13 IJ 13 13 23 23 23 23 23
4 14 14 14 14 14- 24- 24- 24- 24
5 15 15 15 15 :15 15 25 25 25
6 16 16 16 16 16 16 16 26 26
7 17 17 17 17 17 17 17 17 27
8 18 18 18 18 18 18 18 18 18
9 19 19 19 19 19 19 19 19 19
b) NO TRANSFER
x = _"--. ~ 3 _4 __ 5 __ 6 _21_1 _7 __ rl _ 9 ~
- 22 -
Line L = [Y B L + m. ]
m = 3 I 4- 15 6 7
Av = 26
Line L = [B L + mY]
m =

Av = 26
Line L = [Y A. L + m ]
m

Av = 28
Line L = [A L + mY]
m

Av = 28
- 23 -
Line L =; [K Y].
Line L =:[Y kJ
Av = 2.1
Line L = [Y Y]
Av = 31
Line L ='[Y Ox]
line L =; [0 x Y ]
x =
1 1 2 1 3 1 ~ 1 5 __ .
,fI'
- 24- -
a) TRANSFER
Line L = [Y T L + m ]
m = 9 0 I 2 3 5 6 7 8
I hl
20 21 22 L3 25 26 27
2t! ]
[!!J
Av = 22.25
b) NO TRANSFER
Line L = [Y U L + m ]
m =
Av = 20.5
Line L = [Y W L + ]
m =

Av = 27
L = [W L + n Y 1
n =
Av = 27
- 25 -
Line L = [B L + n W L + m
m 2 3 567890
7

8

9
o

I
2
3

5
6

Line L = [ W L + n B L + m ]
8
9
o
I
2
3

5
6
7
2 3 567 890




-26 -
Av = 22.5
Av = 22.5
Line L = [W
\
L + n A L +m ]
q.
5.
Line L = [ A L + n W L + m )
m 2 3 567890
7
I---I---pr.c....y" /.
8
I---+--+---f-&o--v.
9


I
2

3

4-

5
6
- 27 -
Av =
Av =24-.5
x
Li n ~ L = [W L + n 0 x
67890
If
5
6
7
8
9
L i n ~ L = [0 x w L +m ]
I
2
3
It
5
6
7
d
9
7 8 9 0 I 2 3 ~ 5 6
~ + - 4 - ~ - + - 4 ~ ~
~ + - 4 - ~ - + ~ ~ r - 1 - - r ~ ~
~ n ~ ~ ~ - ~ - ~ ~ + - ~ - r ~ ~
- 28 -
~
Av = 17 + x
Av = 17 + x
a) Transfer
b) NO TRANSFER
Line L = [W L+n W L+m]
n2 3 4- 5' 67890 I

7
8
9
o
I
2
3
4-
5
6


t---t---t---t



Line L = [ W L + n T L + m ]
n
2 3 4- 5 6 7 8 9 0 I
m
I II 21 21 21 21 21 21 21 21 21
2 12 12 22 22 22 22 22 22 22 22
3 13 13 13 2'" \) 23 23 23 23 23 23
4- 14- 14- 14- Pi 24- 24- 24- 24- 24- 24-
5 15 15 15 15 15 25 25 25 25 25
6 16 16 16 16 16 16 26 26 26 26
7 17 17 17 17 17 17 17 27 27 27
8 18 18 18 18 18 18 18 18 '28 28
9 19 19 19 I J 19 19 19 19 19
0 20 20 20 20 29
20 20 20 20 20
9 o I I
- 29 -
Av -== 23.5
Av = 20
L i np L [IN L + n K
n =
5 1
6
7 I . ~ 1
9
1
0
"1

~
Av = I ~
LinA L = [K W L + m
rn =
Av = I ~
L i np L = [ VI L + nUL + rn J
n
2 3 4- 5 6 7 8 9 0 I
m
J 9 19 19 19 19 19 19 19 19 19
0 10 10 20 20 20 20 20 20 20 20
I II II II 21 21 21 21 21 21 21
2 12 12 12 12 22 22 22 22 22 22
J 13
1 r')
\J 13 13 13 23 23
'1 ,)
L.v 23 23
4-
14- 14- 14- 14- 14- 14- 24 24- 24- 24-
Av WU = 18
5 15 15 15 15 15 15 15 25 25 25
6 16 16 .16 16 16 16 16 16 26 26
7 17 17 17 17 17 17 17 17 17 27
8 18 18 18 18 18 18 18 18 18 18
- 30 -
It is not generally possible or desirable to program each line of a routine in
minimum time; ther'efore, the following order of precedence should be alloted to
the sections of a data processing run.
I. Instructions which process each individual item.
2. Instructions', whd.ch conceIln-: the -transfer of blocks
into and out of the computer.
Consider the following routine which is designed to place consecutive two-word
items into temporary storage.
100 B 110 A* 109
101 C 110 U 110
102
103 tape read
routine
109 001002 000000
110 V70300 W 200
III processing
112 routine
100 B 113 A* 1 0 ~ ) '
101 C 113 U 113
102 tape read
routine
105 001002 000000
113 V70300 W 200
114 processing
115 routine
MAXIMUM
31
19
50 MC item
121 seconds/tape
MINIMUM
11
12
23 MC item
56 seconds/tape
-31-
Thus, the total savings obtained by changing two lines of instructioo$ is of the
order of 65 seconds per tape processed.
Practically every routine on UNIVAC makes use of generalized overflow for control
purposes$ It is, therefore, useful to have this frequently-used routine in mini-
mum form. Compare the following:
000
050
051
052
053
000
008
009
010
all
012
R 052 U 050
B 052
C 052
[
000000
A 053)
U 052
- ]
000001
R 012 U 008
BOlO
Call
000000
[
[
A 012
U all
000001
- ]
- ]
AVERAGE
10
21
21'
52 MC
MINIMUM
8
11
. .JL
31 MC
Additional tables have been prepared for instruction pairs involving multiplica-
tion by a known multiplier. Because these tables are quite extensive, only the
most common instruction pairs, that is, BM and MB, are shown. However, as ex-
plained below, instruction times for KM and MK prototypes can also be looked up.
The tables have been prepared in a fashion similar to that used for the other
arithmetic instructions. However, multiplication is not a fixed length operation;
the number of stages required depends on the number of additions performed during
the multiplication, which in turn, depends on the multiplier digits.
Figure 2A shows the stages required for multiplication involving no additions,
i..e .. , multiplication by zero. A sample problem is solved for the instruction
pair
B 2 M 6
in Figure 2B. This instruction pair requires 31 minor cycles, as shown. (The
multiplication stages are
If additions are involved, one minor cycle is added for each addition. Figure 2C
shows a sample problem
B 2 M 6
where the multiplier is a single digit requiring two additions. Actually,
the additions occur somewhere between M5 and M15, but they are attached to the
end for descriptive convenience.
-32-
If an instruction of the type K is involved, instead of type B, these tables may
still be used o If K occurs as a LHI, use the appropriate table for BM with n = 2.
If K occurs as a RHI, use the proper: table MB wi th m = 70
The following chart shows the number of additions required for various multiplier
digitse
Multiplier Digit
o
1,3
2,4,6
5,7,9
8
Number of Additions
o
1
2
3
4
If the mul tipl,;t:ftr c/::>otain$l: mpre than one digit, the total number of additions is
equal to the sum of the additions required for all the digits o For instance,
Multiplier
25
75
333:33333333
-33-
Number of Additions
5
6
11
A B
C
Stage TSD Stage TSD Stage
TO 0
PTS
0 TS
MI (Mul tiplier -.rX) I yTO I
yT?
TO 2

2
t
TS
M2 3 *: TO 3 * TO
TO 4 4
M3 5 5
TO 6 * MI 6 * MI
M4 7 * TO 7
* TO
TO 8 M2 8 * M2
M5 9 * TO 9 * TO
M6 0 * M3 0 * M3
M7 I * TO I * TO
M8 2 * M4 2 * M4
M9 3 * TO
3 *TO
MIO 4 * M5 4 *M5
MIl 5
* M6
5 * M6
Ml2 6 * M7 6
* M7
Ml3 7 * M8 7 * M8
Ml4 8 * M9 8 * M9
TO 9 * MIO 9 * MIO
Ml5 0 * MIl 0 * MIl
I * M12 I * Ml2
2 * Ml3
2 * MI3
3
* Ml4
3 * Ml4
4
* TO
4 * TO
5 * Ml5
5 * Ml5
6 aTO 6 * Al
7 aTon 7 * A2
8 PTO
8 a TO
9 9 a
0 0 TO
I
I TS
Figure 2
PART II
MINIMJM LATENCY
TABLES
-35-
NO ADDITIONS
Line L = [ B L + n M L + m ]
n23 LJ.!5 67890 I
m
9

o


2
3
LJ.
5
6
7
8

8
9
o
J
2
3
LJ.
5
6
7
Line L = [ M L + n B L + m ]
2 3 LJ. 5 6 7 8 9 0 I




ONE ADDITION
Line L = [8 L + n M L -+ m ]
2 3 LJ. 567890 J
m
8
9
0
I
2
3
LJ.
5
6
7
- 36 -
Line L = [M L + n B L + m ]
23 LJ.5 67890 I




2
3
LJ.
5
6
7

7
8
9
0
I
2
3
4-
5
6
3
4-
5
TWO ADDITIONSI
Line L = [B L+n M L + m ] Line L = [M L+n B L + In ]
2 3 LI- 5 6 7 8 9 0 I
m
2 3 4- S 6 7 8 9 0 I
8
9
0
I
2
3
4-
5
6
7
THREE ADDITIONS
Line L = [ B L + n M L + m ] Line L = [M L + n B L + m ]
- 37 -
IFOUR ADDITIONSI
Line L = [ B L + n M L + m ] Line L = [M L + n B L + m ]
3 6
~ + - + - ~ ~ - + ~ - - ~ ~ + - ~
ll. 31 7 31
IFIVE ADDITIONSI
Line L = [ B L + n M L + m ] Line L = [M L + n B L + m ]
3
- 38 -
I SIX ADDITIONS
Line L = [B L + n M L + m ]
Line L = [M L + n B L + m ]
SEVEN ADDITIONS I
Line L = [ B L + n M L + m ]
L i L = [ M L + n B L + m ]
mn 2 3 4- 5 6 7 8 9 0

n
2 3 If 5 6 7 8
m
8
3 'l:;Q'//.;;,,"/hQ'//. 9

0
('ffffff.'ff .....
5 I
....
6 2

7 3
I--+--+---t--........ ... // /
8
4-

9 5

o
6
7'
- 39 -
I EIGHT ADD I T ION S I
Line L = [ B L + n M L + m ] Line L = [M L + n B L + m ]
n
2 3 4: 5 6 7 8 9 0
m
I
2
3
4:
5
6 3

7 4-

8 5

9 6

0 7
NINE ADDITIONS
Line L = [B L+n M L+m] Line L = [M L+n
B
L + m ]
0 I
n
2 3 4- 5 6 7 8 9 0 I
m
8


9
0
I
2
3
4:
5
6
41
7
- 4:0 -
70 EfficieQ1 CodiJC!.9, Example
The following problem will further illustrate the use of linear and minimum la-
ten.cy codinga> It:is designed as a subroutine of CII payrOlll computation" A net
pay is supplied to the routine in rAo Compute the number of bills and coins of
the vax'i,cn.lls required to make up a pay envelope as follows I
0 OOOOOOOxxxxx
{fo.
total
1 OOOOOOOOOOOx number of $100
2 OOOOOOOOOOOx
QO QO
$20
3 OOOOOOOOOOOx
00 00
$10
4 OOOOOOOOOOOx
eo
$
5
5 OOOOOOOOOOOx
00
"
$ 1
6 OOOOOOOOOOOx
00 00
25
7 OOOOOOOOOOOX
09 00
10
8 OOOOOOOOOOOx
00 QO
5
9 OOOOOOOOOOOx
00 OQ
l
103 F 116
'H 200 Amount to 200 11
104 .4 000
G 201 # $100 bills to 201 11
105 e 200
L 192 11
106 .3 000
X 000 11
107 C 142 laos digit + 143 - R. H. address 142
F 184 11
108 E 200
02 000 11
109 A' III
C 115 Uni to s digit + 130 - R. H. address 115 11
110 F 112
U 118 8
III
U 143
112 000 000
000 all
113
000 002
114
000 001
115 [ 8/17
]
116 000 000
001 000
117
118 E 200
K 000 21
119 P 191
31
120
C 206 *IoPo x 4 = # of quarters 11
121 X
01 11
122 K
P 193 **19 e Pe (F <i> P Q X 10/4 = # 0 f dime s ) 41
123 C 207
X 11
124 K
P 189 I. Pe (LSD x 20/10/4) = # of nickels 31
125 C 208
X 11
126 K
P 190 Ie Po (LSD X 100/20/10/4) = # of pennies 31
*I.P"
= Integral Part
**F. p. = Fractional Part
-42-
127 C 209
U 115 8
128
129
130 V 202
U 14G 10
131 V 173
U 140 9
132 V 154
U 140 18
133 V 175
U 140 17
134 V 170 Pick correct comb. of 10 & 20 bills
U 140 16
135 V 172
U 140 15
136 V 178
U 140 14
137 V 180
U 140 13
138 V 182
U 140 12
139 V 181
U 140 11
140 W 202
y
200 21
141 EXIT M
SUBoM
142 [w 204
U 130 ] 9/18
143 V 155
U 142 9
144 V 156
U 142 18
145 V 167
U 142 17
146 V 158
U 142 16
147 V 169 Pick up correct comb. of 5 & 1 bills
U 142 15
148 V 162
U 142 14
149 V 161
U 142 13
-43-
150 V
151 V
152 V
153
154
113
153
164
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
000 000
000 000
000 000
000 000
000 000
U 142
U 142
U 142
000 003
000 001
000 001
000 003
000 001
000 001
000 001
000 004
000 002
000 004
000 002
000 000
000 002
000 001
000 000
12
11
10
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
000 000
000 000
000 000
000 000
000 000
000 000
000 000
000 000
000 000
000 000
000 000
004 000
W 204
000 000
000 000
000 000
000 001
000 001
000 003
000 000
000 003
000 001
000 004
000 000
aoo 100
000 002
000 005
000 000
'U
13,0
000 025
000 000
000 000
-45-,
STUDENT. EXERCISES
le.) Cooe and time an efficient routine which will place into
rA the smallest of the 4 quantities in memory location

20 Code and time an efficient routine which will add the
300 words in the memory starting at location 1000
Code. and time an efficient routine which will check a
tape. of one-w()rd items for ascending sequence Q Assume
the standard convention is followed o

