Stored-program computer
A computer whose instructions live in the same read-write memory as its data, encoded as ordinary numbers. Because the program is just numbers in memory, the machine can be given an entirely new task by loading new numbers — no rewiring — and can even read and modify its own instructions.
Глубинный архив пока написан по-английски — выверенные переводы входят в план развития. Встроенный переводчик браузера хорошо справляется с этой страницей.
✦ Постойте, правда?
ENIAC, the room-sized machine that came just before the idea, was "programmed" by physically re-plugging cables and throwing hundreds of switches; setting up a new problem could take days. In June 1945 John von Neumann circulated a 101-page "First Draft of a Report on the EDVAC" that described storing the program as numbers in memory alongside the data. The first machine actually to run a stored program was the Manchester "Baby," which executed its first program on 21 June 1948 — the moment software became a thing you could load instead of a thing you had to build.
Что это
A stored-program computer keeps its program in the same memory as its data, both written in the same binary numbers. Nothing about the wiring says "this number is an instruction and that one is data" — the machine simply fetches whatever it is pointed at and treats it as a command. Because instructions are just numbers, they can be loaded, copied, generated, and altered like any other data. That single design choice is what makes a computer *general-purpose*: one physical machine can run a spreadsheet, a game, or a compiler, depending only on which numbers currently sit in memory.
Почему это было важно
Earlier calculating machines were specialized: to change the task you changed the machine. ENIAC (1945) could be rewired for new problems, but "rewired" meant days of plugging cables. The stored-program idea collapsed that setup time to the time it takes to load a file, and it separated software from hardware for good — the machine became a fixed, mass-producible object, and the interesting, endlessly varied part became the numbers you fed it. Every computer since, from mainframes to the phone in your pocket, is a stored-program machine.
Что это открыло
Everything downstream in computing assumes it. Programming languages exist because a program is data a *translator program* can read and rewrite. Operating systems, which load and switch between programs, are only possible when programs are loadable numbers. Self-modifying code, compilers, interpreters, viruses, and the entire software industry all rest on the flat, unpretentious fact that in a stored-program computer, instructions and data are made of the same stuff.
Минимальная работающая версия
A processor that endlessly repeats a fetch–execute cycle: read the next instruction from memory, decode what it means, do it, then read the next — with instructions and data sharing one addressable store. This is the "von Neumann architecture."
Эта статья ждёт своего полного изложения — картографы работают. Её место в графе уже подтверждено.
Нити, проходящие через эту способность
Why couldn't the Romans build a computer?7 шаговИсточники
- — John von Neumann, 'First Draft of a Report on the EDVAC' (1945)
- — Alan M. Turing, 'On Computable Numbers, with an Application to the Entscheidungsproblem' (1936)
- — Martin Davis, *The Universal Computer: The Road from Leibniz to Turing* (2000)
Что-то не так на этой странице? Каждое утверждение здесь должно выдерживать проверку. Предложить исправление →