Assume we have a functional unit that can do one operation every 100 nanoseconds.

- For the base system,
- What is its latency?
- What is its thruput?

- If we then break it into 5 stages; each stage takes 20 nanoseconds, plus 5 nanoseconds to latch the registers between stages, and we can run it as a 5 stage pipeline.
- What is its latency?
- What is its maximum thruput?
- What is its actual thruput, if we always and only have 3 operations that we can pipeline at one time. (So 3 operations, then flush the pipe; then 3 more, then flush the pipe)?

- Suppose we have the same 5 stage pipeline, but we have a dependency between stage 4 and stage 2, so that an input for stage 2 for operation i+1 is the output of stage 4 for operation i.
- What is its latency?
- What is its maximum thruput?