COLLATION METHOD OF
SORTING
This section assumes the reader is familiar with the general principles of
collation as described in the paper ~ A Brief Description of Sorting Methods
for the Univac System." The following discussion is intended as a further
examination into the manner by which collation is achieved on the Univac.
As the intent is instructional, the methods described are fundamental ones
basic to all collation runs, but, of course, this section is not intended
as a primer of all the shop tricks employed in speeding up collation rou-
tines. For simplicity in the presentation of these principles, two-way
collation will be describedo
Plates 1-5 describe a two-way collation for ten-word items. The following
symbols used in these flow charts are defined belowo
T ~ A
a
A ~ T a
A
A.
1
La
C-+Tc
2,S,/.,==
Tape mounted on Uniservo a
The next sequentially available block from Ta is
brought into the computer in a forward direction.
Same as above, except the tape moves in a back-
ward direction
The block from Ta (consists of 6 items)
The ith item of A (i = 1, 2, ... , 6). The items in
a block are numbered from top to bottom.
The keyword of Ai
The label in the identification block of Ta
The kth C item (k = 1, 2, ... , 6)
The collation of 6 C items
The C block is written on Ic
Sentinels; The following collation sequence applies
j < A ~ < 2 < = for all A ~
The flow charts assume that the input tape to be sorted is in the standard
data tape format described in the section on identification blocks and re-
runs. This input tape is assumed to be mounted on T5 but any other servo
will do (except Tl or T2 unless these servos are also changed).
The instructions are assumed to be in the memory.
At the start of the problem certain counters and variable connectors are
set which will be described as they are encounteredQ Next, an "S" sentin-
el is placed in the key word position of a block, -{: S-+Af r- ' and this
block is written twice on the fir$t output tapes, Tl and T2. This is necessary
since these tapes will later be read backwards and the sentinel blocks are there-
fore needed to indicate the front erid of the tapese
At "Tape labeV,!,. SCP the routine requests the operator to type in the desired
input tape label, La. The first block on T5 is then read into the computer and
t11e lable recordect on tape is compared with the label typed in by the operator.
If the labels do not T5 is rewound with interlock
9
and the tape label and
typed printed on If.the operator finds that he typed
rectI y, hH?y force tuna .. at breakpoint 1. The rOtltine will then re-
tyrn and allow type the correct label. If the input tape in-
cq!:rectl Y labeled, . he may mOllnt correct pe, forc. e transfer on breakpoint 1
and the routine will return to When is reached we are assured that the
input tape to be sorted is the one. The tape label, La, suitably
modified to indicate that the items are sorted, will be placed on the final out-
put tape.
The logical operations between (0 and 6Q) are commonly referred to as the "in-
ternal sort"e The purpose of the sort is to place the items of each
block in order and then (in two-way collation) to split this input into two
nearly piles.
At C1) the first data block from 15 is read into the computer. This block is
called UA
99
:' its individual items are Al, A
2
, .. "., A6' anyone of which is iden-
tified by Ai. N(;:xt, the key word of the last A item of the block is compared
with a Z sentinel for the end of tape teste If this is not the last block on
the input tape, the items of A are arranged in ascending sequence by one of the
methods to be described later. This ordered A block is written on Tm (m = 1,
ini tiall y); one is added to f::., which is the input block counter; and since the
variable connector aI' is set, output tape designation is Chan. ged from servo
I to 2" Control is returned to whence it follows that each block from T5 has
its individual items ordered and IS then written on either Tl or T
2
0 When the
first sentinel block is read, e posi ti.oD of the last valid (nonsentinel)
item is determined Then at 4 the remaining item positions are filled
wi th Z sentinels!, variable connector 2 is and control is transferred to @
where the block is sorted and written on the current output tape.
At the number of blocks to be sorted
9
a , is examinedo The purpose of this
examination is to-determine whether the first Uexternal sort" or merging will be
in or order. Since the final output should be in ascending
the first merge must be an ascending one if the number of data passes
required is odd and it must be a descending one if the data passes required are
even. This fact is clarified in the following table which shows that each exter-
nal pass over the data changes its The number of data passes can be
obtained Aom the block counter ll,. If the block counter is such that control is
sent to \ZJ sort is set to do an ascending merge; but if control is
transferred to a descending merge is set up_
-2-
Cumul.ative Consecutive Input Input Sequence Consecutive Output Type of Output
Data Passes Blocks in Sequence Required Where Blocks in Sequence Sequence for
Total Data Passes Total Data Passes
CX:Id Even CX:Id Even
1 1 *A *A 2 A D
2 2 A D 4 D A
3 -A
D A 8 A D
4 8 A D 16 D A
5 16 D A 32 A D
6 32 A D 64 D A
7 64 D A 128 A D
8 128 A D 256 D A
9 256 D A 512 A D
I
(\')
10 512 A D 1024 D A
,
11 1024 D 2048 A
* Fixed because of ascending internal sort
When the items are arranged in ascending fashion from the top of the block to
the are in sequence from bgtton to Ii
fore possible to perform a descending merge first when the internal sort pro-
duced an ascending sequence for the items in a block. To avoid confusion the
item positions in the block are always numbered from top to bottom (AI, ,A
2
, ,
A6 for ten-word items). To pick up the items from the bottom, the external
pass counters, i and j, are set to 6; the additive necessary to
alter i and j for the next item is set to -1; the constant, L, to which i and
j are compared to determine when the last item of a block has been used is set
to 1; and n, the reset for i and j, is set to 60 Note in this routine the
above is the normal of the counters. Only if the first external merge
is an ascending one ,and only. on the fi,rst pass, are,tl:te ,sf?t to
the values indicated In either case, we first go to 1 where
sentinels are plaqed 6n the two output tapes and,' after setting the counters
appropriately, transfer control to@ where the external sort proper begins.
As indicated at 0, an input tape containing only'one or't1JVo to be sort-
ed must be given special Consider the case for only block of
data. (This a partial block since the first block containing sentinels
on T 5 was 1n& ).. rst, we note that the block ,i. s . sorted, but
without the ID block. At 24 this single data block is read back into the com-
puter; control is sent to su routine 4 where the ID bl6ck'is on T
e

Control is returned to the current the data block is wtittenon Te along
with a full sentinel block; the tapes are rewound and then is stopped.
If ther. .were only t. wo blOCkS .. Of .. dat.a, cO .. ntrol .. w .. O .. UId b .. t .. rans.fe .. rr. ed ... tg
4 where the labe+. is. placed 01}, the output t?pe and the W is ,
set since only one merge is necessary. After, returning to \Zj, an merge
is performed. It should be noted that subroutine 1 is not entered; hence, no sen-
are placed at the top of the output tape.
The following is a description of the merge:
At @ the external collation begins. External collation is merely a succession
of "two-way titapet9, merges which are actually successive merges for many strings of
blocks. The merges alternate with each complete pass between ascending and de-
scending sequence. The prime problem in external collation is the determination
of the end of a data In the desirable case, the strings all consist of
the same number of blocks. This number is called the string iengtho By counting
each block merged and comparing this count with the string length counter, the
end of the string is easily determined. For all strings to be of equal length,
the number of data blocks to be sorted must be a power of twoo Usually, this
does not occur and therefore strings of partial length will be encountered. As
shown in the following example, these partial strings are always at the beginning
or the end of a tape.
-4-
Internal
--.s ort --1'\
--x x
x x
x x
x x
x x
x x
x x
x x
x x
x x
x x
(11
xx (1x (nxx
x x ( x x x x
x x x x x x
Q) xx x ( xx xx
g' xx g' xx xx
xx xx xx
01 xx g xx en xx g xx
x x :.8 x x x:.8 x x
-g xx -g xx
xx xx xx
xx xx xx xx
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x0J:
x x
x
x x
x x
x x
x x
x x x x x
x x x
Input
x Represents a data block
All merging is done by the items within each string
Strings of blocks are spaced to between strings
-5-
x
x
x
(
x
01
X
H
(
X

01
X
C
X
rl
'"0
X
C
(
x
u
(/)
X
<
X
X
X
X
X
X
X
X
X
X
X
X
X
X
x
Output
The end of partial strings at the front of a tape can be detected by the
reading-in of a sentinel_block recorded at the of the tape. The
length (number of blocks) of the strings at the back end of the tape may
be obtained by counting the number of blocks written. Since only the count
for the last string is necessary, this is a feasible procedure.
Initially, the block counters for and and the output block
are set to zero. The input string length, R, is set to one
(one block"strings). The input tapes, Ta and Tb, are set to read Tl and
T
2
, respectiVely, and the output tapes Tc Td are set to T3 and T4.
At @ the two-standby block PI'9CedUre f.or information from.Jape is
set up. The first data block t a is read into A (1- Ts b sigl}ifies
backward read). The second block from Ta is read into the A standby block,
A. The first block from Tb is read into B, the second Tb block is read into
rio Y, which indiates the location to which the block in rl2riill be trans-
ferred, is set to B, the B standby block.
At OJ) , the variable connector is encountered. The connector was set to
first merge is to be a descending one and to 2 if it is to be
ascendings At the item with the greatest keyword is selected and sent to
working storage. At the item with the smallest keyword is selected. In
either case, we go to subroutine 2, where the last item selected is sent to
the current output position, Ck. Note that the output block, C, is always
filled from the top_ Also note, that when each C block is written on the
current output tape, T@, a count is kept by adding one After return-
ing from subroutine 2, the next input item is obtained to replace the one just
put in the output blockdo Whenever an input block is exhausted, control is
transferred to either or These routines set up the two-standby
block read routine so as to make available the next A or B block, as required.
In subroutine 3, we find the read routine itself.
Upon returning from subroutine 3, the appropriate input block counter, or
is increased by one, and the new A or B block is obtained from the stand-
by position. As mentioned before, the end of an input string is indicated by
either the input block counter becoming equal to the string length counter, R,
or the presence of an S sentinel in the block just transferred into A or B
from the standby block position.
Since at least one block of data must be present on both Ta and Tb for the
first external merge, it is necessary to consider the end of an input string
of exactly R blocks. When an input block from one input tape is processed
completely, say the A input. is increased by one for the block just pro-
cessed and the next A block is brought from the standby block, A, into posi-
tion A. Next, the input block just obtained is examined for sentinels. If
-6-
none are presenti it means the end of has not yet been reached. Assuming
this is the the block counter 1S compared with the string kngth R.
If we have reached the end of a string, control is transferred to Since
E, is set initially, we will set variable connectors and E() to and
y 2. Control is returned to @ to send out the rema1ning B items of the
?tring. When the. B string is completely processed, = R and control is
transferred to@,theendofstringoperations.Herethe input output
block counters are reset; the end of string connectors (() ct) . are
reset; and the output tape switched. The routine then returns to 11 to
merge the next strings.
At some time the condition will be obtainedo This condition sig-
nals end of input'string as well. Consider the case where the end of Ta is
first . C?ntrol is sent,to .where contents of E are in
posltlon of A6.(A
1
1f 1S the f1rst external merge be1ng done
i
l1
ascending sequence) I f an ascending merge is being done F= 99 = to, other-
wise E = /.. 9
2
is then set and control transferred to @. 'The /. or = in Aj
and the new variable connector settings force all the remaining B items of the
current string onto the output tape. Two conditions may then arise: 1) This
is the last string on T
b
; or 2) There is just one more string left.
In the first an S sentinel block will be read on Tb and will be
transferred to Since 92 is set, control is shunted to the end of
tape operations. If is set this was the last data pass and the data is
completely Therefore Z sentinels are written on To and all tapes are
rewound. If is is at least one more data pass. In this
'case, the i.nput item counters are reset. -{ 0 : C determines which is the
f;
current output tape. If this were T
c
' for is set to the differ-
ence between the length of a complete output string (2R) and the actual length
of the last output string. Since it is impossible to have a partial string on
back of both output is reset to zerOG The end of tape and end of
string connectors are reset, and then the servos are interchanged so that the
output tapes for this merge are the input tapes for the next. At @ the
contents of E and F are interchanged, since the type of merge (ascending or
descending) will be The variable connector serves to alternate
the setting of <D At @t the input string length counter, R, is doubled
since the output strings of the just completed merge are now the input strings.
If the new input "string length is greater than or equal to half the number of
input blocks, the next merge is the last, in which case the tape label is re-
corded on the next output tape and set to write Z sentinels on T@ at the
end of the next pass.
If this is not the last merge pass
9
then the new output tapes must be posi-
tioned past the two sentinel blocks. Since the two-standby block reading
scheme is employed 9 the reading heads must be at the front of Ta and Tb.
-7-
But, because of the slightly different rates of acceleration and deceleration
for reading backward and writing forward, it is necessary to read one block
forward if writing is to be done following a backward read. This is the pur-
pose of rewinding Te before recording the ID block. Thus, instead of re-re-
cording sentinE!l blocks on the new output tapes, the tapes are repositioned
with their read-write heads forward of the sentinels.
The variable connector CB) is employed only in the exceptional case when the
first merge is an ascending one and we were consequently selecting from
the top to the bottom of the input If this were the case, was
set, and control is tranSferred. to to reset counters to their normal
form" In eithe'r case, we eventually return to Q5V where the next pass is
startedo
This explains the case where the last strings on Ta and Tb were reached at
the same time" In the case where the end of and the end of the current,
but not the B string is reached, there is one ing B string to
merge with no A stringo Reaching the end of Ta causes G to be
set as in the first condition described. After the last lock of toe B string
is processed, a ck be read into position B. SiB R,
control is transferred Q2) where y29 previously set, sends us the
end of string operationso From there control is transferred to to begin
merging the new B string with the supposed A string. The t or the = in Ag
forces all the B items onto the output the Tb sentinel block is
obtained" Then is transferred to and since G2 is still set
and from there to 9 the end of tape operationso
2... Two fa!. .Eerformi!29, Collation
Two methods in common use will be described as means for sorting internally
a block of items into an ascending sequenceo The first of these methods is
a general one and is nearly identical with the external collation technique.
The flow chart'attached displays this method as adapted for two-word items o
The essence of the method consists of dividing the input data exactly in half
and then considering it as composed of one-item strings and finally merging
the one-item strings of each half together to form two-item strings. These
two-item strings are then merged to form four-item strings and the process
continues in like manner until all items are in single sorted stringo
The problem of partial strings which is encountered in external collation
is easily avoided here since there are 30 two-word items per block and, as
32 = 2
5
, we can simply attach two = sentinel items to the end of our block
when it is in the memory and sort 32 items o The sentinels will end up at
the end of the sorted items and may then be omitted when the block is
written on the output tapeso This is a feasible scheme in internal colla-
tionS! since the amount of time required in handling the two dummy items is
relatively smallo It should also be noted that it is not necessary to
-8-
change the merging from ascending to descending at the end of each pass since
the front of the merged items is just as readily accessible as the end of the
items. This, of course, is not the case in external collation where the first
items merged are separated 2 minutes (tape rewind time for t tape) in time
from the last i.tems mergedo Since the last items merged in external sorting
are immediately available to the computer, efficiency dictates that the tapes
be read backwards.
The block of 30 A items to be sorted are read into the memory and a pair of
= sentinels are positioned below the last item
9
making a total of 32 items.


