Software Development at the Eckert-Mauchly Computer Company Between 1947 and 1955

Iterations – Norberg – Software Development at EMCC 

 18 

The structure of this compiler resembled the earlier compilers with the 

added feature of the Pseudo-code.

50

 The set of easily accessible 

subroutines was contained in a library in alphabetical order. Information 

from tape was read in the same sixty word units, called blocks. Arithmetic 

was done in floating point. Data was expressed in “two-word” form, 

which represented the complete numeric quantity to be expressed. The 

first word contained a numeric quantity without the decimal point and the 

second word gave the information for the placement of the decimal point. 

The data and instructions occupied the same block locations as in the 

previous compilers.  

 

In solving a problem such as  

 

 

 

 

Y = e

-x2

 sin cx 

 

Where x ranged between –0.99 and +0.99 in increments of 

∆

x = 0.01, y 

needed to be determined for each of 199 values of x throughout the range. 

Consider only one such calculation in the set to appreciate the flavor of the 

Pseudo-code. One operation was to increase x by the increment 

∆

x and 

test to determine if all the values of x throughout the range have been used 

to calculate the y result required. The description of this operation is 

“ADD to A LIMIT.” Three lines were needed, expressed in a three-

address code.  

 

 

 

1 2 3 4 5 6 7 8 9 10 11 12 

  AAL 

 

(x

i

)   (

∆

x)   (Lx) 

 

 

1CN  0 0 0 0 ( 

≠

 OPN #) 

 

 

2CN  0 0 0 0 ( = OPN #) 

 

The first line required 3 relative working storage addresses. The values 

needed were x, 

∆

x, and Lx, the last being the limit of the range of x. X for 

this problem was known to be in 000, 

∆

x was in 002, and Lx was in 004. 

Thus, the first line was expressed 

 

 

 

AAL 000 002 004 

 

On the basis of this instruction, the routine would add x (in 000) to 

∆

x (in 

002) and place the sum (the new x) back in 000. This new x would then be 

tested against the limit of x (LX in 004) to determine whether or not the 

limit had been reached. All the coder was required to do was send 

“control” back to the beginning and calculate the next y with the new 

value of x. This was the purpose of the 1CN line. The symbols 

≠

 OPN # 

meant that if the new x was not equal to the limit, control would be 

transferred to the operation number (OPN #) placed here. Since actual 

calculation began with Operation 1, this line would read 

 

 

 

 

1CN 000 0 00001 

---

Iterations – Norberg – Software Development at EMCC 

 19 

 

Five digits were allocated to the operation number because the compiler 

could handle up to 99,999 operations. When the limit for this problem was 

reached, control went to the next operation, in this case Operation 11. 

 

 

 

 

2CN 000 0 00011 

 

In all, there were 35 Pseudo-code instructions, including an end coding 

instruction. Figure 4. Later compilers for business program compilation 

received names, Math-Matic (a form of A-3) and Flow-Matic, in response 

to requests of the sales staff. Flow-Matic allowed the user to write 

instructions in English pseudo-code, which UNIVAC I could translate and 

use to generate the program. Figure 5 shows the set of instructions for 

Flow-Matic, a compiler that became an important input to the design of 

the later higher-level programming language COBOL. 

 

Dupont was the first company to use A-0. From there, it spread to the 

David Taylor Model Basin, where Betty Snyder now resided, to the Army 

Map Service, and the Census Bureau, all purchasers of UNIVAC I 

computer systems.

51

 

 

The Mauchly group determinedly tried to convince the community to use 

these techniques. There was the Philadelphia meeting mentioned above 

where Hopper spoke on A-0. Mauchly repeatedly addressed groups in 

various professional settings. Not surprisingly, he participated in the 

symposium on large-scale digital calculating machinery at Harvard in 

1947, where he spoke about “Preparation of Problems for Edvac-type 

Machines.”

52

 In September 1949, he presented details about the UNIVAC 

system to the American Chemical Society. Mauchly was invited to address 

the Chesapeake Section of the Society of Naval Architects and Marine 

Engineers in January 1951 on the subject of the computer’s value in 

various engineering problems of interest to the society. Before the 

American Gas Association-Edison Electric Institute Joint Accounting 

Conference in April 1953, Mauchly focused on the system’s usefulness to 

business and noted a new training program offered by Remington Rand.

53

  

 

Hopper was on the lecture circuit at least as much as Mauchly. In May 

1952, she spoke to the Association for Computing Machinery (ACM) on 

“The Education of a Computer,” a title she used often but with slightly 

revised text each time to keep up with developments in EMCC. For 

example, as we noted above, one talk was to the Symposium on Industrial 

Application for Automatic Computing Equipment held in Kansas City, 

MO, by the Midwest Research Institute in January 1953. Richard 

Ridgeway delivered a paper on “Compiling Routines” to a meeting of the 

ACM in September 1952 in which he did a detailed analysis of EMCC 

compilers. The Census Bureau organized a workshop on coding and 

programming for July 1953, attended by several Remington Rand 

---