microprogram

Microprogramming is a method of implementing a computer's control logic by a sequence of small, portable instructions called microinstructions, stored in a control store or microcode memory. A microprogrammed control unit uses a microprogram counter to step through microinstructions that orchestrate the datapath signals needed to execute each machine instruction. Each microinstruction encodes control signals for the ALU, registers, buses, and memory access, as well as sequencing operations such as conditional branches to other microinstructions.

The approach contrasts with hardwired control logic, where the control signals are generated by fixed combinational circuits. By changing the microcode, the same hardware can support different instruction sets, add new instructions, or fix bugs without redesigning the control circuitry. The concept dates from the 1950s and 1960s, with Maurice Wilkes and others forming the basis of microcode theory, and it achieved widespread use in many CISC machines, notably IBM System/360.

Microinstruction formats are typically described as horizontal or vertical. Horizontal microcode uses wide fields to directly control many signals, at the cost of larger control memory, while vertical microcode uses compact encodings that rely on a microdecoder to expand instructions into a sequence of micro-ops, saving control store space at the expense of decoding logic.

In modern CPUs, microcode remains common for implementing complex instructions, managing obsolete modes, and applying errata fixes. Some architectures also use a micro-op layer, where complex instructions are translated into a sequence of simpler micro-operations, which are then executed by a mostly hardwired pipeline. Microprogramming thus provides flexibility at the cost of some performance versus fully hardwired control.