- -
- 10 .-
Anyone of the first 16 items is labeled 'Ai where i = 1, 2, , 16. The
last 16 items are also labeled A, but anyone of these is indicated by Aj
where j = 17, 18,0.09 320 inpgt string counters are fla and llb, while
the string length is R. At Al (the superscript 0 designates the key
word of the item) is compared with AX7 the smallest being stored as the
output item C c' The string counter I:l a or b is increased (which one de-
pends on item was the smallest) and tested against the string length
R. When an input string is exhausted, variable connectors are set to force
into C the remaining the other string" At the end of both strings
control i.s transferred to where i is tested against 17. If the test
indicates inequality, there are more strings to merge and control is re-
turned to 00 If i = l7
g
then the last string in each set of 16 items
has merged., The input string is examined. If R = 16, then a
32-item string has just been produced and C contains the sorted A blocke
If this is not the case
9
R is doubled and i, j9 k are reset and the 32 C
items become a new set of 32 A itemso Ai and Aj are defined as above and
the merging is repeated"
It is apparent that the method just described is a perfectly general one
and is adaptable to any item size" It is, however, a lengthy one, and
the following method is generally used for larger item sizes" The method
is illustrated by coding it for ten-word items. The input block is assumed
to be in 700-759 and the sorted items (ascending sequence)will be stored in
cells 640=699.,
The basis of the method is quite simplee The first item of the block is
selected and compared with each of the remaining items of the block" The
routine assumes that the input block is already in ascending sequence and
so assumes Al be placed in 640
9
A2 in A6 in 690, unless
-9-

lIa=lIb = 0
R=i=k=1
j = 17 . ell
8 .! A,_lIS! .@ .! ,+1-, H'b"-:-+--("',::1-
0--{E)
'lwS_Ckl



---0
10
otherwise instructed. The first item, AI' is selected and compared to all the
remaining items, Ai., The routine then aads ten to the assumed output destina-
tion of Al for every Ai which is found to be less than or equal to AI. At the
same time ten is subtracted from the assumed destination of each such Ai-
After testing all Ai against AI' the address of the output item, in which Al
is to be stored? is For example, if the sequence of the input item
codes is
Item Code
Al
176
A2
21
A3
87
A4
9
A5
930
A6
21
Al will be stored in cell 680. A2 is then compared with the remaining Ai
(i = 3, 4, 6) in a similar maDner. Thus, is stored in 660.
The following coding example is the heart of the Internal Sort for
items:
020 L 700
021 B 710
022 B 070
023 C 070
024 A 061
025 B 720
Input Block
Output Block
T 025
S 061
B 041
C 041
T 029
---------
026 B 071
S 061
027 C 071
B 041
700, ,759
640, ,699
-11-
028 A 061 C 041
029 B 730 T 033
._--
- -
030 B 072
031 C 072
032 A 061
033 B 740
--
- ---
034 B 073
035 C 073
036 A 061
037 B 750
038 B 074
039 C 074
040 A 061
041 [Y 700
042 B40070
043 C 041
044 A 062
S 061
B 041
C 041
T 037
S 061
B 041
C 041
T 041
S 061
B 041
C 041
Z 649
L 71cl
B 046
G 046
045 B 042 A* 064
046 C 042 U 021
061 000000 000010
062 000000 000004
-12-
063
064 010001 000010
.
070
y
710
Z
071 Y 720
Z
072 Y 730
Z 670
07:? Y 740
Z 680
074 Y 750
Z 690
-13-
11 = l1a=
a -I, b=2, c=3, d=14, k=1
m=l, r=14,9=C,
.a
l
l .Yi .8,. V, "E, .11j
.7] I 8, ..cp,
26 ) I


.
T 5 WRONG
NT

La WRONG

RWD* T5 H rl - A f-----..0
PLATE I
l
-
~
-
~
~
-
-
>" ~ ro ro
C\J
UJ
I-
~
-
a..
-
~
t
I"
CD
lID
....J
~
CD
.0
I-
CD
l
::
Ul
::;=
t
<.-
t
.0
I-
ID
1< VI
l
I-
l
"
<
8
~
I
~ ---{I)
N . ~ .
~ G-----
~ . ~ @.A
PLATE 3
16
RWO* T
a
, Tb,ToTd "OUTPUT ON Tel" - SCP
~
2R-!J.c_!J.a
2R -!J.c-!J.b
~ ! b-w.s,H I d-bH !w.s l - d H E-w.sIH F- E Hw.s
,
- F ~
~
~
PLATE IJ
UJ
I-
~
Q..
SERVICE ROUTINES
Ie Introduction
The service are designed to be a practical aid to the programmer
and to improve his efficiency in the use of the computer. The routines of
interest here fall into two catagories, correctors and diagnostic routines.
The corrector routines are VIII,' which is a manual corrector operat-
ing from the Supervisory Control, and AC-2, AC-3, and AC-4, which are auto-
matic correctors proceeding from pseudo-instructions on tape.
In the class of diagnostic routines we have the Codedit, Codecheck, and the
Analyser which prepare, in edited form, the input coding and also to some
extent an analysis of the codingo In addition, there exist a series of rou-
tines known as the Automonitor Routines which give a detailed account of
the actual running of a program.
These routines, with the addition of certain more specialized service rou-
tines and some of the engineering test routines make up the service routines'
tape which is generally available at every computer. Proper use of the serv-
ice routines will nlaterially decrease the amount of programmer and computer
time spent on program checks and revisions e
20 1Qcator
The Locator is always the first routine on the service routines tape. Its
function is to read into the computer any of the routines which follow it.
After an initial is performed, the Locator prints out WHAT BLOCK and
sets up a 10 instructiono Type in a word in the form 000 000 OOB where
BBBB is the block humber of the routine desired. The Locator will in effect
perform an initial read on the routine desired and stop prior to transfer-
ring control to line 000 of that routineo To proceed with the desired rou-
tine
9
merely hit the start
If the block number of the desired routine is not known, depress breakpoint
a at the start of the routine. Force transfer when QO sets up in the static
register. The SCP (set at normal) will print out the block number of each
service routine on the tapeQ
The service tape is normally mounted on servo one. If it is used
on any other servo
g1
force transfer on breakpoint one after the initial read.
The SCP will print out 9 SERVO UOOOOOe Type in the servo number of the
servo used in the form SSSSSSSSSSSS and the Locator will make the necessary
modifications in itself and in the particular service routine desired.
THE CORRECTORS
3. Ylll,
It, is sometimes nec:essary for a programmer to assemble a tape from a group of
tapes or make on an existing tapee These operations occur most
frequently during the ftdebuggingU stage of programming. In order to avoid a
great deal of duplicated effort and to simplify these operations, we make use
of the MARK service routine o
The MARK VIII routine operates from control words typed in from the Supervi-
sory Control. Its basic operation is to copy a tape or a portion 6f a tape
onto another tape.*'
Two types of manual control are exercised by the programmer. These are Break-
point Options (BKPr), and type-ins as set up by the routine.
Breakpgint Options
To use these options depress the appropriate breakpoint selector switch and
force transfer when the computer stops on that breakpoint number.
After the reading of the MARK VIII instructions or at any time after a Clear C
operation, there are two breakpoint options available:
- Change .Q.f. Servos: MARK VIII (except when the special methods, de ...
scribed below are used) normally copies from servo 2 to servo 3. To use any
other servos, force transfer on BKPT 10 The SCPwill print 9servoU00044 and
set up a 10 m instruction. Type in a word of the form IIIIIIOOOOOO where I is
the input servo number and 0 is the output servo number.
BKPT 1 - Change 2f Write To change the write density from IOO/inch to
20/inch or from 20/inch to 100 inch, force transfer on BKPT 7. The write den-
sity is not changed by the Clear C operation. After the routine changes the
write density, it prints out the new density as
730100A00036
530100A00036
or
20!inch
100/inch
*NOTE: The procedure for correcting a tape on Univac requires that the entire
tape be recopied inasmuch as a new block inserted in the midst of other
'information will render the following block unreadable due to the posi-
tion of the erase head and the differences in the servois read and write
acceleration times Q
-3-
Control
After breakpoint options, if any, have been exercised, the SCP prints out
.6blockillmitE) At this point there are the following operations which may
be executed at the programmerOs discretion:
a e Skip through a tape
b" Copy a tape
Correct a tape
de Merge tapes
Skipping Through A Tape: There are two methods available, special and
regular.
Special Metho<i,.. After the SCP types out t
f
type in a
word'as YSOOOOOOBBBB, where S is the servo number and BBBB is the
number of blocks to be skipped.
_Regular Methoci. After the SCP types t, type in a
word as OOOOOOOOBBBB, the SCP will then print out 83next90rder. Type
in SSSSSSSSSSSS and the operation will be executed., When either of
these methods is employed the first word of every skipped block is
printed out on the SCP. (A switch on the Supervisory Control may be
set to skip these printouts.) In either case the control is trans-
ferred to the beginning of MARK VIII againo
Copying oS. There are two methods available, special and regular.
Special After o6blockfllmit appears on the sep, type, in
where X is the input servo, Y is the output servo, and
BBBB is the number of blocks to be copied",
Regular Methodoo After Q6block&lmit'9 type in OOOOOOOOBBBB,where BBBB
is the number of blocks on the tape to be copiede The SCP will then
print out 83next90rder!j) and set up a type in. Type in ZZZZZZZZZZZZ.
Correcting Tape.
After is printed, type in 00000000BB8B, where BBBB is the
number of blocks on the tape to be copied. The SCP will then print out
83next90rder. Type in the block and word number of the first word to be
corrected as follows: 00BBBBOOOOWW
9
where BBBB is the block number and WW
is the word number
o
(The first block is 0001, but the first word in the
block is 00
0
) The SCP will print out the old word and set up a type in.
-4-
Type in the new w()rd. The SCP will then again type out 83next90rder. Type
in the next block and word number and proceed as above until all the correc-
tions have been entered. After the last correction, type in ZZZZZZZZZZZZ
instead of a block and word number, and the routine will then complete the
copying operation and proceed to the ending routine. The corrections must
be in sequence by block number, but not necessarily by word number.
Merging Tapes
Tape merging is executed by a series of the above operations, taking advan-
tage of the fact that it is not necessary to rewind the tapes.
Ending Routine
Upon completion of the copy or correct operations, the computer prints
To rewind the input tape, hit the start bar once. (To avoid
rewinding the input set up a skip instruction in the static register when
the rewind instruction appears on the next timee) The computer will stop
with a U 044 set up in the static register. Hit the start bar once to read
the output tape bclckward a number of blocks equal to the block limi t ..l!ll
typed in. To change this block limit, skip the U 044 and type in, when
called for, the number of blocks desired to be read back as 000000008B8B.
The SCP will print the first word of every block read backwards. Upon com-
pletion of the backward reads, the computer will rewind the output tape.
When using the MARK VIII routine, the programmer should note that at any
time a Clear C operation starts the routine anew, except for changes in
write density and servo. Should manual intervention be necessary, it is
useful to know thalt the input block in the memory is at 100 - 1590
",,5-
FROM LOCATOR
FROM CLEAR C
B,W ORDER
SPECIAL ORDER
CHANGE SERVOS
SKIP ORDER
COpy ORDER
CORRECT ORDER
COpy B-1 BLOCKS FROM
TYPE IN BLOCK LIMIT N
SKIP ORDER
SKIP H BLOCKS ON INPUT TAPE, PRINT-
ING THE FIRST WORD OF EACH BLOCK
COpy ORDER
COPY TO N BLOCKS FROM THE
INPUT TO THE OUTPUT TAPE
Rwd* INPUT
READ OUTPUT TAPE BACK N BLOCKS, PRINT-
ING THE FIRST WORD OF EACH BLOCK
SKIP SR
TYPE IN BLOCK LIMIT N
Rwd* OUTPUT
INPUT TO OUTPUT TAPE I -I
PRINT WORD W
OF BLOCK B
TYPE IN CORRECTION FOR
WORD W OF BLOCK B
FLCM CHART MK V II I
000 [61 000
001 L 003
002 R 060
003 Yii030
004 F
005 50 012
006 B 080
007 R
008 R 086
009 50 041
010 B 08:2
011 L 042
012 .6BLOC
013 B 096
014 R 098
R 000 ]
B 001
Ql 051
Z 090
Q7 048
10 080
T 025
U 094
F 055
10082
T 032'
T 021

