Field Guide · concept

Also known as: ISA

An instruction set architecture (ISA) is the contract between hardware and software: the set of instructions, registers, and data types a processor understands, separate from how any particular chip implements it.1

Fetch read instruction Decode read opcode Execute registers + ALU Write back store result ISA defines the opcodes and registers · microarchitecture decides how the cycle is built
The processor repeats a fetch-decode-execute-write-back cycle; the ISA is the fixed vocabulary of opcodes and registers that this cycle operates on, while the microarchitecture is the separate engineering of how fast and how cleverly the cycle runs.

Overview

The ISA defines what a program can ask the CPU to do — the opcodes, the registers, how memory is addressed, what data types exist — so that machine code written for an ISA runs on any chip that implements it. It is a stable interface: a binary compiled for x86-64 in 2010 still runs on an x86-64 chip made today, because both honour the same contract.

The microarchitecture is the separate question of how a given chip carries those instructions out — its pipeline depth, caches, branch predictors, and execution units. Two chips can share an ISA yet differ wildly in speed and power because their microarchitectures differ. This split is what lets Intel and AMD both build x86 chips, or dozens of vendors all build ARM chips, that run the same software.

RISC versus CISC

ISAs split loosely into two philosophies, though modern designs blur the line internally:

Trait RISC CISC
Instruction count Few, simple Many, complex
Instruction length Fixed Variable
Work per instruction Small, uniform Can be large
Memory access Load/store only Many ops touch memory
Examples ARM, RISC-V x86

The major families today are the CISC x86 and the RISC ARM and open RISC-V. In practice even x86 chips decode their complex instructions into simple RISC-like internal operations, so the distinction is now more about the external contract than the silicon.

Where it fits

The ISA is why a compiler must target a specific architecture, and why a binary built for x86 will not run on an ARM chip without recompilation or emulation. It builds on the von Neumann idea of a stored program the processor fetches and executes. For GopherTrunk this is a daily reality: Go cross-compiles the same source to x86 servers and to the ARM CPU in a Raspberry Pi capture node, each producing native machine code for that ISA.

Sources

  1. Instruction set architecture — Wikipedia, on the ISA as the hardware/software interface and its separation from microarchitecture. ↩

See also