Skip to main content

Full text of "llnl :: pdp1 :: LRL PDP-1 UsersGuide Aug67"

See other formats

Raymond De Saussure 

August 1, 196T 


Introduc_ti,CT 'ITie PDP-1 is a binary digital computer with a 5 

microsecond memoi-y cycle and U096 words of I6 bit core storage. 
Auxiliary storage is available in the form of four potter tape 
units (200BPI dencsity), and two IB^f 729-IV tape units (200, 556, 
80OBPI density). Although an older machine -- first introduced in 
late i960 -" it is particularly versatile in I/O conversion, and 
many formfs of data handling are available here that cannot be 
found elsewhere at LRL, 

£iy modem standards, the PDP-1 is slow and has minimal 
storage. Arithmetic handling is poor; there is no floating point 
hardware, and it lacks a compiler although two assemblers PAP 
".r.d PA! arc a-ailablc. These difficul' i:;c r.rc of less importance 
zo data conversion and this is the prime function of the machine. 
Most input-output operations take in the order of milliseconds, 
T.o that a 5 microsecond cycle is not nrohibitive. On the other 
hand, extensive arithmetic operations should be moved to the 
larger oneratinp systems where they properly belonpr. The ideal 
use of the PDP"1 is as a "data-pump" such as the conversion of 
ARCTT napi-r tape to hifh density ma.gnetic tnnp without repard for 
format, the ta^e then being processed and msni-oulated on other machines, 

Commuiii cation between the FDP-1 and other computers at LRL 
must be either by means of mapnetic tape or else by punched cards, 
L'y ffi'- the :..,.;,»-. efficient method is mapnetic tape, and nunched cards 
wrc varrmiteci oisly if extensive hand manipulRtior of the data is 
cont.f?L';pl«.terK Market ic taper, are interchangeable between machines 
only under certain specific conditions which ^re well defined. 

- 2 - 

A aeven ch,aniiel tape --all the magnetic tanes at LRL are 
seven channel-- con- is t;- of a Beries of lines each of which contains 
six binar-y bits '■^f Information mid a seventh parity bit such that 
the total number •'>[" "<ne" bits is always even or odd. Binary tapes 
are written in odd narity, and RCD tapes in even parity. Thus, for 
example, an l8 bit vord from the PDP-1 is represented on magnetic 
tape by three parity checked lines of six bits each. Words are 
vritten con tin nous iy until a record gap is reached. From the tape 
a.ione , inhere is no way of determininp the number of bits in the 
originating machine. It could as easily be tvo words of l8 bits 
each, as one word of 36 bits. Therefore, to read a tape on a 
different machine than written, one hnr. oniv t,r> mnke oprtain tb>=it 
the total number of bits in the record is evenly divisible by 
the number of bits in the word of the computer doinp the reading. 
On this basifi, it may be seen that a l60 wn-^d record written on 
the PDP~1 ma;,'- be read by a maximum number of machines currently in 
LRL computations as lonp as the read and write operations are in 
the same density. Various machines at LRL have the following 
word ienfrths: Stretch (6^!)^ 709 is ( 36) , 3600 (US), 6600 (6o), 
PDP-] (l8). The Larc aCGeptr4 only a BCD tape (even parity), and 

l/0_ Je yjjjes avai iable 

'Ibo frl'njU:p ip^-.; ;,|,,>^.p.-) f.r,,jj pjr,py,^ j ^ attached to the PDP~1: 

Maraietlc tar>es fPottei ) Four tape units are available to read 

f-cMannei j.ow rienLiity v.iBizueti c tape reels of P^tOO' caoacity. 
This Pives p. capacity of 1.92 minion PDP«1 words exclusive 

3 - 

of record paps for each tape^ A record pan occupies 
mrprcxirpntoly th'/' r=a!r.«' srmce ae fifty iS-bH- words. The 
right hand tape unit is a read only unit utilized for systems. 

Magnetic tapes (TBM) Two tape xmits are available to read 

T-channel magnetic tapes at densities of 200, 556, or 800 
bits per inch. Capacities are proportionally greater so 
that the high density tape vill hold about 7.68 million 
PDP~1 words, again exclusive of record paps. 

Card reader-punch (IBM model 1^102): Readin;^ speed - BOO cards /min. ; 
punching speed - 250 cards /min. 

Line printer (Anelex); Prints at 600 or 103^+ lines/min. , 
iciu cfiar. / line. 

Typewriter (Soroban): 9.5 char. /sec. input or output. 

Pipital plotter (Celcomp 565): X,Y plots on a 10" drum in ,01" 

-,^ -A. 'J -■) — — -•— 4- 

ovcuo (ik< .J.J 1110. PCI j.)UA.^^^^ 

Paper tape reader (Digitronics ): Optical reader for 8-channel 

paper tape, Can handle reels up to 10,5: diameter. Fan- fold 
containers are 5x7" and will hold roughly 1/3 to 1/2 of a 
1000' box. Tape is read at UOO lines/sec. 