Q 017
U 097
CODING .EQB. ll!l
Rewind Service Routine Tape
000000 U 000
Force Transfer for Servo Change
]
Tr.ansfer Read and Write Orders to
Working Storage
Force Transfer to Change Write Density
.6BLOC Kfl LMIT -7 SCP
000000 OOXXXX --+ 080 (Block Limit)
Transfer Control if Special Order
Transfer Control to Read First Block
83NEXT SCP
Next Order 082
Transfer Control if this is a Copy Order
Transfer Control if this is a Skip Order
Transfer Control if this is the Incorrect
Block
Transfer Control to Write Current Block
-7..,
U 094
016 H 096
U 013
017 L
B 082
018 T 000
;6 037
019 E 082
A 035
020
U 085
021 B 036
L 080
022 50 100
Q 000
023 R 095
U 094
024 A 036
U 022
025 H 082
05 000
026 A 080
C 084
027 E 080
H 080
028 R 0601
U 052
029 YZZ030
Z 090
030 R 095
U 094
Transfer Control to Read Next Block
Increase Block Counter
Transfer Control if this is the Last Block
Fabricated Corrector
First Word of Input SCP
Transfer Control if Last Block
Transfer Control to Read Next Block
Increase Block Counter
Prepare Servo Change Control
Prepare Block Limit
Transfer Control to Servo Change Routine
Transfer Read, Write, and Block Control
to WS
Transfer Control to Read First Block
-8-
031 L 029
032 L 080
033
034
[73 100
[ 12 000
035 50 lOa
036 000000
037 [ 53 100
038 50 038
039 [82 000
040 101111
041 83NEXT
042 9SERVO
043 50 012
044 B 080
045 23 000
046 50 100
047 S 036
U 010
U 013
A 036 ]
30 100]
10 100
000001
A 036 ]
900END
U 042 ]
111111
9CRDER
U 044
10 080
L 036
30 100
Q 041
U 045
Transfer Control to Copy Test
Transfer Control to Finish Copying
Write Last Block
500038 900END ~ SCP
Stop
Rewind Input Tape
Transfer Control to Test Output Tape Option
,,6BLOC KL1 LMIT --+ SCP
000000 OOXXXX --+-. 080 (Block Limit)
First Word of Input Block---. SCP
Transfer Control if Last Block
Decrease Block Counter
-9-
048 B 037
F 033
049 G 037
C 033
050 50 037 Present Write Density ---+ SCP
U 003
051 50 0 4 ~ ~ 9SERVO U 00044 --+ SCP
10 084 IIIIII 000000 -'084 (Input-Output Servos)
052 F 040
B 084
053 E 034
H 034
054 E 039
Ii
039
055 B 084
J61111
056 E 037'
H 037
057 E 03:::1
H 033
058 E 0'41
H 04J.
059 E 04:1
H 045
-10-
EXAMPLE PROBLEM)' f.Qi .MMi!S V I I I
Problem 1=
BJlks 1 through 15 of tape 0
Blks 13 through 17 of tape B --+ tape D
Blks 1 through 5 of tape tape 0, correcting
Blk 1 word 07
Blk 3 word 05
Skip through 0 to test its readability
Solution
Assume tapes A, B, C, D are mounted on servos 2, 3, 4, and 5 and that the MARK
VIII is in the memory" The operatoris actions are:
Ie Type in Z25 000 000 015
Hit start bar to rewind tape A
3. Clear Cand hit start bar
4. Type in Y30 000 000 012
5"
Type in '235 000 bOO 005
60 Hit start bar to rewind tape B
7"
Set brkpQt I, clear C, and hit start bar
8. Force transfer, release brkpUt 1, and hit start bar
90
Type in 444 444 555 555
lOa, Type in 000 000 000 005
11. Type in 000 001 000 007
12. Type in correction for blk 1 word 07 of tape C
13., Type in 000 003 000 005
-11-'
14. Type in correc:tion for blk 3 word 05 of tape C
15. Type in ZZZ ZZZ ZZZ 'ZZZ
16... Hi t start bar to rewind tape C,
170 Clear SR to and hit start bar
1841 Type in 000 000 000 025
Problem .2,=
Blks 1 through 2 of tape A C
Blks 1 through 5 of tape C
Blk 3 of tape C
Blks; 8 through 19 of tape B C
Skip through C t'o test its readability and copy at 20!inch
for
Solution
Assume A, B, and two blanks are on servos 2, 3, 4, and 5. The operator's actions
are:
leo Type in 224 000 000 002
2. Clear C and hit start bar
3e Type .in 234 000 000 005
40 Clear C and hit start bar
5('1 Type in Z24 000 000 001
6
0 Hit start bar to rewind tape A
7. Clear C and hit start bar
8. Type in Y30 000 000 002
-12-
9. Type in 234 000 000 012
10. Hit start bar to rewind tape B
11. Clear SH to zeros and hit start bar
12. Type in 000 000 000 020
13. Clear C, set brkp't 7, drop interlock on T4, hit start bar
14. Force transfer, release brkp't 7, and hit start bar
15. Type in 245 000 000 020
16. Hit start bar
17. Hit start bar
-13-
4. AC-3
Autocorrector 3 is a single purpose routine designed to perform tape correction
only. AC-3 has several features which serve in reducing the possibility of
clerical errors as well as programmer and computer time ..
Unlike "MARK VIII, which operates from Supervisory Control type-ins, the AC-3
is tape controlled. Thus, when only tape correction is to be done and when a
large number of corrections are to be made, the AC-3 is more conservative of
Univac time than the MARK VIII. In addition to the value of tape controlled
operation, the AC-3 has several other features designed as a programmer con-
venience. They are:
a. The corrections need not be recorded on the control
tape in block number sequence as Ae-3 sorts the
corrections before they are applied"
b. Before each correction is actually applied, it can
ei ther be printed on the SCP or wri tterhon a tape,
providing a permanent record of the corrections is
made ..
c. In conjunction with item b, options are available
for making minor corrections to the control tape
information"
The control tape which is mounted on Uniservo 4 consists of a series of two-
word items in the following format:
OOB BBB 000 OWW
eee cee eec eee
Where BBBB is the number of the block to be corrected .( 1 ~ " B B B B ~ 2000) ,
WW is the word within that block to be corrected (00 < ww < 59)j and
eee eee cee eee is the correction for that word.. For example, if word 51
of block 27 is to be changed from BOO 301 COO 397 to BOO 351 HOO 397 the
correction item would be
000 027 000 051
BOO 351 HOO 397
A maximum of 329 correction items can be handled at one time with Ae-3111 To
mark the end of the correction list on tape 4, the sentinel ZZZ ZZZZZZ ZZZ
must follow the last correction item.. The remaining information following
the sentinel in the block is ignored by AC-3.
The operating instructions for Ae-3 are listed below.
1.. The tapes to be mounted are as follows
Uniservo
1
2
3
4
5
-14-
Tape
Service routine containing Ae-3
Tape to be corrected (see 3)
Blank tape for corrected output
(see 3)
Ae-3 control tape
Blank tape for the correction
listing if option under step 3
is to be exercised
2. Initial read Ac-3 through use of Service Routine Locator. Computer
will stop after first block of AC-3 is in the memory,
3. Several conditional transfer breakpoint options are now available:
a. To print the AC-3 operating instructions depress breakpoint
zero, set the SCP selector switch to normal, and operate the
start baro
The computer will stop on a QO. Force transfer and operate
the start bar0 The AC-3 operating instructions will print
on SCP and the computer will stop again on QO. At this
point QO should be released and other options, if they are
to be exercised, should have their appropriate breakpoint
buttons depressed. Operate start bar to continueo
b. To modify the Uniservos assigned to the tape to be corrected
and the blank to contain the corrected version depress break-
point one and operate the start bar.
The computer will stop on a Ql. Release breakpoint one,
force transfer and operate the start bar. Computer will
print
In Out Servo
and stall on an input ready. Type in the new Uniservo
i'designations in the following form
xxx xxx yyy yyy
where X is the input servo and Y is the output servo. The
computer will automatically modify the appropriate instruc-
tions and continue.
Co To have the correction list written on tape 5 depress break-
point 3 and operate the start bar. This option will be ex-
ercised when step 5 is reachedo
4. The computer will then print out
block limit
and stall on an input ready. Type in the number of blocks on the
"tape to be corrected" in this format
000 000 OOX XXX (1 < XXXX < 2000)
5. If breakpoint 3 has been depressed, computer will stop on Q3. Re-
lease breakpoint 3 and force transfero If the correction list is
known to be correct (e.g. no errors in unityping), the SCP output
selector switch should be placed on skip to conServe computer time.
Operate the start bar to continue with AC-3.
If an SCP output is desired, the print selector switch should be
set to check and the margins set for 39 digits.
-15-
6. When the corrections have been applied, the computer will rewind all
tapes and stop.
7. If the option for writing the corrections on tape 5 has been
cised, the uniprinter settings should be
a. Print selector switch set to check
b. Margins set for 36 digits
Occasionally it may be desirable to modify the correction information on the
control tape without the necessity of retyping it or doing a MARK VIII correc-
tion. This, of course, could arise when an error in listing the corrections
has been made or in unitypingG To expediate this correction of the control
tape, a set of additional breakpoint options have been provided. These break-
points, if option is to be exercised, should be set on step 3 above.
To modify a correction, depress the appropriate breakpoint listed below.
a. To modify the block and word number of a
correction depress breakpoint four.
b. To modify the correction itself depress
breakpoint five.
c. To skip this correction entirely depress
breakpoint six.
When the computer is about to apply a correction, it will print the old word,
the block and word number, and the new word, and will then stop on the appro-
priate breakpoint. Operate the start bar each time the computer stops on the
breakpoint until the correction to be modified is printed out. Then for mod-
ifications a and b above, force transfer, release the breakpoint and operate
the start bar. The computer will stall on an input ready. Then
a. To modify the block and word number, type in the
new block and word number in the same format as
on the control tape. This new block number must
not be less than the replaced block number nor
larger than the block number of the correction
item to follow (this may require scanning the
list of corrections manually).
b. To change the correction itself, type in the new
correction. If both a block and word number and
its correction are to be changed, both breakpoints
four and five must be used.
For skipping a correction simply force transfer, release the breakpoint and
operate the start bar.
AC-3 may be rerun at any time by operating the clear C switch. This causes
the input, output, and control tapes to be rewounde Tape 5 will not be rewound.
A schematic chart of AC-3 is appended.
-16-
FROM LOCATOR
-.
FROM CLEAR C ~
PRINT
OPERATING
INSTRUCTIONS
IS THIS CORRECTION
A SENTINEL?
NO
SELECT CORRECTION
WITH LOWEST REMAINING
BLOCK NUMBER B
CHANGE
INPUT-OUTPUT 1-1 __ ~
SERVOS
COpy TO N BLOCKS FROM
THE I HPUT TO THE OUTPUT TAPE
COPY TO B-1 BLOCKS FROM
THE I NPUl TO THE OUTPUT TAPE
WRITE ON T5 OLD WORD,
BLOCK AND WORD HUMBER,
AND HEW WORD
TYPE IN BLOCK LIMIT N
READ OUTPUT TAPE
BACK N BLOCKS
PRINT OLD WORD
BLOCK AND WORD
NUMBER AND
NEW WORD
FT
READ ALL
CORRECT! ONS
INTO
MEMORY
SELECT CORRECTION
WITH LOWEST
BLOCK NUMBER B
~
RWO* CONTROL
HPE
TYPE IN NEW
BLOCK AND
WORD NUMBER
RWD INPUT AND
OUTPUT TAPES
TYPE IN
NEW
CORRECTION
MAKE CORRECTION
FL,;W CdABT AC-3
SERVICE ROUTINES
003
Like the Autocorrector 4 is a single purpose routine designed to per-
form tape correction onlYQ The distinguishing feature of AC-4 over the
Mark VIII and AG-3 is that it provides a check on the corrections to be ef-
fected o As in AC-3
9
the correction items are recorded on a control tape
and thus computer time is conservedo
The control tape consists of a series of three-word items having the follow-
irAg formatg
OOB BBB 000 OWW
xxx xxx XXX xxx
ece ece eec GGe
Where BBBB is the number of the block to be corrected (1.s BBBB < 2000), WN
is the word wi thin the block to be cox'rected (00 < WN 59) 9 XXX ,XXX XXX XXX
is the lncorrect word and Gee occ ece cec is the correct wordo For example,
if word 51 of block. 27 is to be corrected from BOO 301 COO 397 to BOO 351
HOC 397 the correction item would beg
000 027 000 051
BOO 301 COO 397
BOO 351 HOO 397
Following the last item will be the sent:lnel word ZZZ ZZl Z'ZZ ZZZo
The remaining information in the correction data block will be ingnored o
The (corl'action i terns must be rceeorded on the control tape in ascendi.ng order
by block numbero
BrieflY9 the mode of operation of AC=4 is thi.SJ8 the tape to be corrected is
copied onto '0 blank output tape until an incorrrect block is located Cil The
word to be corrected is then compared wIth the word the programmer expected
to be there
g
as indicated by the second word of the correction itemo If they
agxee
g
the correction is made by replaclng that word of the input tape by the
third word of the correction item and the is; X"epe:a:ted for the next
correction itemo If the comparison al disagreement, it may indicate
ei an error in listing the block and word number or in the unltypi.ng of
..,18-
where X is the input servo and Y is the output ser-
vOo The computer will automatically modify the ap-
propriate instructions and continue
o
4. The computer will then prin.t
Block Limit
and stall on an input readyo Type in the number' of blocks
on the "tape to be corrected" in this
000 000 OOX xxx (1 < XXXX < 2000)
50 The computer will then proceed with corrections
9
comparing
word two of the correction item with the word in the desig-
na.ted location before the correction is made
o
Step 8
11
below,
will cover the operation when discrepancies are detected o
60 When all correcti.ons have been applled\-) the input tapes will
be' rewound o The output tape will be read backwards to check
its legibility and then rewound o
70 AC-4 may be restarted at any by clearing Co
8
0
If, during the application the correction list
9
a discrep-
ancy is found between the second word of the correction item
and the word in the designated location
9
the following infor-
mation will print on SCP and the computer will stop:
Block and word number
The word occupying that locatlon
The word expected to be in that location
The word to be put in that locati.on
If the AC..,4 operating inst.ructions are desired\} depress QO
Sind operate the Start Bare The rcomputer will stop on a QOo
Force transfer
g
release the breakpoint and operate the
Start Bare The operating instructions will be printed on
and the computer will again stopo
Se!veral options are now g
Ao To disregard the discrepancy sod proceed with the
correction
9
depress breakpoint two
9
and operate the
Start Baro The computer will stop on a Q20 Force
transfer, release the breakpoint and operate the
Start Bare
B. To change the block or word number, depress breakpoint
three and operate the Start Bare The computer will
stop on a Q3. Force transfer, release. the breakpoint,
and operate the Start Bar. The computer will print on
SCP
New Blk, Word
and stall on an input readyo Type in the new block
and word numbere This new block number must not be
less than the old onenor greater than the block n u ~
ber for the next correction item. The computer will
apply this new correction and continue in normal
operation.
c. To skip this correction
9
simply operate the Start Bar.
-21-
I'\)
I'\)
FROM
LOCATOR
FROM
CLEAR C
SELECT NEXT COR-
RECTION ITEM FOR
BLOCK B WORD W
COpy TO BLOCK
N FROM TH E INPUT
TO THE OUTPUT
RWD* CONTROL
AND INPUT
TAPES
IGNORE
DISCREPANCY
II
COPY TO BLOCK B-1
FROM THE INPUT I .. I
TO THE OUTPUT
N
SKIP THIS
CORRECTION
TO CORRECT
B AltO W
TYPE IN NEW BLOCK AND
WORD NUMBER, BW
TYPE IN
BLOCK LIMIT N
MAKE
CORRECTION
2
2
FLOW CHART
FOR AC-ll
6. MA-2
Occasionally, it is desirable to merge the data from two tapes on a word-by-
word basis as contrasted to the block merging accomplished by the Mark VIII.
Examples of this might be:
a. copying a subroutine into a master routine where the
subroutine does not take up an intergral number of
blocks o
b./correcting tapes on which words have been left out
or extra words addedo
c. merging data from two tapes where block-by-block
merge is not feasible.
The Automatic Line Merge, MA-29 is a service routine designed for this pur-
pose. Certain other features have been included in MA-2 to increase its
flexibility and usefulness.
The basic operation of MA-2 is to transfer words from one or two input tapes
designated A and B to an output tape Ce The uniservos for A, B, and Care
specified by an :initial control word. This contror' word and the control words
following it are normally unityped onto a control tape which is mounted on
although a breakpoint option may be exercised to use another
to have the control words typed in manually through SCKo Other
breakpoint options are available for minor corrections of the control words
which are printed.on the-SCP before they are executedo
There are three different types of control words which MA-2 recognizeso The
first of these must be the first control word received by MA-2. It is called
an initial control word and has the following format
ABC 000 000 000
where A and B are the two input uniservo numbers (use zero for B if there is
only one input) and C the output uniservoe An initial instruction causes the
MA-2 tape instructions to be appropriately adjusted.
The second type of control word is a pseudo instruction instructing MA-2 to
perform one of the five following operations:
1. N words from tape A or B to
2. Skip N words on tape A or Bo
3. fill tape C with'N tOspecialu wordse
4. Add the next N words from the control tape
(.or SCK) to tape C.
5. Read tape C backwards to test its readability,
also, print the total number of blocks on the
tape, the first word of each block, and then
rewind the tape with interlocko
-23=
These pseudo instructions are in the following format
C is a control digit specifying one of the five operations noted above.
In particular if
c == Z
A copy operation is to be performed
C :: Y A skip operation is to be performed
C :: F A fill operation is to be performed
C = A An add operation is to be performed
C = B
A read back operation is to be performed
Si is the inputtape and it will be either the letter A or B for those
operations requiring an input tape.
So is the output tape. It is always the letter C for those operations
requiring an output tape.
T is of significance only when a fill operation is to be done. In
these cases it will be a digit specifying the kind of "special" word
to be used in the fill operation. If
T == a
T == 1
T = 2
The fill word is 000 000 000 000
The fill word is tit ttt ttl itt
The fill word is ZZZ ZZZ ZZZ ZZz
as many as seven other fill words may be by inerting the
additional fill words in words 13 to 19 of block 3 of MA-2. If this
is done, T has the range 0 < T < 9. Nei ther Si nor So need be spec-
ified on a fill operation.
Together specify the number of words affected by the pseudo instruc-
tion.' If the number of words to be affected are less than 60,
BBBB = 0 ,and 00 < LL :5. 59, while if the number 'of words are 60 or
more BBBB = number of 60 word multiples and LL = remaining words.
For example, ,if the number words, N,
N == 15
N == 75
N = 120
Then LL == 15 and BBBB :: 0000
Then LL =. 15 and BBBB := 0001
Then LL = 00 and BBBB ::: 0002
The third type of control word is the normal Univac instruction. That is,
if a control word (exclusive of the initial control word which is the servo
designator) does not have as its left-most digit the letter Z, Y, F, A, or
B it is treated as a Univac instruction and executed. In particular a
control word
620 000 900 000
-24-
will cause tape 2 to be rewound and Univac stopped. Do not expect the con-
tents of the working registers to remain unaltered between the execution of
successive words. This means a control word sequence
LOO 500 BOO 900
000 000 QOO 051
will not produce the expected result of comparing (500) and (900) for equali-
ty.
The operating instructions for MA-2 are:
1. Tapes to be mounted are -as follows
Uniservo
1
Tape
Service routine containing MA-2
Control tape (if options under
steps 3b and c are not exercised)
Input tape A
4
A
B
c
Input tape B (if a second input is
used)
Blank tape for output
2. Initial read MA-2 through use of Service Routine Locator.
Computer will stop after the first block of MA-2 is in the
memory. Depress breakpoint seveno
3. Several conditional transfer breakpoint options are now
available:
a. To print the MA-2 operating instructions
depress breakpoint zerop set the SCP se-
lector switch to normal, and operate the
Start Bar.
The computer, will stop on a QOo Force
transfer and operate the Start Biro The
MA-2 operating instructions will print on
SCP and the computer will again stop on
QO. At this point QO should be released
and other options
9
if they are to be ex-
ercised, should have their appropriate
breakpoint buttons depressedo Operate
the Start Bar to continue.
b. To modify the uniservo assigned to the
control which is normally #4, de-
press breakpoint one and operate the Start
Bar.
The computer will stop on a Ql. Force
transfer, release breakpoint one, and
... 25-
operate the Start Bar. The computer will print
on SCp,
TTTTTTTTTTTT
and stall on an input readyo Type in the new
uniservo designation for the control tape in
this form:
'XXXXXXXXXXXX
where X is the uniservo number. The computer
will automatically modify the appropriate i n ~
structions and continue.
c. To modify MA-2 for manual operation through
the SCK, depress brekkpoint three and operate
the S tart Bar.
The computer will stop on a Q3. Force transfer,
release breakpoint three and operate the Start
Bar. The manual operation of MA-2 is described
in step 4b.
4. The computer will then begin the word-by-word merging in the
following manner, depending on whether,or not the manual op-
tion of 3c was exercised:
a. If the control tape is used, the merging pro-
ceeds automatically with MA-2 selecting its
instructions from the control tape o Just be-
fore each order is executed,it is printed on the
SCP thus providing a permanent record of the
cOrrections is madeo The merging i ~ stopped
only by a 90000 instruction on the control
tape.
b., If the manual option has been exercised, the
computer will print on the SCP
ABC
and stall on an input ready$ Type in the
initial control word (servo specifications).
The computer will modify the appropriate tape
instructions and stall 'again on an input readyo
Type in the next instruction. The computer
will continue calling for instructions until a
90 000 instruction is supplied.
5. The routine' may be rerun at any time by clearing C. Oc-
casionally, it may be desirable to modify the instructions
-26-
on the control tape without the necessity of retyping it or
doing a Mark VIII correction. This could, of course, arise
when an error in listing the instructions or in unityping
them has been made. A set of additional breakpoint options
have been incorporated in MA-2 to facilitate the modifica-
tion of the control tapeo These if their op-
tions are to be exercised, should be set on step 3 above:
a. To insert an instruction, depress breakpoint
four. Each time the computer stops on a Q4
operate the Start Bar until the instruction
just printed on SCP is the one in front of
which a new instruction is to be-inserted:
When this occurs, force transfer, release
Q4 (unless further insertions are to be made),
and operate the Start Bar.
The computer will print
INSERT
and stall on an input ready. Type in the
additional order.
b. To change an instruction, depress breakpoint
five. Operate the Start Bar each time the
computer stops on -8 Q5 until the instruction
to be changed has ju.t been printed on SCP.
then force transfer, release Q5 (unle$s
further instruction changes are to be made),
and operate the Start Bar.
The computer will print
CHANGE
and stall on an input ready. Type in the
corrected instruction.
c. TO,delete an instruction, depress breakpoint
six. Operate the Start Bar each time the com-
puter stops on a Q6 until the instruction to
be deleted has just been printed on the SCP.
When this occurs, force transfer, release Q6
(unless further deletions are to be made),
and operate the Start Bar.
The computer will print
DELETE
and pass on to the next instruction.
-27-
Certain instruction errors are automatically detected by MA-2:
If a partially filled output block is still in
the computer at the time a B instruction (read
output tape backwards) is received
