Overview
In computer architecture a word is the natural unit of data used by a particular processor. It denotes a fixed number of bits that the CPU can process in one operation and commonly corresponds to the width of registers, the arithmetic logic unit (ALU), and the typical size of machine instructions.
Characteristics
Word sizes vary between architectures. Modern general-purpose processors most often use 32-bit or 64-bit words, while embedded systems may use 8, 16, or 24 bits. The chosen word size affects several system properties:
- Register and ALU width, determining integer range and arithmetic behavior.
- Memory addressing and the maximum directly addressable space.
- Typical instruction encoding and data alignment rules.
Historical context
Early computers used a wide variety of word lengths; some machines adopted unconventional sizes that matched application needs or hardware constraints. Over time a trend toward powers of two (8, 16, 32, 64) simplified software and hardware design, and compatibility pressures encouraged adoption of standardized sizes across families of processors.
Uses and examples
Word size determines how data types are represented and how many memory bytes are consumed by typical values. Terms derived from "word" are common in programming and documentation: halfword (half a word), doubleword (two words), and quadword (four words). Systems that are word-addressable index memory by words rather than bytes, while most modern machines are byte-addressable.
Notable distinctions
It is important to distinguish the architectural word from related concepts such as the computer address size (which sets the address space) and the byte (an 8-bit storage unit on most systems). Endianness and alignment rules influence how multi-byte words are laid out in memory and can affect portability and performance.
Understanding word size helps explain limits on integer ranges, pointer sizes, and the interface between hardware and software across different computing platforms.