Paper tape punci'; (Tr-lpty-ie ) : Punches B-chennel fan-fold paper tape 

at a rate of 63.3 lines /sec. Punches from 1000' box container. 

%lar tape nunch (Friden): Punches 8-channel mylar tape at a rate 
of about. 10 char, /sec. Innut and outnut are held on spindles. 

Titmd tablet: I'toaofmii&B X»Y coordinataa x>f manually held metallic 
pen usually operated in conjunction with the visual CRT. May 

- u _ 

also be used with paner overlays. Resolution of 100 points/inch 
or. P 10. 2h X 10.?!i" metqllic grid. 

Visual CRT (DEC type 30): Point display only, except for horizontal 
and vertical grids with a 50 microsec. settling time. Vectors 
and characters must be constructed from points. This is a 
zero centered grid of size 102U x 10?U. 

Precisir^n CRT (TEC type 31): This tube is inte/^ral to the Fireball 

system, and is currently a Litton J.itlOS. A partially reflecting 
membrane splits the output lig^ht in+o a primary and reference 
branch with appropriate optics into which 35inm. or Ux5" film 
may be placed. Light passing through this film is sensed by 
rriC.ins of a photomultiplier tube ...a J cither returned to the 
computer as a yes-no pulse or else sent through an analog to 
digital converter for density readings. The yes-no response 

renui "rSS H-ntir-nifl tnnf.f='1 v SO mi nrnQf^nnnnc •not" nnin+ fm« a 

i*096 X ii096 matrix. Persons interested in using the Eyeball 
should refer to other nnpers on the sub.^ect, and should also 
hold direct discussions with both the programmers and 
Eyeball engineers. 

Analog to Digital Converter (Redcor 632): Specified as a UOKC 

converter, this is tised at a 25KO conversion rate {kO microsec.) 
and returns 10 bits + sign. It is currentl^y operational only 
on the single Eyeball, and has not yet been completed in 
the dual system. Accuracy is in the order of 10%, although 
this figure must be defined carefully. 

Mouse (Knglebart bug) (LRL design): This is a nreliminary model of 

- 5 - 

a mamJal device that nay be rolled about to reflect an 
X,Y position on the visual CRT. As such, it has properties 
reflecting those of both the lipht nen and the Rand tablet, 
I'he resolution is 102i) x IO2I4. flampling is accomplished 
by a finp;er-OT)erated microswitch. 

Analog to Digital Converter (LRL model): This converter handles 
30 bits at a 33.3KC conversion rate. One channel is 
currently used with the mouse, and the other three are still 

Eyeball Film transport (Vought); Capable of reading 35mr film from 
reels of 3,5" diameter (100' reels) at an actual recommended 
fsneed of about 8 frames /sec. Tbi^ r>in registered transnort 
is used on the Eyeball system, usually on the primary arm. 

Camera Film ( Vought): A pin registered transport for unexposed 
35mm. film. Advances at a rate of 20 frames /sec. 

Uhsprocketed Eyeball transport (LRL design): Capable of moving 
unsprocketed film for Eyeball reading. Pecently completed 
and designed to replace the old Vought transport, the full 
properties of this device are not yet known. 

Light pen (LRL design operating through a DEC 370 nhotomultiplier) : 
This is a light-weight tip— switch desifpi attached to the 
visual CRT. 

Light pen (DEC): An older model operated by a foot pedal is still 

available. There is little reason to use this earlier version, 

polaroid camera: Desipned to lock into the reference arm of the 
Eyeball, this may be used to obtain immediate photogra'ohs 

- 6 - 

from the precision CPT. 
Voice input (LRL desiprn): A sound pover telephone of conventional 

design has been attached to feed into the Redcor ADC to 

allow digitizing of sound signals. 
Un a vailable _I /0_ 

As important as the I/O that exists on the machine are those 
devices and features that are absent. The PDP-1 is unable to read 
DEC tape, opaque charts, film larger than ^x5", or disc packs. As 
mentioned previously, the anount of memory mani-oulation is limited 
by core size, arithmetic capability, and lark of disc storage. 
Faster graphics are available elsewhere in the form of the DD80 , 
and further nlans are being developed in other areas for graphics 
capability including opaque reading. 
General I/O description 

The PDP-1 has what would now be termed a single level 
priority interrupt and which is called the Sequence Break. When 
enabled, the I/O sends a completion pulse which traps the computer 
to location zero with appropriate register storage. The conditions 
relating to the various I/O devices are determined through inter- 
rogation of the status registers in which specific bits in an 

xu uxxj ixt:xu ttxe acZ ux' uxeSUeu i/U xnuludot: out: cah-uo t5T,H,i#t; Oi uiie 

device. For example, a "one" in bit 3 of status register zero 
indicates that a typewriter key has been struck. The typewriter 
is also one of the devices connected to the sequence break so that 
the user has the option of either trapping to the sequence break 
or alternatively staying outside of the trap mode and periodically 

- 7 - 

checking bit 3. Other devices attached to the sequence break are: 
paper tape reader (except for the read-in mode), line printer 
(printing and spacing), card Dimch, card reader (buffer), light pen, 
and IBM magnetic tape units (end of word count, and .lob terminated). 