p
the com-
puter prints
Te BKS xxxx
OUTPUT BLOCK
NOT FULL
Hyy 5zz UOOIO?
where xxxx is the total number of blocks on C
and zz + 1 is the number of words in the par-
tially filled output blocko The computer will
stop on a Q70 If the remainder of the block is
to be filled with zeros, simply operate the
Start Barolf other corrective action isnec-
essary, force-transfer
9
set one of the break-
points listed in the error conditions noted
and operate the Start The compu-
ter will stop on the selected breakpoint and
the operations discussed in the preyious para-
graph should be followedo
b. If a copy or skip order is supplied to MA.,,2
without listing the input servo
9
the computer
will print
ERROR SET BP
andstopo Set one of the breakpoints 49 5 p
or 6 and operate the St.art Bar o The computer
will stop on the selected bI'eakpoint and the
6petations for that breakpoint
should be followedo
The following examples will illustrate the use of MA-2:
Example 1
Prepare an output tape, C, containing the
following parts of input tapes A and B:
BLK 1 W 00 to BLK 2 W 16
BLK 1 W 10 to BLK 1 W 19
BLK 2 W 17 to BLK 2 W 59
BLK 2 W 00 to BLK 2 W 49
from A
from B
from A
from B
Let tapes A, B9 and C (blank) be mounted on
servos 29 3
9
and 59 respectivelyo The con-
trol tape (or manual type-ins) would then
contain
=28=
Instructions
235 000 000 000
lAC 017 000 001
YBO 010 000 000
ZBC 010 000 000
ZAC 043 000 000
YBO 040 000 000
ZBC 050 000 000
620 000 630 000
BO.o ()OQ.,OOO 000
900 000 000 000
Example 2
Explanation
A =: T 29 B :: T 3 ,C = T 5
Copy 77 words from A to C
Skip 10 words on B
Copy 10 wordsf?;om B to C
Copy 43 words from A to C
Skip 40 words from B to C
Copy 50 words fromB to C
Rewind A and B
Read C back and rewind
Stop
.Tape A-contains 4 blockso The first 6 words are
heading) the remainder is datao Transform the
heading into an identification block filling the
remainder of the block with zeroso Follow this
1D block with the data placing Z sentinels after
the data and add two sentinel blockso
Let tape A be mounted on servo 2 and a bl.nkp C,
be mounted on servo 50 The control tape would
then contain
Instructions Explanation
205 000 000 000
ZAC 006 000 000
FOO 054 000 000
ZAC 054 000 003
FOO 206'000 002
BOO 000 000 000
620 000'900 000
Example 3
A := T2, C :: T5
Copy 6 words from A to C
Fill C with 54 words of zerQ8
Copy 234 words from A to C
Fill C with 126 sentinel words
Read C back and rewind
Rewind A and stop
Tape A is a unityped program tape, 2 blocks in
lengtho In proofreading the unityping it was
found that words 33 and 34 of block 2 were left
out, and everything beyond these word$, there=
fore, has been the last two words be=
ing zeroo These words should be BOO 220 LOO
135 and AOO 235 TOO 120 respectivelyo Tape B
=29=
is a,19 block program tape. Word 12 of block 10
of tapeBis to be corrected to read BOO 110 HOD
132. Tape A is to replace that part of tape B
from word 20 of block 15 to word 19 of block 17.
In addition, a block of ignores is to be
to the completed program tapeo
Let tapes A and B be mounted on servos 2 and 3 and
a blank tape, C, be mounted on 30 The control tape
contents are:
Instructions
235 000 000 000
ZBC012 000 010
YBO 001 000 000,
ADO 001 000 000
BOO 110 HOO 132
ZBC 007 000 005
YBO 000 000,002
ZAC 033 000'001
AOO 002 OOOiOOO
BOO 220 LaO', 135
ADO 235 TOO 120
ZAC 025 000 000
ZBC 040 000,. 002
Faa 100 000 001
620 000 630 000
BOO 000 000 000
900 000 000 000
7. HERB I
Explanation
A T 2' B:: T39 C =: T5
Copy 612 words from B to C
Skip the word to be corrected
Add 1 word .to C
Word to be added
Copy 307 words from B to C
Skip 120 words on B
Copy 93 words from A to C
Add 2 words to C
Word to be added
Word to be added
Copy 25 words from A to C
Copy 160 words from B to C
Fill C with 60 ignore words
Rewind A and B
Read C back and rewind
Stop
HERB 1 is a single purpose routine designed to automatically compare for
identity the contents of two tapesQ Whenever the routine finds a discrep-
ancy, the two words and their block and word number is printed on SCPo An
option is provided for producing an output tape that is a synthesis of the
input tapes.
There are a number of computer operations that may warrant the use of an
auto-comparator such as HERB 10 Some of these are listed below:
a. as a check on the manual operations of MK VIII such
as ,copying a tape or correcting a tape.
b. as a means of spotlighting the differences between
two supposed identical routineso
c. as a means of verifying the accuracy of a duplicated
unityping job and at the same time providing a cor-
rected output.
-30-
The opera-ting instructions for HERB I are listed below
1. The service routine tape is mounted on Uniservo 1.
The two tapes to be compared are mounted on servos
A and B (any servos may be used). If an output
tape is desired, mount a tape on servo Co
2. Set the SCP Selector Switch on computer digito Set
the margins for
a) No output tapeg 39 digits
b) Output tape: 52 digits
3. Initial read Herb I through use of the Service
Routine Locator. Computer will stop with the
first block of HERB I in the memory.
4. If the output tape feature is desired, set break-
points 4 and 5.
5. Operate Start Bar.
6. Computer will stall on an input readyo Type in a
control word in the following format:
AAA BBB CGG xxx
where A and B are the uniservo numbers of the two
tapes to be C is the uniservo number of
the desired tape (C 0 if no output tape option
is being used), and xxx is the number of blocks on
the input tapes to be compared.
7. The computer-will begin comparing A and B word-by-
word for identityo If C 0, the routine will copy
each A word passing the identity test onto the out-
put tape. The operation of the routine when a dis-
crepancy between A and B is detected is covered in
step 10.
8. When the two input tapes have been compared the
desired number of blocks, the input and output (if
used) tapes are rewound and the computer stopped.
9. To rerun at any time clear Co This will cause A,
Band C (if used) to be rewound and the routine
will pick up at step 6.
10. In case a discrepancy between A and B is encounter-
ed by the routine
9
three words will be printed on
SCP in the following sequence:
-31-
1. word from tape A
2. block and word number
of the discrepancy
3. word from tape B
if the output tape option is not being used, the
routine returns to the comparison of the remain-
ing words on the input tapes.
If an output tape option is used, the computer
will stop on breakpoint 4 after the printouto
The following options are now available:
a. To allow the A word to appear on the
output tape, operate the Start Bar.
The computer will stop on breakpoint
5. Operate the Start Baro The A
word will again be printed on SCP.
b. To allow the B word to appear on the
output tape, operate the Start Bar.
The computer will stop on breakpoint
5. Force transfer and operate the
Start Bar. The B word will again be
printed on SCP.
c. To allow a new word to appear on the
output tape, ,force transfer and op-
erate the Start Bar. , The computer
will stall on a type in. Type in
the desired word.
After one of the above options have been exercised,
the computer will continue with the comparisons.
-32-
PROGRAMMING MANUAL FOR THE
HIGH-SPEED PRINTER
1. General Characteristics
The Remington Rand High-Speed Printer converts information stored in
the form of magnetic pulses (the Univac XS-3 code) into a visible
printed recordo The speed of conversion is much greater than anything
heretofore commercially availableo Some general characteristic of the
printer are listed below:
Characters printed per line
Lines printed per minute,
optional at
Horizontal character spacing
Vertical character spacing,
optional at
Number of different printable
characters
Printing format control
130
200, 400
Q
or 600
10 per inch
6, 3, or 2 per inch
51
plug board and paper loop
In brief, the High-Speed Printer operates in the following fashiono
Coded data is read from the magnetic tape in groups of 120 digits c a l l ~
ed blockettesu Each blockette is stored in a memory with the data
being in the same coded form as on the tapeo A continuously revolving
shaft, on which are mounted the 130 typewheels (actually 65 double
wheels), also carries a commutator which Causes to be generated se-
quentially the code for each of the characters on the typewheelso
This coded representation is sent to a 120 place comparator which also
receives in the same code the contents of the 120 place memoryo When-
ever agreement occurs, which may, of course, be in many places of the
comparator at once, Signals leave those places on separate wires.
Opposite each typewheel there is a hammer which can be driven against
ito The paper, with an inked ribbon or special carbon paper in front
of it, passes between the typewheels and hammers, and it is by making
a hammer dtive the local portion of the paper against a typewheel that
printing is achievedo
- 1 -
The comparator output signals mentioned above ultimately release the
energy which drives the hammers. Before doing so however
f
they pass
through a plugboard by means of which they can be routed to any de-
sired hammer. Thus any memory position can be made to print out in
any print position.
Time is allowed for one complete typewheel revolution so that every
character in the memory has a chance to print. Then the paper is
a new blockette of information is read in and the cycle re-
peated.
Physically, the High-Speed Printer is composed of four units: the
Uniservo, the Printer Unit, the Memory Unit, and the Power Supply
Unit. These units are interconnected by means of cableso Figure 1
is a picture of the printer. .
The above description of the printer implied that the magnetic tape
input must be in blockette form. On the middle right quarter of the
Univac Supervisory Control is a series of ten Block Sub-
Divider Buttons. Each button corresponds to a Uniservo. When one
or more buttons are depressed, each block written on the correspond-
ing Uniservos will be in blockette form. That is, the computer will
automatically interrupt the write instructi.ons for those Uniservos
for a short period after each multiple of 120 digits has been
Thus each block of 720 characters. will be composed of 6 sets of 120
character blockettes, a small blank space appearing between each
blockette.
Since the Tape-To-Card Converter requires magnetic tape record-
ed in blockette form, but at a somewhat different spacing for its
efficient operation, the-Block Sub-Divider Buttons for Uniservos
1, 2, 3, 4, 5, 6, 7 contain the appropriate delays for High-Speed
Printer tapes while the buttons for Uniservos 8, 9, - have delays
appropriate for the Tape-To-Card Convertero
To prepare High-Speed Printer tapes on the Univac, the programmer
need merely depress the Block Sub-Divider Button corresponding to
the Uniservo he has selected from among those listed above to re-
ceive this output. All write orders for these Uniservos should be
5n m's. Because of the extra space needed between blockettes, a
1500' reel of tape can contain up to 7500 blockettes (= 1250 blocks)G
In making time estimates, the programmer should note that because
of the reduced number of blocks there are fewer starts and StOPSi and
thus a full tape recorded in High-Speed Printer blockettes can be
made by the computer in 3.5
- 2 -
20 Printing Format Control Features
In addition to the ability to print the contents of a memory location
anywhere on the page, the High-Speed Printer also possesses several
other highly useful editing facilitieso
Zero Suppressione It is frequently desirable to suppress the print-
ing of zeros to the left of the first significant (non-zero) digit in
a resulto It may be required to do this in several areas along a line
of printingo By means of the plug board on the printer
u
this may auto-
matically be done in as many as 18 independent arbitrary areas(or
fields} 0
Fast Feedo On many forms there will frequently be areas where no
printing is to occuro Rather than step through these areas a line at
a time without printing anything, which would not only be slow but
would also require putting "blank" blockettes of information on the
tape q thus wasting tape and computer time as well
o
the paper can be
continuously moved at high speed from one printing area to the nexto
This operation is known as "Fast Feeding" 0 It is controlled by means
of a punched paper loop which is placed on the printer and moves in
synchronism with the papero The Fast Feed can be started by means of
a hole punched in the paper loop (in a certain channel) or by
any of several special coded combinations written on the magnetic
tapeo It is stopped by sensing the appropriate punched holes on the
paper loopo
Multilineo Normally the contents of the memory (one blockette or 120
digits) is printed out on a single lineo By means of the "Multiline"
symbol q placed at the beginning of a blockette of information, and in
conjunction with the plugboard
a
a blockette can be broken up and
printed out in as many as six lineso This feature may
permit great savings of tapea An obvious use is in name-and-address
printingo
- 3 -
Multiple Printingo By means of the plugboard, the contents of a
memory location can be printed in not just one but in two or
three places across the lineo In addition, if the High-Speed
Printer is operating in Multiline, the additional printings do
not even have to be on the same line (subject to certain re-
strictions to be discussed later). Of course
o
no more than
130 characters can be printed on anyone line.
30 Wiring For Single Line Printing
For printing of this nature; no Multiline
v
Multiple Printing,
or Zero Suppression; only the top panel of the plugboard is in-
volvedo The complete plugboard is shown in Figure 30 The
first 5 columns of holes on the left of the top panel of the
plugboard are numbered 1-24, 2 5 - 4 8 , ~ 0 0 9 97--120 0 These
holes are connected to the comparator output lines and are in
one-to-one correspondence with the 120 positions of the memoryo
In the next section of the plugboard to the right9 there are 130
pairs of holes
Q
each pair corresponding to a printing positiona
The two holes of a pair are vertically one above the other and
are internally connected togethero (The reason for using two
holes is to have a free one into which a second jumper can be
put so that the comparator signal can be sent to more than one
pointo This is necessary in Multiple and Multiline Printingo)
It is only necessary now to connect by means of jumpers the set
of 120 holes, corresponding to the memory locations g to the set
of 130 pairs of holes (that is, to one hole in each pair) cor-
responding to the hammers or printing positionso This can be
done in one-to-one or any other desired patterno If it is not
required to print certain memory locations, or if it is known
that certain locations will always contain non-printing charact-
ers, these may be left unplugged.
For most purposes, 6, 8, or 10 inch jumpers will be found besta
Short jumpers make for a tight, neat-looking board but have
the disadvantage that it becomes difficult to enter the heart of
a heavily plugged-up area to change or add a jumper.
For electrical reasons only, it is necessary to place a few
jumpers in the Multiline section of the plugboard even though no
Multiline Printing is to be doneo Connect (upper panel):
- 4 -
32A to 33F
33A to 34A
338 to 348
33C to 34C
33D to 34D
33E to 34E
There is one more thing that must be done before the board can be
used. The.bottom four holes of the 16th column of the board deter-
mine the line spacing
v
ioeo, whether single space, double space or
triple s'paceo A jumper is plugged from the bottom hole (marked
"out") to anyone of the three above (marked "l"v "2", "3" and
corresponding to single space, double space and triple space, re-
spectively)
40 Modes of Operation
Under normal operation, the High-Speed Printer will stop printing
(and reading tape) whenever a blockette is read which contains the
printer stop symbol E (1 11 0000) 0 The blockette containing the
stop symbol is completely read into the memory but is not printed.
If a blockette is read into the memory which contains a printer
breakpoint symbol P (0 11 0001) and the breakpoint switch is in the
b'reakpoint position, the printer interprets the (3 as a z::.. The
blockette containing the f3 or l can be printed it desiredo
When the "Print, No Read - Read, No Print" switch is placed in the
.tRead, No Print" position
o
the printer will read a tape without
printing until a stop or breakpoint symbol is' ettcounteredo It is
also possible to skip down a tape in either the forward or back-
ward direction and stop when a Fast Feed I symbol <0 01 1111),
printer stop. or breakpoint symbol is encounteredo
There are 63 possible UNIVAC code combinations, only 51 of which
appear as characters on the typewheels of the High-Speed Printer.
The other 12 characters are normally not printed and are either
completely ignored or serve as editing symbols, such as the stop
and breakpoint mentioned above.
At certain times, however
Q
it may be desirable to have a visual
record of all the characters on the tape including the normally
non-printing oneso This can be done by operating the Computer
Digit switcho The effect of this is to cause each bloekette to
print out in two lines.
- 5 -
On the first line only the normal printing characters appearo On
the second line
Q
any normally non-printing symbols which may be
present in the blockette will print, and only such symbols. They
will print as regular characters and can b ~ identified by the
following table.
Pulse Code Name Computer Digit
Print-Out
1 00 0000 Ignore 5
o 00 0001 Space 6
o 01 0000 Multiline E
1 01 1110 C
o 01 1111 Fast Feed I D
0 10 0000 Tab N
1 10 0001 Tab 0
1 10 0010 Fast Feed II P
o 10 1111 Fast Feed III M
1 11 0000 Stop V
o 11 0001 Breakpoint W
o 11 1110 Fast Feed IV T
Note that in each pair of lines comprlslng a blockette
o
one and only
one character should be printed in every column.
50 Zero Suppression
Zero suppression is controlled by the lower panel of the plugboard.
As mentioned previouslYQ up to 18 fields of arbitrary length may be
employedo
- 6 -
It should be remembered that the zeros are suppressed on read-in
to the memoryo not on print-out, so that the way in which the
memory is plugged to the printing positions must be borne in mindo
Single Line Printing Onlyo The holes in columns 26 through 31 on
the lower panel of the plugboard (PB II) are numbered 1 to 120 and
correspond to the 120 positions of the memoryo or the 120 digits
coming from the tapeo To start a zero suppression field in "normal"
or Single Line Printing, plug from the hole corresponding to the
address where the first zero is to be deleted (if present) to the
first of the 18 numbered holes in column 25 {hole 25A) 0 Then plug
from hole 23A Qabeled "Start (SL)")9 to hole 24Ao If there is to
be more than one field to undergo zero suppression, plug the start-
ing positions of the successive fields to the 2nd, 3rd, 4th, etco
holes in column 25 and then plug 23B to 24B, 23C to 24C, etc.
To end each fie1d g plug from the address corresponding to the last
zero to be deleted (if everything has been zero up to that point
g
of course) to successive holes in column 320 Then plug 33A to 34A
o
and if several fields are to undergo zero suppression, plug 33B to
34Bg 33C to 34C
v
etco
The fact that the High-Speed Printer can be run in either "Mu1ti-
linetll or Lin'e" (Normal) fashion has already been discussed
o
Since these two modes of operation
g
both of which may occur in a
run, in general produce different formats, it may be desirable to
observe one set of zero fields on single line printing and a
different set when on Multiline. There are three possible cases
to con.sider o
Fields for Single Line and Multiline Printing different. To do
this g plug the fields for Single Line Printing as described above.
They will not be observed when the printer is on Multiline. For
the Multiline fields
Q
plug the starting positions successively to
holes 25Rg 25Qv 25P
t
etco (working upwards) 0 Then plug 23S to 24S
(Labeled teStart (ML) '9) 9 23R to 24R, etc.
o
terminating one hole be-
low the topmost hole plugged in column 25 (in the Multiline group).
Plug the ending positions for the Multiline fields into holes 32R
Q
32QQ 32Pg etco, again working upward, and then plug 33S to 34S,
33R to 34R
Q
etc
oo
terminating, as before, one hole below the top
hole plugged in column 32. The Multiline fields will not be ob-
served when the printer is on Single Line
.... 7 -
Fields for Single Line and Multiline identical. If the fields to
undergo zero suppression are exactly the "same, that is, occupy the
same positions in the blockette, for Single Line and Multiline, the
plugging is simplified. Plug the fields as for Single Line. Then
run a jumper from the first unused hole in column 23 to the hole
marked "Start (ML) tt which is 24S. Likewise, run a jumper from the
first unused hole in column 33 to "End (ML" , hole 34So All the
fields will now be observed both on "Normal and Multiline printing.
Some fields common to Single Line and Multiline. When only some of
the zero suppression fields are common to both Multiline and Single
Line Printing while others are distinct to one or the other, the
situation is somewhat more complex. Here "Y" jumpers must be used
for the common fields. The stem of the tty" is plugged into the
address location and then one fork is plugged in with the upper
<Single Line) group in column 25 (or 32) and the other fork is
plugged to the lower (Multiline) section of column 25 (or 32)0 The
other (non-common) fields are plugged as before. The jumpers between
columns 23 and 24 and between columris 33 34 are run as previouslyo
Two more points with respect to Zero Suppression should be noted:
1. If one Zero Suppression field follows another
immediately (e.g., 22-37, 38-51), the ending
point of the first field need not be plugged.
2. If a single digit field is to undergo zero
suppression, a "Y" jumper must be used. The
stem of the tty" is plugged into the address
location and one fork goes to column 25 (start)
while the other goes to 32 (end) 0 If the single
digit field is followed immediately by another
field (as in algebraic sign followed by number)
then, as stated above, the ending of the single
digit field need not be plugged and a straight
jumper can be used ( to plug the
6. Fast Feed
There are four magnetic tape symbols which may be used to initiate a
Fast Feed. They are:
- 8 -
Pulse Code Name Modified Unityper I and
Unityper II Symbol
o 01 1111 FF I
@l
1 10 0001 FF II
I
o 10 1111 FF III
?
o 11 1110 FF IV =
The Fast Feed symbol used is not printed and it will be observed
as a Fast Feed instruction only if it is placed in the first digit
position of a blockette. The fast-feeding is accomplished before
the blockette containing the symbol is printed.
As mentioned in Section 4, it is desirable that the first block-
ette of a new form (or form group) should employ an FF I: other-
wise there is no restriction on which Fast Feed symbols are used o
When started, a Fast Feed will continue feeding paper without
printing or reading until a hole is encountered in that channel of
the Paper Control Loop corresponding to the Fast Feed symbol (e.g.,
a Fast Feed started by a FF II symbol is stopped only by encounter-
ing a hole in channel 2 of the Paper Control Loop). A Fast Feed
can also be started by punching a hole in channel 5 of the Paper
Control Loop. This type of Fast Feed is stopped only by a hole
sensed in channel 6. This "Loop Controlled
tt
Fast Feed is useful
in taking care of "overflow" from one form to the next. A Loop
Controlled Fast Feed is never started until the first digit of the
blockette currently being read is sensed. If this should be a
Fast Feed symbol, it takes precedence over the channel 5 hole; i.e.,
the magnetic tape controlled Fast Feed is obeyed, the Loop Control-
led Fast Feed is ignoredo
A special punch with sprocket hole aligning keys is used in pre-
paring the Paper Loop Control tape. To punch the start of a
