computer architecture

Imagine you are designing a car, but you get to split the job in two. First you decide what a driver will ever see and touch: a steering wheel, three pedals, a gear lever, and exactly what each one does. That is a promise to every driver, and it must stay the same even if next year's engine is electric instead of petrol. Computer architecture is the same kind of work for a machine that runs programs: it is the study and design of the machine and, above all, of the contract between the hardware and the software that runs on it.

More precisely, the word architecture, in its narrow technical sense, means the instruction set architecture, or ISA: the complete list of operations a processor promises to understand (add these two numbers, load this value from memory, jump to this place if a result was zero), the registers a program can name, how memory is addressed, and how numbers are encoded. This is the abstract contract that software sees. A program written to this contract will run on any chip that honours it, today or in twenty years.

It matters because the contract outlives the silicon. Programs compiled for a given ISA decades ago can still run on a brand-new chip with the same ISA, even though nothing inside is the same. Be careful, though: people use computer architecture loosely to mean the whole field, including how the chip is built inside. In this domain we keep the sharp distinction between architecture (the contract software sees) and organisation or microarchitecture (the hidden way a particular chip delivers on that contract).