From the user's viewpoint, the I/O may be broadly divided 
into three major groups as follows: 

Group 1: Devices attached to a high speed channel (IBM and Potter 

magnetic tapes). 

In this group, it if? necessarj'- to specify an address and a 
word count. Data is then streamed across in a block transfer mode 
and the computer is freed for further calculation. The channel 
operates on a cycle stealing basis which is, in general, unknown 
to the user. The Potter tapes are canable of reading or writing 
a non-contiguous record, which is to say that the record may be 
scattered through different sections of memory. The IBM units 
lack this property. To date this has not been very useful, and is 
rarely implemented. 

Group 2: Devices having their own controller (Card reader-punch, 
printer, typewriter, visual and precision CRT, high speed 
paper tape reader and punch;. 

Into this groun we have gathered the I/O that operates on 
a wait or proceed basis. Tiiree command sets are generally 
available to these devices: a wait command, an enable and proceed, 
and a proceed. These commands take the form respectively of ( )w, 
( )C» and ( ). The wait command causes the machine to pause 

. 8 - 

indefinitely while waitinft for an I/O completion pulse. The enable 
coinmand trippers a coirpletion pulse and continues, while the proceed 
coirar.and simply initiates the I/O and continues leaving any timing 
considerations to the programmer. 

Group 3: Devices attached to the spider, a general purpose, low 

speed control channel (Light pen interrupt. Rand tablet, Friden 
pxmch^ Cal Comp, and Mouse), The IBM Selectrics formerly 
attached have since been removed. 

The Spider is a multi-addressing device operating through a 
low speed channel. A word 5n loaded into the 10 rp-^i^-ter of the 
PDP-1 and then sent over the low speed channel. The first two 
octal digits are decoded to indicate the specific device. The 
state of the device is indicated by a skip- no skip operation 
following the low speed channel call. The low s-need channel 
command destroys the contents of the 10 register so Spider 
programming must be carefully implemented. This approach has 
allowed high expansibility for minor I/O devices with a minimum 
of design engineering. 
Fundamental Arithmetic Concepts 

The PDP-1 has an l8 bit word. The first bit acts as a sign 
bit so that numbers from to 3TTT7T are regarded as positive, and 
numbers from itOOOOO to 111111 as negative. 

The data word 111111 is eouivalent to a minus zero in the 
PDP-1. T>iis follovrs from the arithmetic which is one's complement. 
That is to say that minus numbers in the PDP-1 are obtained by 
reversing all bits in the word. Thus, to change a +3 to a -3, we 
have +3 (000 000 000 000 000 Oil) becoming lllllk (ill 111 111 m m lOO), 

- 9 - 

Counting by ones from -3 to +3 would p;ive the seauence 
77777)4 J77775 ,777776 ,0 »1 ,2 ,3. 

Arithnetic operations ]n the machine will usually reset 
777777 to so that one must deliberately create the number 777777, 
either by a negative shift operation, or more readily by the ORed 
operate commands CLA CMA. The result is that -0 is an excellent 
number to use as an error indicator or sentinel since it cannot be 
accidentally derived arithmetically, Hovever, care must be used 
in all operations involving -0 since many commands operate in an 
abnormal manner with respect to this number. 

One's complement arithmetic is a tricky but arithmetically 
sophisticated system, which Is only funy n-n-nreci ntpd 
after extensive use. For examnle, if we take a -5 (777772) -md 
add a +2, the result is a -3 (77777*4). Note that the number scale 
has moved upward in the minus ranpte. This becomes even more 
apparent if we take the number 377777 and add 1 so that it becomes 
UOOOOO (-377777), In effect, we have wrapped aroiaid the word. The 
overflow condition reflects this logic, for overflow is obtained 
on the transition from 377777 to i*00000 (a change of the sim bit) 
rather than when 777777 goes to 0. In the ADD oommard, the overflow 

numbers yields a rerult of the opposite r,ign. Similarly in the 
s^'btract comrruid SUB, the flip-flop is set, when two unlike-signed 
numbers are subtracted, if and only if the sign of the result does 

- 10 

B oo !/ ■ nr c o mrn rin d r. 

The Boolean fimrtionf? ^irp obtairied by the lopi r:n.l comiranf^s 
AND, XOR, TOR and the comnl'^tnent instruction CMA from the opf;rate 
froup. In theory, these commands alone would suffice -for profrairiming, 
hut in practice they are I'lnch more limited. The primary use of the 
AND instruction ia as a mask. An AND (7TTT), for example (actually 
progrnmmnd as AND N snd N, 7T7T) will Insure that the accTimulator 
contains only the rirht 12 bits, which is to say the address field. 
Wie PDP~3 has linuted byte manipulation. One may deposit either 
the address or instruction part of the word through the DAP or DIP 
instructions (the latter is almost useless). The entire word must 
be loaded, however, and the TO is immune to such manipulation 
or masking. 

