2.8Beyond
The idea
Superscalar processors go beyond one instruction per cycle by starting several independent instructions together.
Think of it like this: Opening two checkout counters instead of one.
Why it exists
Once a pipeline reaches CPI ≈ 1 the only way to go faster is to issue more than one instruction per cycle.
How it works
- Hardware looks at the next few instructions and finds independent ones.
- It issues them together to different units.
- Dependencies and hazards limit how many can go at once.
- VLIW lets the compiler pick the bundle instead of the hardware.
Worked example: 4 independent instructions
Scalar pipeline: 8 cycles. 2-issue superscalar: 6 cycles. In steady state it completes 2 instructions per cycle (CPI → 0.5).
Now play with it
- →Compare the two tabs: how many instructions start in cycle 1?
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | |
|---|---|---|---|---|---|---|---|---|
| I1 | IF | ID | EX | MEM | WB | |||
| I2 | IF | ID | EX | MEM | WB | |||
| I3 | IF | ID | EX | MEM | WB | |||
| I4 | IF | ID | EX | MEM | WB |
4 independent instructions: 8 cycles.
Check yourself
What is the best-case CPI of a 2-issue superscalar processor?
Exam notes (the dense version)
Superscalar (multiple issue/cycle), VLIW (compiler-scheduled), dynamic scheduling (scoreboard, Tomasulo), mostly conceptual.
Previous year questions: Instruction Pipelining & Hazards (15)
PYQs are grouped by topic, so this is the full set for Instruction Pipelining & Hazards; try the ones that match this subtopic. Open all 15 on the topic page.