Instruction Format

of electronic computers; the number of addresses in an instruction. The instruction format is determined by the nature and coherence of the operations being carried out, the length of the machine’s word format, and the capacity and structure of direct-access storage. Instructions are divided by number of addresses into multiple-address, single-address, and zero-address instructions. In multiple-address instructions, the instruction format may be uniform or nonuniform. The uniform instruction format is characterized by identical structure and length in all addresses of the instruction. With the nonuniform instruction format, the same part of the instruction contains complete addresses (based on the entire capacity of the direct-access storage) and short addresses (linked with part of the direct-access storage). In many cases the short addresses pertain to special ultrahigh-speed direct-access storage units of limited capacity. For general-use computers (for example, the BESM-6 or the Ural-16), single-address instructions are most common. They make it possible to expand the working section substantially or to assign two instructions to each cell. The extreme case is the zero-address instruction (zero instruction format), in which the address part in the instruction is entirely absent, and operands are either assigned to the instruction itself or specially assigned to the direct-access storage and are selected according to a previously determined law. The zero instruction format is most frequently used in combination with other command structures (for example, in the American Burroughs 5000). A distinction is made between the constant instruction format, when the number of addresses in an instruction is set during the design of the computer, and the variable instruction format, which foresees operational change in the instruction format during the process of computations depending on the nature of the procedures being carried out (for example, the American Burroughs 8501). The variable instruction format substantially increases the performance of the computer, but its realization involves design and circuitry complications. The variable instruction format is implemented most easily during sequential and parallel-sequential data processing as, for example, on the Minsk-23 computer.