The exclusive OR (XOR) is used most frequently to reverse 
the sign bit. It also has some use for reflection in graphics 
work. Reversing the first bit reflects about the center line, 
reversing the second bit reflects about a quarter screen line, and 
so forth. 

TlTe inclusive OR has only limited and specific use. 

The CMA is used to change sign on the entire word. Thus, if 
N is in the accumulator, We CMA comm.and will result in -N in 
the accumxilator. 
Mi ct o in Rtru rations 

Both the skip proup and the operate proup on the PDP-1 belonft 
to the class of commands known generally as microinstructions. This 
means that the programmer can assemble his own command within a srroup. 

- 11 - 

and in effect have the -port-lons operate concurrently without loss 
of time. ITiis is accomplished in either assembler simply by placing 
the commands next to each other with a space between. As many 
commands from one group as practical may be ORed in this manner. 
Examples are as follows: 

SPA SPI (5 microsec.) Skip if either accumulator or 10 

is positive. 
CLA CMA CLI (5 microsec.) Load 777777 into the accumulator 

and clear the 10, 
Skip commands may not be ORed with operate group commands. 
Indirect comman ds 

The PDP-1 has multi-level indirect addressing. Care must 
be used not to cascade past the desired nvimber. The following 
sequence will load the contents of storage location 7777 into 
the accumulator rather than loading the contents of A (17777) 
into the acciimulator. 

LAC I A A, 17777 

A correct use of the indirect command is as follows where 
the contents of B are loaded into the accumulator: 

B, 777777 
An alternate possibility is the use of the execute command, 
XCT, to obtain single level indirect addressing. Thus, the above 
coiild simply be written: 


B. 777777 

- 12 - 

The indirect has a special meaninr; when used with the LAW 
command, namely it is used as a minus sign. By means of either 
assembly system, LAW .1 N will load the accumulator with the 
one's complement (minus) value of N, whereas LAW -N will load the 
contents of the location -N, For example, LAW I 3 loads the 
accumulator with 77777^ while LAW -3 loads the contents of 
location 777,^. 

The indirect command is used with the skip ^roup to indicate 
a reversal of the skip. SZF I 2 is interpreted as don ' t skip on 
flag 2. Any skip command may be treated in this manner. 
Unlisted commands 

Several commands are available Jn the assemblers that are 
not listed in the manual. Among these are the following: 
NIX Y (Negative Index) op code hk (10 microsec. ) The C(y) are 
replaced by C(y)-1 which are left in the accumulator. The 
previous C(AC) are lost. Overflow is not indicated. If 
the original C(y) equals 777777, the result is 777776. 
The NIX command is wired directly into the PDP-1. 
SNA (Skip on non-zero accumulator) op code 650100 (10 microsec.) 
If the accumulator is not zero, the program counter is 
advanced one extra position and the next command in 
sequence is skipped, 
SNO (Skip on non-zero overflow) op code 65IOOO (lO microsec.) If 
the overflow flip-flop is not zero, the program counter is 
advanced one extra position and th* next instruction in 
sequence is skipped, 
SMI (Skip on minus 10 ) op code 652000 (lO mi.crosec. ) If bit 

- 13 - 

of the 10 is a one, the profrram counter is Indexed one 
extrp, position and the next instruction in sequence 
is skipped, 
SNS (Skip on non-zero svitch) op code 65OONO (650010,650020, ...) 
(10 microsec. ) If the selected sense switch is non-zero, 
the program cotmter is advanced one extra position and the 
next instruction in the sequence will be skipped, 65OO7O 
will skip if an^r sense switch is up. 
SNF (Skip on non-zero flag) op code 65000N (65001,65002, ...) 
(10 microsec.) If the selected program flag is non-zero, 
the program counter is advanced one extra position and the 
next inBtruction in the sequence will be skipped. 65000? 
will skip if an^ sense flag is turned on. 
Revised commands 

LAT (load accumulator from test word) op code 762200 (5 microsec.) 
Loads the contents of the test word switches into the 
accumulator. It is no longer possible to OR the accumulator 
with the test switches. 
PDP-l Systems 

The systems tape is customarily kept mounted on the right 
hand Potter tape unit, which is used as a read only unit. Depressing 
the read-in switch at the console causes a loop paper tape to boot- 
strap itself into the upper part of memory and execute. This reads 
the next record from the system tape 0, and from that record determines 
the distmice to the nearest loading-routine record on the systems tape. 
The systems tape searches for the loading record, locates it, reads 

- 11^ - 

it into memory, and executes the loader. This causes the on-line 
typewriter to produce rhe statement ^^e^tD. At this point, the 
user types a three letter code followed by a slash. The loader 
searches its memory for this code, determines where that routine 
is located on the systems tape, moves to and reads that record, 
and begins execution of the called- for routine. The three letter 
codes accepted by systems are the last three alphanumerics typed 
prior to the slash. An IllegaJ^TD statement can be caused either 
by erroneous typing or by lack of that routine on the systems tape, 
A list of these codes is maintained in the PDP-1 console area. 

