Code_ The Hidden Language of Computer Hardware and Software - Charles Petzold [89]
But beginning in the early 1940s, vacuum tubes began supplanting relays in new computers. By 1945, the transition was complete. While relay machines were known as electromechanical computers, vacuum tubes were the basis of the first electronic computers.
In Great Britain, the Colossus computer (first operational in 1943) was dedicated to cracking the German "Enigma" code-making machine. Contributing to this project (and to some later British computer projects) was Alan M. Turing (1912–1954), who is most famous these days for writing two influential papers. The first, published in 1937, pioneered the concept of "computability," which is an analysis of what computers can and can't do. He conceived of an abstract model of a computer that's now known as the Turing Machine. The second famous paper Turing wrote was on the subject of artificial intelligence. He introduced a test for machine intelligence that's now known as the Turing Test.
At the Moore School of Electrical Engineering (University of Pennsylvania), J. Presper Eckert (1919–1995) and John Mauchly (1907–1980) designed the ENIAC (Electronic Numerical Integrator and Computer). It used 18,000 vacuum tubes and was completed in late 1945. In sheer tonnage (about 30), the ENIAC was the largest computer that was ever (and probably will ever be) made. By 1977, you could buy a faster computer at Radio Shack. Eckert and Mauchly's attempt to patent the computer was, however, thwarted by a competing claim of John V. Atanasoff (1903–1995), who earlier designed an electronic computer that never worked quite right.
The ENIAC attracted the interest of mathematician John von Neumann (1903–1957). Since 1930, the Hungarian-born von Neumann (whose last name is pronounced noy mahn) had been living in the United States. A flamboyant man who had a reputation for doing complex arithmetic in his head, von Neumann was a mathematics professor at the Princeton Institute for Advanced Study, and he did research in everything from quantum mechanics to the application of game theory to economics.
John von Neumann helped design the successor to the ENIAC, the EDVAC (Electronic Discrete Variable Automatic Computer). Particularly in the 1946 paper "Preliminary Discussion of the Logical Design of an Electronic Computing Instrument," coauthored with Arthur W. Burks and Herman H. Goldstine, he described several features of a computer that made the EDVAC a considerable advance over the ENIAC. The designers of the EDVAC felt that the computer should use binary numbers internally. The ENIAC used decimal numbers. The computer should also have as much memory as possible, and this memory should be used for storing both program code and data as the program was being executed. (Again, this wasn't the case with the ENIAC. Programming the ENIAC was a matter of throwing switches and plugging in cables.) These instructions should be sequential in memory and addressed with a program counter but should also allow conditional jumps. This design came to be known as the stored-program concept.
These design decisions were such an important evolutionary step that today we speak of von Neumann architecture. The computer that we built in the last chapter was a classic von Neumann machine. But with von Neumann architecture comes the von Neumann bottleneck. A von Neumann machine generally spends a significant amount of time just fetching instructions from memory in preparation for executing them. You'll recall that the final design of the Chapter 17 computer required that three-quarters of the time it spent on each instruction be involved in the instruction fetch.
At the time of the EDVAC, it wasn't cost