Fast Feed, punch a hole in channel 5 on the line corresponding to
the first line to be skipped. To stop the Fast Feed, punch a hole
on the line where the first line of printing is to occur after the
Fast Feed. Channel 6 is used for stopping if the Fast Feed was
started with a channel 5 hole. Channell, 2, 3, or 4 is used for
a tape-started Fast Feed.
- 9 -
The longest loop the High-Speed Printer can handle is 22 inch-
es; the shortest is 11 incheso Of course, the punching for one
form can be repeated several times, if desired, <and if the form
is eleven inches long or less). Indeed, this is preferable
since greater loop life will resulto
There are two restrictions that must be observed in the use of
Fast Feed:
10 A Fast Feed, whether loop or tape started, must
always cause at least one line to be skipped.
2" A channel 5 hole should never be punched on the same
line with any other hole.
During a Fast Feed operation, the paper is moved at a rate of
20
tt
/second, or, in other words, 7200 lines per m i n u t e ~
7. Multiline Printing
The Multiline symbol ~ <0 01 0000) is used to put the High-Speed
Printer into the Multiline mode of operation. It should be
placed in the first digit position of the blockette to be multi-
lined, unless there is a Fast Feed symbol required too, in which
case the Fast Feed symbol is placed first, the Multiline symbol
second.
The printer returns to normal operation at the end of each multi-
lined blockette so that even if consecutive blockettes are to be
done in Multiline, a new Multilirte symbol must be placed in each.
The Multiline symbol will not be printed.
Multiline operation is entirely a function of the plugboard and
the plugging operations may be broken into three groups:
- 10 -
10 Selecting the number of lines in which a blockette
is to be printed.
2. Selecting the memory positions which are to be
printed on each lineo
30 Selecting the printing location on each line for
each charactero
Selection of the number of lines. This plugging consists of
four steps and only the upper right hand section of the plug-
board is under concern.
10 Run a jumper from the hole marked "Home" (33F)
to the hole in column 32
Q
counting down from the topu
the number of rows corresponding to the number of
lines into which the blockettes are to be divided in
Multiline"
20 For rows above the hole selected in (l)Q run jumpers
between corresponding holes in columns 32 and 33"
30 For rows includina and below the hole selected in (l)u
run jumpers between corresponding holes in columns 33
and 34v down to and including row E ~
40 If the number of lines selected is 2 to 6
9
connect
holes 32S and 32T (marked 'tOne Linen) with a jumper;
if the number of lines selected is Iv omit the jumper
between these two holeso
The following examples will illustrate the required plugging:
Example 1: Four line printing
10 Connect 33F to 320
20 Connect 32A to 33A; 32B to 33B;
32C to 33C
30 Connect 330 to 340; 33E to 34E
4() Connect 32S to 32T
Example 2 ~ One line printing
1. Connect 33F to 32A
2. No plugging required
3. Connect 33A to 34A; 33B to 34B;
33C to 34C; 330 to 340
33E to 34E
40 No plugging required
= 11 =
Example 3: Six line printing
1. Connect 33F to 32F
20 Connect 32A to 33A; 328 to 338
32C to 33C; 32D to 33D
32E to 33E
30 No plugging required
40 Connect 32S to 32T
Selection of the memory locations to be printed on each line.
In Multiline work th'ere is a set of 15 relays brought into play,
different ones of which can be energized on different lines under
the control of the plugboard. Further, by plugging the memory
locations to the of these relays, the characters to be
printed out on each line can be selected simply by the relays
which have been chosen to be energized for each line.
Each relay has 12 contacts and thus can handle 12 digits. The
first step, then, is to decide how many relays are required to
print each line and to plug them accordingly. To avoid con-
fusion, relays be chosen sequentially. Thus, if 40 digits
are to be printed on the first line, 48 digits on the second
line and 6 digits on the third line, relays 1, 2, 3, and 4 (room
for a total of 48 digits) should be plugged to operate on the
first line; 5, 6
Q
7, and 8 should be plugged for the second line,
and relay 9 for the third line.
This is accomplished as' follows (see upper right hand portion of
the plugboard): Holes 32G, H, I, J, K, and L (labeled ttML Counter,
1, 2, 3, 4, 5, 6
ft
, respectively) give output signals in turn when
the printer is on line 1, 2, 3, etce, of a Multiline operation.
Holes 33G and 34F (internally connected together), 33H and 34G,
331 and 34H, .. 0, 33U and 34T are pairs of holes going to the 15
relays coils in order. They are labeled 1 to 15 on the plugboard.
To operate relay 1 on line 1, run a jumper between holes 32G and
34F. Then to also pick up relays 2, 3, and 4 on line 1, as re-
quired in the example, run jumpers between 33G and 34G, 33H and 34H,
and 331 and 341. This, of course, could be extended if more relays
were required to be picked up on line 1. (As many as 10 relays
may be used on any given line. This gives sufficient contacts--
120, to handle the entire memory.)
For line 2, again assuming four relays are to be picked up, plug
32H (line 2 output) to 34J (relay 5 coil). Then plug 33K to 34K,
33L to 34L and 33M to 34M.
-12-
On line 3, only one relay is assumed required. Pick this up
by connecting 321 to 34N.
One important rule should now be If, on any given line
in a Multiline operation, zero, one, or two relays are used,
then a resistor must be plugged in on that line. This is easily
done by running a jumper--
a. From the hole in ,the pair corresponding
to the second relay in the chain (if 2 relays
are used), or
b. From the unused hole in the pair corresponding
to the single relay (if only one relay is used),
or
c. From the line output hole itself (holes 32G to L)
if no relays are used (i.e., nothing is to print
on that particular line) to one of the six holes
in the section marked "Resistors" (holes 32M to R).
Thus, in the example above, since only one relay is used on
line 3, a jumper should be run from hole 330 to 32M. Note
carefully that this plugging of resistors applies only to the
lines within the group size selected for the Multiline opera-
tion. Thus, in the above example, where a 3 line Multiline
was assumed, nothing at all, need be plugged to the last three
output lines (32J, K, and L) since they will never be excited.
This should not be confused with the situation where, say, in
a 4 line Multiline it is not desired to print anything on line
2. Here, no relay need be connected to line 2, but a resistor
should be.
Having wired up the relay coils, the next step is to send the
memory locations (actually, comparator outputs) to the proper
relay contacts. The 180 relay contacts (12X15) go to 180 pairs
of holes which are located in columns 17 through 31 in the
upper plugboard panel. One column represents the contacts for'
one relay.
Now, unless the run is exclusively Multiline, the plugboard
will have already been wired for Single Line Printing, as
described in Section 3. That is, the holes in columns 1
through 5 (labeled ttFrom Comparator") will be wired to the
-13-
paired holes in columns 6 through 16 (labeled "To Single
Lines Relays"). Thus, each memory location can, in general,
be picked up from the unused hole in the pair here. If
certain memory locations do not print in Single Line work
(and are to print in Multiline work), or if there is to be no
Single Line Printing at all, then the memory locations must
be picked up from the first five columns directly. It is
then simply a question of taking those memory locations which
are to print on the first line and plugging them sequentially
to the relays which have been assigned to the first line, and
similarly for the other lines. Remember that the holes cor-
responding to the relay contacts are paired.
As an example, suppose the first 40 memory locations are to
print in Single Line Printing in positions 1-40. Then hole
lA would have been plugged to 6A, IB to 6C, lC to 6E, .. ,
2P to 9G. Now suppose these same 40 memory locations are
also to print on line 1 of a operation. The first
4 relays (columns 17, 18, 19, and 20) would have been allo-
cated for this purpose. Then the plugging would be: 6B to
17A, 60 to 17C, 6F to 17E, .. , 9H to 2OG. Notice that
many contacts on the fourth relay are unused. Information
to print on line 2 would be plugged starting with the
relay (column 21).
Selection of the printing location for each character.
The other sides of the Multiline relay contacts go to a set
of 180 pairs of holes on the lower plugboard panel (columns
7 through 21). Again, each column corresponds to a relay.
Thus, column 17 on the upper panel is the "input" side of
relay 1, while column 7 on the lower panel is the "output"
side of the same relay. The individual holes in the
columns are likewise in one-to-one correspondence. Thus,
a comparator signal entering the first (or second) hole in
column 17 of the top panel will appear at the first (and
second) hole in column 7 of the bottom panel when relay 1
is energized (i.e., on line one).
The first five columns and the first ten holes of the
sixth column on the lower panel (making 130 holes in all)
are connected one-to-one to the 130 thyratrons which drive
the printing hammers. The procedure now is to connect
the line 1 relay outputs (however many relays may be used
for line 1) to the thyratron holes in the manner desired.
Now on line 2, some of the characters may need to be
-14-
printed ino as yeto unused positfonso These are plugged directly to
the corresponding thyratrons o Other characters (or more p r o p e r l y ~
memory locations) may need to be.printed in positions already print-
ed on line 10 To do thiso plug from these particular holes in the
line 2 relay output set to those unused holes of the pairs of the
line 1 relay output
o
where the other hole of the pair is already
plugged to the desired thyratrono
A similar procedure is followed for line 3 and the following lines Q
There will always be a free hole in a relay output pair into which
a jumper from an output on a later line can be plugged to achieve
printing in the same position as the earlier line
o
As an example of the above pluggingo consider the following exercise:
Print the first three memory locations in the first
three print positions on line 1; the next four memory
locations in the first four positions on line 2; and
the next 5 memory locations in the first five posi-
tions of line 30
It is assumed that the first Multiline relay has been assigned for
line 10 the second for line 20 and the third for line 30 Further
o
that the comparator outputs have been picked up and are plugged
into
o
in order 0 l7Ao 17C
o
l7Eo 18Ao 18C
o
18E
o
18G
o
19Aq 19C
o
19Eq
19G
o
and 191 (all on the lower panel).. The required plugging then
is (all on the lower panel):
7A to lAo 7C to IBo 7E to IC
8A to 78
0
8C to 70q 8E to 7Fo 8G to 10
9A to 8B
o
9C to 80q 9E to 8F
a
9G to 8H
o
91 to IE
8.. _Multiple Printing
(line 1)
(I ine 2)
(I ine 3)
,Plugging procedure for Single Line operationo By means of the paired
holes in the section of the upper panel marked uTo Single Line Relaysttv
a memory location can be readily printed out as many as three times on
a line ..
- 15 -
to do this
o
simply plug from the unused hole in the first position
chosen to one of the paired holes (preferably the top) in the
second place where printing is desiredo If triplicating is re-
quired
o
run a jumper from the unused hole of the second position
to the third positiono
Example: Print the-contents of the first
memory location in positions 10
610 and 121.
Plug lA to 6A
6B to llA
lIB to 16A
If some of the memory positions for Single Line Printing are common
to Multiline Printing also
o
it is desirable to do the Multiple
plugging for Single Line firsto then the last unused hole in a pair
may be connected to the Multiline relayso
Plugging procedure f6r Multiline operationo In Multiline work
o
a
character cannot only be printed out 3 times on a given line
o
but
then can be printed out up to 3 times again on any other line in
the groupo This is done by 1 ~ c h a i n pluggingtQ, as noted above
o
in
the .fTo Multiline RelaystQ section of the plugboard
o
By connecting
the output to contacts on relays (or a single relay), which close
on the same line, the digit may be printed up to 3 times on that
same lineo Then by extending the chain -of plugging to relays
which close on other lines
g
the same character may be repeated on
those lines o On the lower panel
o
the positioning of the characters
on the line is taken care of as in regular Multiline work (dis-
cussed in Section 7)0
Example: Print out memory location I in
positions I and 61 on Single Line
operation. On Multiline operation
print out memory location 1 in
positions 10 25, and 49 on line 10
and in positions 25 and 49 on line
2.
Assume relay 1 is energized on line 10 and relay 2 on line 20 Then
the plugging is
o
by panel:
= 16 =
On the upper panel
Q
plug IA to 6A
6B to llA
lIB to l7A
l7B to l7C
170 to 17E
17F to l8A
l8B to l8C
On the lower plug 7A to IA
7C to 2A
7E to 3A
8A to 7D
8C to 7F
Note that by this method it is possible to have, at most,
180 print=outs from one blockette in Multiline, since there
are only 180 relay contacts. However, there is complete
flexibility in the positioning of the information on each
line and, of course, complete independence between Single
Line and Multiline Printing.
There is another system for doing multiple printing on
Multiline which allows the entire memory to be triplicated,
if desired (i. produce as many as 360 print-outs from
one blockette)0 However
Q
in using this method there must be
no Single Line work at allor, if Single Line operation does
occur, one must accept the same duplicating or triplicating
pattern as is plugged for the Multiline work.
This system is particularly useful where each line of the
Multiline group is printed directly under the line above
and the group as a whole is duplicated or triplicated across
the pageo This system is used as
10 Plug for straight Multiline work as in Section 7,
ignoring the Multiple Printing problem (ioeo, simply
plugging for one printing) 0
20 On the lower panel 0 in the ttFrom Multiline Relays"
section, outputs going to the same print position
on different lines will have been chain plugged
together and run to one set of thyratron holes
o
However, each of the outputs for the last line will
have a free holeo If only duplicating of the group
is required, run straight jumpers from these free
holes to the appropriate thyratron holes. If
-17-
triplicating is required, tty'9' jumpers are
used to connect each free output hole to
two thyratrons o
Print the first memory location
on line 10 the second memory
location on line 2 and the third
on line 3D On all three lines
the printing is to be in tripli-
cate
9
printing in columns 1, 49,
and 970 (Use the first Multiline
relay on line 10 the second on
line 2 and the third on line 30 )
On the upper panel, plug lA to 17A
IB to 18A
lC to 19A
On the lower panel
o
plug 7A to lA
8A to 7B
9A to 8B
9B to 3A and SA (tty" jumper)
Note that if the printer should go into Single Line Printing
and if something were plugged to print (on Single Line) in
either positions 49, or 97
1
it would print in all three
positions. This might be all right but, if independently
some other memory location should be plugged to another of
these three points, there would be trouble. (The printer
would stop with a Print Check Error.) Thus
g
great care must
be exercised in using this method of Multiple Printing if
Single Line operation can also occur in the run o
9. Checking Features
The error detection circuits on the High-Speed Printer may
be considered in the three sections through which information
recorded on tape is 1- reading of data from tape,
2- storage of data in the memory, and 3- printing of the
memoryo
-18-
Error detection in reading tape. Each digit read from tape
is given a binary bit count. If the number of binary ones
present is odd the character has been read from tape cor-
rectly. If an even count is detected, the remainder of the
blockette is read, but the Odd-Even Check Error is set and
the printer stops. Nothing is printed.
Further, each blockette read from tape must contain exactly
120 digits a If a blockette is longer or shorter than 120
digits, the printer will set the 120 Check Error and stop.
Nothing will be printed.
Error detection in storage. Each of the digits coming from
the tape is placed in the proper position in the memory by
virtue of the set of 120 address lines, which are sequen-
tially excited by the function table which decodes the Main
(or Address) Counter. If an address line should fail to
rise to the signal level when it is supposed to, the corres-
ponding digit coming from tape could not enter the memorYe
The Address Check circuits look for this and g if an address
line should fail to be excited at the proper time, the
Address Check Error is set and the printer stopped.
Error detection in printing. The fundamental checking of
the printing is done by means of the group of 130 check
thyratrons working in conjunction with the All-Out Detector.
At the end of each print cycle, all 130 check thyratrons
should be extinguished and the All-Out Detector looks for
this. If some column fails to print when it should or prints
an incorrect character, its check thyratron will be on when
probed. This causes the machine to stop with the Print Check
Error set.
The determination of whether all check thyratrons have been
extinguished or not is made by a circuit called the All-Out
Detector. If this circuit should fail in a certain way, a
steady All-Out signal would be given and printing error would
fail to stop the machineo To guard against this, the All-Out
Detector is itself checked every line and, if not functioning
properlyo the printer will stop with the All-Out Detector
Error seto
In addition to the major checking circuits described above,
there are other checks applied to the control circuits and
paper feed mechanism to ensure complete accuracy of printing.
The description of these checks requires a detailed under-
standing of the printer logic.
-19--
10. Example Problem
As an example of simple plugging for the High-Speed Printer, suppose
a series of paychecks are to be printed. The desired check format is
shown in Figure 4. The blockette layout is shown in Figure 5. The
simple nature of this example permits us to do Multiline Printing only,
with one blockette per check. An explanation of the net pay field in
the blockette is called forQ It is obviously desirable that the
amount of the check be printed in as unalterable a manner as possible.
This is easily done if, say, twenty dollars and fifty cents is printed
as $20.50 and not as (assuming pay up to $9990 99 is permitted) $ 20.50
which would be the result from simple zero Thus, the net
pay field is assumed to be pre-edited by the UNIVAC when it is inserted
in the blockette. Thus, the above example would appear as _$2050 in
the blockette.
A. Selection of the Number of Lines (6line Multiline):
Plug (Upper Panel)
32F to 33F
32A to 33A
32B to 33B
32C to 33C
320 to 330
32E to 33E
32S to 32T
B. Selection of the Multiline Relays:
Line Relays Plug (Upper Panel)
1 1\l2
q
3 32G to 34F
33G to 34G
33H to 34H
2 none 32H to 32M
3 4
11
5 321 to 341
33J to 34J
33K to 32N
-20-
4
6
11
7
0
8
0
9 32J to 34K
33L to 34L
33M to 34M
33N to 34N
5 10\111 32K to 340
33P to 34P
33Q to 320
6 12\113\114 32L to 34Q
33R to 34R
335 to 345
C. Comparator Outputs (Memory Locations) to Relay Contacts:
Line Field Plug (Upper Panel)
1 check number IB to l7A
17B to 185
lC to 17C
17D to l8U
10 to 17E
l7F to 18W
IE to 17G
17H to 19A
IF to 171
17J to 19C
IG to 17K
17L to 19E
net pay 2B to, 17M
2C to 170
20 to 17Q
2E to 175
40 to 17U (decimal pt)
2F to 17W
2G to 18A
date IN to IBC
18D to 19G
10 to 18E
18F to 191
3B to 100 (dash)
18H to 19K
1P to 181
18J to 19M
lQ to 18K
18L to 190
-21-
48 to IBM (dash)
18N to 19Q
1R to 180
18P to 195
IS to 18Q
l8H to 19U
3 regular pay 20 to 20A
2P to 20C
2Q to 20E
l7V to 200 (decimal pt)
2R to 201
25 to 20K
withholding tax 3H to 20M
31 to 200
3J to 200
20H to 205 (decimal pt)
3K to 20U
3L to 20W
bond deduction 3U to 21A
3V to 21C
2m to 21E (decimal pt)
3W to 21G
3X to 211
overtime pay 2T to 22A
2U to 22C
2V to 22E
21F to 22G (decimal pt)
2W to 221
2X to 22K
FICA tax 3M to 22M
3N to 220
228 to 22Q (decimal pt)
30 to 225
3P to 22U
insurance 41 to 22W
4J to 23A
22R to 23C (decimal pt)
4K to 23E
4L to 23G
name 5A to 231
58 to 23K
5C to 23M
5D to 230
5E to 23Q
5F to 235
5G to 23U
.. 22-
SH to 23W
SI to 24A
SJ to 24C
SK to 24E
SL to 24G
5M to 241
SN to 24K
SO to 24M
SP to 240
SQ to 24Q
SR to 245
S5 to 24U
ST to 24tv
SU to
,,'25A
SV to' 2SC
5W to 25E
5X to 25G
net pay I7N to 251
I7P to 25K
17R to 25M
I7T to 250
4C to 25Q (decimal pt)
I7X to 2S5
l8B to 25U
S medical pay 3C to 26A
3D to 26C
3E to. 26E
25R to 26G (decimal pt)
3F to 261
3G to 26K
union dues 3E to 26M
3F to 260
26H to 26Q (decimal pt)
3G to 265
3H to 26U
other dedo 3Q to 26W
3R to 27A
26R to 27C (decimal pt)
35 to 27E
3T to 27G
6 first adjustment lH to 28A
II to 28C
lJ to 28E
27D to 28G (decimal pt)
lK to 281
lL to 28K
second adjustment IT to 28M
-23,-
lU to 280
IV to 28Q
28H to 285 (decimal pt)
lW to 28U
IX to 28W
third adjustment 2H to 29A
2I to 29C
2J to 29E
28T to 29G
2K to 29I
2L to 29K
badge number 4M to 29M
4N to 290
40 to 29Q
4P to 295
4Q to 29U
4R to 29W
45 to 30A
4T to 30C
40 From Relay Contacts to Print Positions
Line Field Plug (Lower Panel)
1 check number 7A to lKI1
85 to 35
7C to lL!] 8U to 3T
7E to lMo 8W to 3U
7G to IN!] 9A to 3V
7I to lOa 9C to 3W
7K to lP
o
9E to 3X
net pay 7M to 2D
70 to 2E
7Q to 2F
75 to 2G
7U to 2H
7W to 2I
8A to 2J
date 8C to 25
0
9G to 5X
8E to
2Ta
9I to 6A
8G to 2U, 9K to 68
8I to 2Vo 9M to 6C
8K to 2W\I 90 to 6D
8M to 2X
II
9Q
to 6E
80 to 3A
l1
9S to 6F
8Q to 38
11
9U to 6G
3 regular pay lOA to 78
lOC to 7D
-24-
10E to 7F
lOG to 7H
101 to 7J
10K to 7L
withholding tax 10M to 7X
100 to 8B
10Q to 2K
lOS to 2L
IOU to 2M
lOW to 2N
bond deduction 11A to 8N
11C to 8P
lIE to 8R
11G to 3C
III to 3D
4 overtime pay 12A to lOB
12C to 100
12E to 10F
12G to 10H
121 to 10J
12K to 10L
FICA tax 12M to lOP
120 to lOR
12Q to lOT
12S to 10V
12U to lOX
insurance 12W to lIB
13A to 110
13C to 11F
13E to 11H
13G to 11J
name 131 to 4E
13K to 4F
13M to 4G
130 to 4H
13Q to 41
13S to 4J
13U to 4K
l3W to 4L
14A to 4M
14C to 4N
14E to 40
l4G to 4P
141 to 4Q
14K to 4R
14M to 4S
140 to 4T
-25-
14Q to 4U
145 to 4V
14U to 4W
14W to 4X
15A to 5A
15C to 58
15E to 5C
15G to 5D
net pay 151 to 5W
15K to 9H
15M to 9J
150 to 9L
15Q to 9N
155 to 9P
15U to 9R
5 medic,a1 pay 16A to 128
16C to 12D
16E to 12F
16G to 12H
161 to 12J
16K to 12L
union dues 16M to 12N
160 to 12P
16Q to 12R
165 to, 12T
16U to 12V
other deductions 16W to 12X
17A to 138
17C to 13D
17E to 13F
17G to 13H
6 first adjustment 18A to 12K
lac to lQ
18E to lR
18G to IS
181 to IT
18K to 10
second adjustment IBM to 2A
180 to 2C
18Q to 7N
185 to 7P
18U to 7R
18W to 75
third adjustment 19A to 16P
19C to 16T
-26-
19E to 16V
19G to 20
191 to 2P
19K to 2Q
badge number 19M to 8T
190 to 8V
19Q to 8X
19S to 98
19U to 90
19W to 9F
20A to 4A
20C to 48
5. Zero Suppression
Plug (Lower Panel)
Start Zero Fields End Zero Fields
261 to 25R, 23R to 24R 26L to 32R, 33R to 34R
27A to 25QI1 23Q to 24Q 270 to 32QI1 33Q to 34Q
27M to 25P
I1
23P to 24P 27P to 32P
o
33P to 34P
27S to 250, 230 to 240 28C to 320
9
330 to 340
280 to 25N
I1
23N to 24N 28H to 32N
I1
33N to 34N
28K to 25M, 23M to 34M 280 to 32M, 33M to 34M,
28P to 25L
o
23L to 34L 28T to 32Lo 33L to 34L
29A to 25K
o
23K to 34K 290 to 32Ko 33K to 34K
29E to 25J\l 23J to 34J 29H to 32J
o
33J to 34J
291 to 251
0
231 to 341 29L to 321
11
331 to 341
29Q to 25H
o
23H to 34H 29T to 32Ho 33H to 341
30A to 25G
o
23S to 34S 300 to 32G
9
335 to 34S
6. Line Spacing
Plug (Upper Panel)
l6X to l6U
, -27-
COLUMNS