It is important that the read-in switch is not hit at any 
time while the systems tape is still ir.cvlr.f;. To do so may cause a 
runaway systems tape, in which case the machine must be stopped 
and the systems tape rewound at the tape unit. A few codes cause 
fairly complex tape movement beyond that described above, so this 
precaution is important. 

Codes that are read into memory from systems load data over 
existing codes. Unless the systems programmer specifically named 
a zero location, unused portions are skipped over during the read. 
For this reason, it is usually advisable to initialize memory with 
a clear memory command (CLM/) from the systems tape. 

The PDP^l lacks a compiler primarily because of memoiy 
limitations. Proprems must be written using either the old assembly 
language PAP, or the newer language PAL. The latter is documented 
in the I/O manual. Some —but not all— of the fundamental 

- 15 

differences betveen PAP and PAL are as follows: 


No. of passes in assembler 
Symbolic notation for "self" 
Maximum alphanumerics in name 


Approx. 18 

Location definition format (zero) NAME 

" " " (non-zero) NAME N 

Designation for indirect /^ 

Designation for new origin ORG ( ) 

Designation for last card END 

Decimal integer format (none) 

Card image given by listing Yes 



6 (first & last 3 
of a longer string) 




*( ) 

(none required) 


No (.justifies into 

In general FAL is slightly faster and has considerably 
better diagnostics, but is more inconvenient to use because of 
the listing format. 
Loop building 

The most efficient loop on the PDP-1 is built with the ISP 
command. To flow through a loop N times, we use the following: 




lAW T (N) 

(ars des-fr^-d) 

where (n) is the actur^I positive inte.r-^r d"s-"red for the loop. 

- 16 - 

The negative of thr IRP rornnFind does not «»xint. Hovever, an 
equivaJent maj' bo obtai.ied through a prograimning trick. For 
example, let us bujld'a loop where N has a value of 3: i.e. the 
loop executes 3 times. Note that the order of the exits is now reversed. 

COUNTO, -3:U00000 



















COUNTO contains the PAL format for the exclusive OR of -3 
and UOOOOO. Note that the ISP is tested first, and then executes 
the loop. 

The LAC . , . , . , IDX portion of the above loops is not 
essential to the loop, itself, but rnther is used to illustrate 
an extremely common feature of these loops without which the 
mechanism is usually trivial, 

A slightly slower loop, but more efficient in certain rases 
may be built by storing a positive count in N, and comparing it 
with another number by means of the SAS or SAD command. 























- IT - 

Subroutines (General) 

Although the following conventions are not universal, they, 
form an outline to at least msny of the available subroutine 
packages, and are highly recommended as a guide to interchangeability. 

1. Subroiitlne packages are given a name between 3 and 6 
letters in length. All internal subroutine names begin 
with the same three letters, preferably the first three 
letters of that name. 

2. Subroutines should not require preloading of special 
locations, but should enter by a calling sequence as 
will be shown later. There should be at most two 
entry points — one is far preferable — and the number 
of exits should be held to a mininium. These exits 
should be of the form Exit, Exit+1, ... and should not 
.lump elsewhere from inside the routine. Entry and exit 
with AC and 10 in fixed format are acceptable* 

3. First card of the subroutine should be a comment card 
giving the name of the subroutine and its date. Subsequent 
comment cards should suffice to define the properties of 
the subroutine. 

it. Second to last ccmment card in the front should list the 
internal subroutines required. This may have the form; 
/ SSR - (Subroutine l), (Subroutine 2), ,,. 
5. Last of the initial comment cards should give the calling 
sequence. 'Thus, in the subroutine example on page 19: 
/ Jsp Beta/ FWA/ WC 

- 18 - 

6. Cards before and after the calling sequence card above 
should give conditions on the AC and 10, or the nature 
of the exits. 

Subroutine Forma ts 

Subroutines on the PDP-1 are connected either by a JSP 
command or a JDA. In both commands, the contents of the accvunulator 
are replaced by the contents of the program counter plus one. Tlie 
difference is that a JDA saves the former accumulator contents in 
the location of this address field, and the accumulator contents 
are lost on executinp: a JSP. The CAL instruction is completely 
useless and is entirely equivalent to JDA 100. Use of a subroutine 
call is illustrated as follows : 





The original accumulator is lost. Exit becomes changed to a 


(Accun, loaded here) BETA, 


The original accumulator is saved irf BETA; the machine transfers 
to BETA+l; and EXIT becomes JMP ALPHA+1 as before. 

