Home * Programming * Data * Word

A Word or Computer Word, is a term for the natural unit of data used by a particular computer architecture. Modern computers usually have a word size to be a power of 2 multiple of the unit of address resolution, likely a Byte, that is two, four, or eight Bytes, which are 16, 32, or 64 bits. Many other sizes have been used in the past, including 8 (a Byte), 9, 12, 18, 24, 36, 39, 40, 48, and 60 bits. Some of the early computers were decimal rather than binary, having a word size of 10 or 12 decimal digits, and some of them had no fixed word length at all.

16-bit Word

Often the size of a word is defined to be a value for compatibility with earlier computers, such as Intel's x86 and x86-64 architecture, which referes a Word from the original 8086 16-bit µ-Processor. Subsequently Intel used the terms Double Word (dword) for 32-bit words, a quadruple word or Quad Word (qword) for 64-bits words, and even a Double Quad Word for 128-bit words. x86 and x86-64 registers may still treated as word registers (ax versus eax or even rax) , while it is recommend to use the native 32-bit double word, because the word-wise access requires a prefix byte to overwrite the default width. SIMD instruction sets like MMX, AltiVec and SSE2 provide operations on vectors of four or eight words inside appropriate SIMD-registers. IBM 360 and successors with 32-bit words, refer 16-bit size as halfword.


On recent 32-bit and 64-bit processors the primitive C datatype short and unsigned short refers to 16-bit words by most compilers for those architectures. In Java, short is guaranteed to have 16-bit. Signed short in C is assumed to use Twos' Complement, but not strictly specified. A Word-type, explicitly type-defined in C, is therefor usually treated as unsigned, also to avoid arithmetical right shift issues:
typedef unsigned char  BYTE;
typedef unsigned short WORD;


unsigned short
C, C++

#include <limits.h>

C, C++, Java

#include <limits.h>



Words stored in memory should be stored at even byte addresses. Otherwise at runtime it will cause a miss-alignment exception on some processors, or a huge penalty on others.


Main article: Endianness.
An issue with words consisting of two or more bytes, is the order, bytes may appear inside a word of memory. According to their usual arithmetical significance, there is a low and a high byte of a 16-bit word, which may either be stored at the lower or higher byte-address in memory. Intel processors were always so called little-endian machines, the least significant byte (LSB) is at the lowest address. Other processors, including the IBM 370 family, the PDP-10 (36 bit), the Motorola microprocessor families, and most of the various RISC designs are big-endian, and store the ‘big-end-first’.

Extracting Bytes

Following C union to extract or synthesize bytes from/in words, is not portable and should be avoided.
union {
   BYTE b[2];
   WORD s;
} u;
u.s = 0xaa55;
assert (u.b[0] == 0x55); // fails, if big-endian
The portable way in C can be done with inlined functions or C preprocessor macros, using arithmetical divide or modulo by 256, aka shift and mask by bitwise 'and' - or for the synthesis multiplication of high byte by 256 plus low byte:
BYTE lowByte (WORD s) {return (BYTE)(s & 255);} // mod 256
BYTE highByte(WORD s) {return (BYTE)(s >>  8);} // div 256
WORD makeWORD (BYTE high, BYTE low) {
   WORD  s = high;
   return (s << 8) + low; // high * 256 + low

See also

External Links

What links here?

Up one Level