10 20 30 11-0 50
YOUR EARNINGS RECORD
CHECK NO.: .... 1 __ -"
EARNINGS
NET I"Ay:r- DATE: ... 1 ___ ....
..---......
REG.I"Ay: ... 1 __ -"
OVERTIME: :1 ==:::!
MEDICAL:
ADJUSTMENTS: 0 ~
DEDUCTIONS
WITHHLDG.TAXd I
FICA TAX:F ~
UNION DUES: c:::::J
BOND:
INSR:
OTHER:
Dc=J Dc:=)
ADJUSTMENT EXPLANATIONS:
1 OVERPAYMENT 3 EXPENSE ACCOUNT
2 UNDERI"AYMENT. 4 COMI"ANY STORE
Figure ij
60 80 90 100 110
CHECK N O . ~ ~ ____ ~
:r:
~
~
...
::!
~
"
, e
. ~
-
~
~
Q.
120
DATE =:J
PAY TO THE ORDER OF: I - ~ ,THE AMOUNT:
---_J
BADGE NO. :1 --]
ON THE ACCOUNT OF
TI-I E B LAN K CO., INC. , N Y , N Y
Treasurer
PR1NT WHEEL POSITIONS
131
I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
u
y
w
x
00 01 10 II
0000
0001 ,
0010
-
.
. .
0011 0 , )
+
0100
I A J /
0101
2 B K S
0110 3 C L T
0111 4 0 M U
1000 5 E N V
lOCH 6 F 0 W
1010 7 G P X
1011 8 H
0 Y
1100 9 I R Z
1101
,
# $ %
1110 &
1ft
IIII (
PRINTING CHARACTERS AND THEIR PULSE CODES FOR
THE HIGH-SPEED PRINTER
Figure 2
2
31 0 :0 I
[0 : - I
5 I
6 [0:
7
8 [ :
9 [ :
2
I" ME: PAY: ]
I
:. I
0 : OJ
FIGURE 5
CHANGES .1.N UNIPR INTER A!iI2 UNITYPER llQ .EllltlS. .QQQ
.EQ!l PROCESS ING OF 12aI6. .EQE l!lQtl S'PEED PRINTER
1. Keyboard for Supervisory Control and Unityper 1
The keyboard and the decoding unit have been changed to allow the typing
onto magnetic tape of any of the 63 pulse code combinations. All indica-
tions of upper and lower case have been removed from the keyboard and
also one of the duplicate numeric sections, namely that one over the al-
phabetic portion. More resistors have been added in the decoding unit to
allow for the decoding of 63 characters where previously only 51 were de-
coded. No other changes in the Unityper have been made. The changes are
illustrated by the accompanying figure and table o
2. Printer Dolly Uniprinter
A new switch has been added on the inside which has two positions: Uprint"
and tfstall
ff
This switch affects only the twelve new pulse code
combinations.
In the ffstall" position, when one of the new characters enters the printer,
the printer stops. The operator determines the character by looking at
the neon lights and prints manually any character he desires, after which
the printer automatically proceeds.
If the is in the "print" position, then when one of the new charac-
ters is encountered, a ";" (semi-colon) is printed in lower case operation
or a ":" (colon) in upper case operation.
One additional change is that the tab operation is indicated by printing a
"v
u
in the computer digit mode of operation.
3. Typing-lll Qf Information the Uniprinter
The Uniprinter operates in the same fashion as before.
operator must remember that to make the printer
press the "$" key; to perform the "unshift" operation,
"7" key; and to perform a "single 'shift" operation, he
"%" key.
-1-
The programmer and
lock", he must de-
he must depress the
must depress the
~ UNIVAC PULSE 90DE
MODIFIED UNIPRINTER
PULSE CODE MODIFIED UNIPRINTER I
Normal
--
Compo
AND S.C. KEYBOARD
L
I U'
Digit
Case
Case._
1 00 0000 i ignore x
o 00 0001 II
space space
o 00 0010
-
.-
-
-
1 00 0011 0 0
) 0
o 00 0100 1 1 .
t
1
1 00 0101 2 2
"
2
1 00 0110 3 3 # 3
o 00 0111 4 4 $ 4
o 00 1000
9
5 % 5
1 00 1001 6 6
*
6
1 00 1010 7 7 & 7
o 00 1011 8 8
v
~
1 00 1100 9 '9
(
9
o 00 1101
,
; : ;
*
o 00 1110 & ;
.. .
. ,
*
1 00 1111
(
;
..
; . .
*
Table 1 - Part. 1
. ';'2-
MODIFIED UNIPRINTER
PULSE CODE
MODIFIED UNITYPER I Normal Comp.
AND S.C. KEYBOARD L U Digit
Case Case
o 01 0000 r Car. Ret.
!
1 01 0001
.,
, ,
t
I
1 01 0010

