Questions: CPU Control Unit

Question 1

A student claims: 'The control unit must be performing the arithmetic in a CPU, since it processes the instruction and decides what happens.' What is wrong with this claim?

• A. Nothing — the control unit does process the instruction, so some arithmetic occurs there
• B. The control unit only generates control signals that configure the datapath; the ALU performs the actual arithmetic
• C. Arithmetic happens in the memory unit, since that is where operands are stored
• D. Modern CPUs have merged the control unit and ALU into a single component

Answer: (B)

The control unit is a decoder and signal generator, not a computational unit. It reads the opcode, determines what the instruction requires, and asserts the right control signals (ALU function select, register reads/writes, memory enables, MUX selections). The ALU then performs the computation using those signals. The control unit never handles data values — only control signals. This is the key architectural separation between control (what to do) and data (the values being operated on).

Question 2

A processor designer needs to build a CPU whose instruction behaviors can be updated through firmware patches after the chip is deployed in the field. Which control implementation is most appropriate?

• A. Hardwired control — faster gate propagation means updates are applied with lower latency
• B. Microprogrammed control — updating instruction behavior means changing ROM contents, not rewiring physical gates
• C. Either approach works equally well for post-deployment updates
• D. Hardwired control — combinational logic arrays can be reconfigured remotely via software

Answer: (B)

Microprogrammed control stores the control signal patterns for each instruction in a ROM (the control store). Changing instruction behavior means updating the ROM — which is precisely what firmware patches do. Intel's x86 processors use microcode updates for exactly this reason. Hardwired control implements the mapping as physical gate networks: changing instruction behavior requires redesigning and respinning the chip. Microprogrammed control trades execution speed for the flexibility to update behavior post-fabrication.

Question 3

For a LOAD instruction, the control unit asserts the memory read-enable signal while keeping the memory write-enable signal deasserted.

• T. True
• F. False

Answer: True

A LOAD instruction reads data from memory into a register. The control unit asserts memory read-enable (telling memory to output the value at the computed address), asserts the MUX select that routes the memory output to the destination register, and keeps write-enable deasserted (no data is being written to memory). Every instruction produces a unique combination of control signal values — this is precisely what the control unit's decoding logic computes from the opcode.

Question 4

Microprogrammed control executes machine instructions faster than hardwired control because it avoids the delays of complex combinational logic networks.

• T. True
• F. False

Answer: False

Microprogrammed control is slower, not faster. Each machine instruction requires a ROM lookup to find its micro-routine, followed by sequential stepping through multiple microinstructions — all adding latency. Hardwired control generates signals directly through combinational gate networks, which propagate at gate speed with no ROM access overhead. Modern high-performance CPUs use hardwired control (or hardwired fast paths with microcode fallback for complex instructions) precisely because speed is the priority.

Question 5

Explain, in your own words, the role of the control unit relative to the datapath. What would happen if the control unit generated the wrong control signals for a given instruction?

Model answer: The datapath contains the computational components (ALU, registers, memory ports, MUXes) but is passive — it needs configuration signals to do anything. The control unit reads the instruction's opcode, decodes it, and generates the exact pattern of control signals that configures the datapath for that instruction. If the wrong signals were generated, the wrong operation would execute: for example, an ADD might write to memory instead of a register, or a LOAD might select the ALU output instead of the memory output — corrupting program state silently.

The control unit is the 'interpreter' of the instruction set. Every machine instruction is a contract: opcode X means 'do this operation with these operands and write the result here.' The control unit is what makes that contract happen in hardware, by asserting the specific combination of control bits that routes data correctly through the datapath. Incorrect signals break the contract — the hardware components all operate correctly but are misconfigured, producing wrong results with no error signal.