Control Unit

The control unit is the conductor of the CPU. It takes each instruction's opcode and turns it into the right set of control signals — the on/off lines that steer data through the rest of the hardware. A control signal might *drive a register onto the bus*, *latch a value*, or *tell the ALU to add*. The control unit does not compute results itself; it tells every other block what to do, step by step.

Two ways to build one

There are two classic designs, and they make opposite trade-offs between speed and flexibility.

1. 1Hardwired control — the signals are produced by fixed combinational logic: decoders and gates wired straight from the opcode to the control lines. It is fast and cheap, but changing the instruction set means *redesigning the circuit*.
2. 2Microprogrammed control — the signals for each instruction are stored as a tiny program (a microprogram) in a small control memory. A microPC steps through microinstructions, each of whose bits asserts one control signal. It is slower (an extra memory read per step) but easy to extend or patch.

A microinstruction is a row of switches

Picture the microprogram as a table. Each microinstruction is one row; each column is a control signal. A 1 in a column asserts that signal during this step. The microPC points at the current row and advances down the table, issuing the signals needed to carry one instruction through fetch, decode, and execute.

Microprogram (each bit = one control signal):

  microPC | PCout MARin MemRead IRin ALUadd RFwrite
    0     |   1     1      0      0     0      0     ; PC -> MAR
    1     |   0     0      1      1     0      0     ; Mem -> IR
    2     |   0     0      0      0     1      1     ; ALU result -> reg

  microPC steps 0 -> 1 -> 2, asserting each row's signals in turn.