A calling sequence may be constructed along the same lines. 
The following sequence would initiate a subroutine that requires 

- 19 - 

both First Word Address, and Word Count ais input: 








Packaged subroutines 

Numerous subroutines alrea<ty exist as decimal decks. The 

majority of these have been written in PAP, but many are being 

converted to the PAL form. In general, conversion is not complex. 

Also available is a code which will input a PAP decimal deck and 

produce a packed version of PAL for decimal input. However, it 

should be noted that the result is not relocatable. 

The following routines represent some of the more useful 

packages , although there are also many other more specialized 

subroutines. The following abbreviations are used: 

FWA First word address 

WC Word Count 

TN Tape number (logical) 

PN Number of records 


- 20 - 

T. Tape handling 

A, IBM tape tmits 

1. JSP TWRITE/ Clfiss one command*/ FWA/ VC 

Writes IBM tape with variable WC, FWA, TN and density. 

2. JSP TREAD/ Class one command*/ FWA/ WC 
Reads IBM tape as above, 

3. JSP WINDI/ SFR NOO/ Sentinel 

Rewinds tape N and unloads or not depending on the sentinel 

h. JSP IBACK/ Class one command*/ RN 

Backspaces a given number of records 
5. JSP READBCD/ Class one command*/ WA/ WC 

Reads an even parity BCD tape. 

B. Potter tape units 


Writes Potter tape with variable WC, FWA, and TN in 
low density. 

Reads Potter tape as above 


Rewinds Potter tape unit N. Mechanically incapable of 

Backspaces Potter tape a given number of records. 

The class one commands control density and tape number, i.e. 
SHD 300 is high density on tape 3, Refer to I/O Manual for 
further details. 

ll. Printer 

1. JSP POCTAL/ FWA/ W. Number of columns 

Prints octal storage of size WC starting at FWA In format 
with variable number of columns. Zero lines are not 

2. JSP POCKETA/ FWA/ WC/ Number of columns 

Similar to Poctal except that zero lines are suppressed, 

3. JSP DOVER/ FWA/ WC/ One-third number of initial blanks 
Printer comments written in XS3 and stored in memory are 
picked up and printed with indicated indenting. 


Initializes printer buffer 
III, Cal-Gomp 

1. JDA CALRYT/ N/ Size/ 

Writes numbers on Cal-Comp. N is the number desired; Size 
is an artificial scaling constant; and is the number of 
90 counter-clockwise rotations. 


Plots a vector from an initial point to X,Y, Z controls 
pen movements. 

3. JSP CALSO/ Number of times for operation 

Entered with plotting operation in 10 and repeats N times. 
IV. typewriter 


Types comments from concise code In memory. 

- 22 - 


T^'pewrlter to printer. Permits progrsunipr to nnnotate 
printout ns desired. Ptrikeovers and backup are perrrltted, 


Octal storage to decimal typevriter, 
h, JBP TIN/ Ircorporated subroutine/ WC 

Inputs decimal numbers froi" typewriter vhile simultaneously 
maintaining an auxiliary subroutine such as a display. 
Available with and without sequence break, 
V, Paper tape 

Punches six by five matrix on paper tape. 
VI, Card reader- punch 


Punches left 12 bits in word as a vertical column. Data 
destroyed as punched, 


Reads cards and assifms a weight to each punched position. 
Useful as part of a Hollerith read, 
yil. CRT and Kyeball displays 


Fast Eyeball scan and playback. 


Fast Analog Digital Converter display of X-sweep, 

Displays a N x N prid. 

- ?3 - 

h. JSP DCM/ FWA/ WC/ X/ Y/ Rize 

Displays comments from memory. Supplementary subroutines 


Displays fast circle centered at origin, based on incremental 

algorithm, and with radius from accumulator. 
VIII, Light pen 


Fast scan of Eyeball with light pen pickup of single point, 


Light pen tracking routine. Requiren initiaJ ization and 
has multiple exits. 


Light pen switch. Displays contents of a bank. 
IX, Arithmetic routines 


Locates maximum number in bank, returning both number and 


Locates minimvim number in bank, returning both number and 


Random n\imber generator. 

Extracts square root, 

Projects X,Y,Z into X,Y plane. 

- 2U - 

X. Prograinming aids 


An internal routine allowing the program to call directly 
from the systems tape without going through read- in mode or 
the typewriter. 


Pseudo-index which is followed by command to be indexed. 
Not valid for skip-type commands such as DIV. Index 
itself is a separate memory location manipulated according 
to normal rules. 

Clears a bank of memory. 

k, JSP BLOCK/ FWA old block/ FWA new block/ WC 
Moves a block of nffimory. 


Internal debug memory dump, 


An internal memory dump printing eight words /line with 

7. JDA FILLUP/ FWA/ WC of filled locations/ Delta on word/ Delta 
on position number. Will generate almost any form of 
linear format. Particularly useful for building checkout 


Displays rif~htsix bits of accumulator as flags. Useful as 
visual numerical counter on console that can be independent 
of calculations. 