.

a 01 0011 ; ; : ;
!
1 01 0100 A a A A
o 01 0101 B b B B
a 01 0110 C c C C
1 01 0111 D d D D
1 01 1000 E e E E
a 01 1001 F f F F
o 01 1010 G g G G
1 01 1011 H h H H
a 01 1100 I i I I
1 01 1101 # ;

;

*
1 01 1110
;
..
; ..
*
o 01 1111
@
.. '
:
,
-
. ,.
*
Table 1 - Part 2
-3-
PULSE CODE
o 10 0000
1 10 0001
1 10 0010.
o 10 0011
1 10 0100
o 10 0101
o 10 0110
1 10 0111
1 10 1000
o 10 1001
o 10 1010
1 10 1011
o 10 1100
1 10 1101
1 10 1110
o 10 1111
MODIFIED UNITYPER I
AND S. C. KEYBOARD
t
tt
1
)
J
K
L
M
N
o
P
Q
R
$
*
?
Table 1 - Part 3
-4-
MODIFIED UNIPRINTER
Normal
'!
Comp. I
L U Digit
Case Case I
tab-----+--v--l
; :
j J
k K
1 L
m M
n N
o o
p
p
q Q
r R
shift lock
,;
unshift
J
K
L
M
N
o
P
Q
R
z
8
~ *
,
~ *
;i
*
I

i1
I'
i'l
i!
;
,I
"



r,
i
MODIFIED UNIPRINTER
PULSE CODE
MODIFIED UNITYPER I Normal
AND S.C. KEYBOARD
L
l
U
.-....... ;.,"'""- ...rt
Case

1 11 0000 l: Prine Stop
a 11 0001

Prine Bkpt.
a 11 0010
. .
.
. , .
1 11 0011 +
+
@
a 11 0100
I I
?
1 11 0101 5 s 5
1 11 0110 T t T
a 11 0111
U u U
a 11 1000 V v V
1 11 1001 W w W
1 11 1010 X x X
a 11 1011
y
Y
y
1 11 1100 Z z Z
0 11 1101 %
single shift
o 11 1110
=
;
:;
1 11 1111
Table 1 - Part 4
* With the switch on "print", these are the characters that will
be printed; on ff s tall
n
, the printer stops.
-5-
Comp.
Digit
Stop
y
;
+
/
5
r
U
V
W
X
y
Z
-
;
*
*
nOD 0
LJI l; 0
I Loopl I
2 I 'STOP i BACK
, ' SPACE
tJ [] G G [:J [:J GSB 5J [:J E] [J [J .
[ SPACE . II. CONlROL BAR. J [J
B
New Keyboard
Old Keyboard
Fig" 1 - Changes in Keyboard of Uni typer I and Supervisory Control
8
\0