operandsize

Operand size is the bit width of the data that a processor instruction operates on. It specifies how many bits are read from a source operand, how many are written to a destination, and how the result should be interpreted by the arithmetic or logic unit. Common sizes are 8, 16, 32 and 64 bits, corresponding to bytes, words, doublewords, and quadwords; some architectures support additional or nonstandard sizes via special instructions.

Control and variation: In many CPUs, the basic data path is fixed, but the effective operand size can be changed by instruction prefixes or special encodings. For example, in x86 the operand-size override prefix (0x66) can switch between 16- and 32-bit operands, and in 64-bit mode the REX.W prefix selects 64-bit operands. The default operand size depends on the execution mode (for instance, 32-bit or 64-bit). Other architectures may fix operand size per instruction or expose separate instructions for wider or narrower data, such as ARM’s 32-bit and 64-bit instruction sets.

Impact and usage: Operand size affects instruction encoding length, register selection, and how values are extended when operands of different sizes are used together. When smaller operands are extended to larger sizes, sign extension or zero extension rules apply. Alignment and memory bandwidth are also influenced, as wider operands require more data per operation and may impact caches and throughput. In practice, compilers and assemblers emit instructions with the appropriate operand size for the target data types and platform.

See also: data width, word size, byte, register, instruction encoding.