- 25 - 

9. Memorize-Optior -Recall package. 

This lies outside the normal format, but permits the programmer 
to make a sequence of decisions from the typevrriter, which 
sequence is then held in memory for subsequent runs. Permits 
program adjustment to know ledge ability of user. 
XI, Timing routines 

1. JSP MMS{Y) 

(Y) is a number divisible by 5 between 20 and 100, and 
represents the number of microseconds delay desired. Routine 
included in timing. 


N is the number of 100 microsecond delays desired including 
XII, Parity routines 


Checks for odd parity, and turns on flag 1 for parity failure. 

2. JDA PARADD/ Parity bit desired 

Adjusts words to odd parity through adding desired bit pattern. 
XIII, Conversion routines 


Converts BCD even parity magnetic tape to Friden Flexowriter 
format. Enters and leaves with character in accumulator, 


Interprets a group of BCD numbers starting at FWA as one 
octal word. 

- 26 - 


Same as above except that XS3 numbers are used, 

Enters with octal in 10} exits with decimal BCD in AC and 10. 
5. JSP 0CTXS3 

Converts octal to XS3. Particularly useful for print routines. 

System packaf^es 

There are many routines on the systems tape that are particularly 

useful for general purposes. These routines are described on sheets 

found hanging near the machine. Among the more useful general routines 

are the following: 

ASP/ General magnetic tape to ptmched card routine. Inputs magnetic 
tape of low or high density and with variable record lengths. 
Will strip a variable number of lead words per record and punch 
the rest with a variable number of columns per card beginning in 
column 1. Right 12 bits of each PDP-1 word are turned vertically 
and punched as a binary column of four octal digits with the 
most significant 'niunbers at the top. This may be used in con- 
Junction with JAB/ to go from paper tape to punched cards. In 
general, this cannot be accomplished without unscrambling on a 
different machine, 

BTY/ Described previously under typewriter. Allows comments to be 
typed directly onto printouts. 

CIM/ Clear memory, neeommended for use prior to any routine. 

- 27 - 

D^G'/ Memory dump. The state of the console should be noted before 
using. In particular, one usually wishes to record the 
program counter, accumulator, 10, program flaps and any 
unusual light indicators such as the defer. The locations 
above 7000 octal are not obtained. Note that the memory 
dump does not destroy lower memory, so that It is often 
possible to take a dump, make a manual console change and 
continue running. The BTY mentioned above can often be 
alternated into the same upper section of memory In this 

ITP/ May be used to print BCD tapes on the IBM tape units. 

LEG/ Loads binary cards. Note that certain upper locations can- 
not be loaded. Binary cards lacking data will often not load 
correctly, and should preferably be stripried from the binary 
deck. Those cards lacking any punches in col. 7 or beyond 
may be discarded without loss. In many cases, it is then 
necessary to enter the starting address (SA) in the address 
keys, hit stop and then start, 

JAB/ Punched paper tape to magnetic tape. This routine is 

designed as a data pump with no attemrot to unscramble the 
user's format, but rather to give him a magnetic tape that 
may be transferred to another rrachine. Variables are 
tape density and record length while words may either be 
packed or xinpacked. 

OTD/ Octal tape dump may be used for printing an odd parity binary 
tape written in any density. 

- 28 - 

pal/, pap/ Assembly routines previously discussed, 

PAR/ Equations input from the typewriter are displayed. Values 
and variables may be modified. 

PBC/ Punches binary cards from memory. A given program may be 
stopped in mid-run, the location of the stop placed in the 
test word, and a PBC executed, Ihe resulting deck will 
restart the code from the stopped location provided that 
the code is entirely contained in locations to 62U6, 

PM8/ Assembly routine producing paper tape binary input for 
the PDP-8. 

VAP/ General purpose assembly system, 

CRT Pro|Trammin p 

Both the visuail and precision CRT have 0,0 at the center. The 
visual CRT has 102*4 points along each axis and the precision CRT 
has U096 addressable points per axis. In both cases the word is 
packed in the left 12 (lO) bits of the PDP word. Since bits to the 
right of this field are ignored by the deflection registers, and 
since commands for the two CRTs may be ORed together, it is easier 
to consider the system as if both CRTs were addressable to h096 
(10,000 octal) and program in this manner, ignoring the fact that 
the hardware does not register this significance on the visual CRT, 

The most negative X to the left is UoOOOOc,, and the largest 
positive X is 377700g. Similarly the most negative Y is UOOOOOg 
at the bottom of the screen and the largest value at the top is 
37TT00q. To continue incrementing on the largest number will 
simply wrap around the CRT and reenter at the left (bottom) side. 

- 29 - 

To step one position we add a delta of lOOg. Notice that in 
crossing 0, this causes a move from 777700 to 000001. This is 
not objectionable, since the display is identical to that for 
000000. Under cert,ain conditions, however, this could accumulate, 
and introduce an error. To decrement a position, it is most 
convenient to subtract 77 (edd 777700), This has the advantage 
that it moves from to 77T700 rather than to 777677 as would have 
been the case had we subtracted 100. Note that the two points in 
question would then display as 7777 and 7776, respectively, which 
are npt identical. 

In displaying a point, we place X in the accumulator and Y 
in the 10, and follow with an appropriate display command. We may 
either use the wait form, such as a DPHW DVHW, or else we may count 
on a minimum time in the program of 35 microseconds before returning 
to our display and use the display and T?roceed commands r^xirh as 
DPH DVH. Further timinpr conPiderntions apply if we intend to read 
a position from the precision tube. Since our registers are set 
up in this manner, it follows that most scans should be prop-rammed 
to step X and only at the pnd of each line to step Y. In this manner, 
we are able to minimize the number of changes in' the 10. 

A fast scan usually requires a delta of about 32.-, to meet 
timing requirements and avoid flicker on the CRT. A non-flicker 
on the present visual system has been observed at a value of 
27 frames/ sec, but this number can vary with the individual, the 
lights and the phase of the moon, so that a number of about 35 or 
even ko frames per sec, is probably advisable for the present system. 

- 30 - 

Although we may physicalJ-y displace points as rapidly a? 35 
tPicroseconds, in actual practice doinfi: useful work, 85 micrnsecondR 
is a fairly ti;?ht programming display loop. If we are reading 
values from the Eyeball, and in particular the analog digital 
converter, about double this figure is necessary. 

A grid command is initiated at a given X,Y and then proceeds 
to the right (top) side of the screen. A useful variation on this 
rule is obtained by sending a display grid and proceed command 
(a DYVH command for example), and then interrupting before full 
drawing time is completed. This gives a shorter vector in the 
desired direction of slightly unpredictable length. When displaying 
points, such a vector may be used to call the user's attention to 
the point display, which may not be obvious. Grid display commands 
containing Y display a vertical line; those with X display a 
horizontal line. All grid display commands reauire an appreciably 
longer timing than normal display commands. The X and Y grid 
commands may be ORed to generate a h^ line, 

A few CRT tricks may accelerate displays. Changing the 
sign of either X or Y reflects the point about the opposite axis. 
For example, X,Y changed to X,-Y is mirrored about the positive 
X axis. As pointed out previously, this is accomplished by 
com.pleroenting Y. Reversing the first bit changes the point by 
one quadrant (half-screen shift), and changing other bits cause 
quarter-screen shifts, eigth-screen shifts, and so forth. Thus, 
to shift a picture from the origin to the lower left may be 
accomplished by reversing the first bits of both X and Y assuming. 

- 31 - 

needless to say, a CRT format packed in the left 12 bits, 
Exchan/^ing X and Y reflects the point about a i*5° axis. 

In a few cases such as a stepping scan, we perform our 
arithmetic with the point in the CRT format. In many cases, 
however, we require more extensive arithmetic operations. For 
this purpose, it is usually most convenient to reverse the first 
bit (an XOR operation against UOOOOO) and then rotate right by 
six places masking out the unused six bits, if we have not already 
done so earlier. This converts our axial scale range from 
UooOOO, 377700 to 000000, 007777. which can be manipulated with 
much greater ease. When we are ready to display, we simply 
reverse the process. 

Translations of a point are readily accomplished by the methods 
above. Unfortunately, no convenient methods have been fotmd for 
rotations, and at present the simplest approach is the brute force 
technique from analytic geometry ~ caveat programmer. 

In constructing program loops for either the yes-no Eyeball 
or the ADC to obtain density responses, much time could be lost in 
waiting for the hardware response. A display command requires 5 
microseconds to execute and 50 microseconds to set the status 
register. From this time another Uo microseconds is required 
before the density is available from the low speed channel. These 
dead spaces should be utilized for useful work; the previous density 
signal may be converted during the settling time of the point, and 
the subsequent point coordinates may be constructed during the 
conversion time of the analog signal. 

- 32 - 

The manner in which the analog digital converter acts may 
be illustrated by the following table and illustration; 





Resulting sifTial 




Voltage level 






The correspondence between the film and the signal must not 
be taken as absolute, since it is a relative matter that can vary 
from day to day and fluctuates somewhat. Relative densities are 
significant; absolute densities are not, Conseouently , it is 
necessary to have nrogranming to determine the -nroper ADC rancre in 
which the program shall run. A subroutine such as TOADX may be 
used for this purpose. 

It is clear that to convert the density to an arithmetic 
number, we may use the same technique as applied to the CRT, 
namely, reverse the first bit except that now we rotate 7 places 
right instead of six, because the field is only 10 bits plus sign. 

- 33 - 

not 12. Thin results in a positive number in which is clear 
film and 1777 is the darkest possible; in other words 102U 
grey levels. 

In passing, it might be mentioned, that the ADC formerly 
yielded a signal ranging from a few digits positive through 
UOOO. This utilized only half of the full conversion range. 
It is still possible to set the level in this manner, if desired.