Pipeline Performance Metrics
You'll learn how to measure how well a pipelined processor performs using key metrics like CPI, IPC, and execution time.
Understanding Pipeline Performance
When we evaluate a pipelined processor, we need concrete numbers to judge its effectiveness. Three fundamental metrics help us do this:
CPI: Cycles Per Instruction
CPI tells you how many clock cycles it takes, on average, to complete one instruction. In an ideal pipeline with no delays, CPI = 1 (one instruction completes every cycle after the pipeline fills). In reality, hazards and stalls increase this number.
Formula: CPI = Total Cycles / Total Instructions
IPC: Instructions Per Cycle
IPC is simply the inverse of CPI. It measures throughput: how many instructions complete per clock cycle on average. Higher is better!
Formula: IPC = 1 / CPI = Total Instructions / Total Cycles
Execution Time
This is the actual wall-clock time to run a program.
Formula: Execution Time = Total Cycles × Clock Cycle Time
Or equivalently: Execution Time = Instructions × CPI × Clock Cycle Time
Real-World Conditions
In an ideal 5-stage pipeline running 1000 instructions:
- Initial fill takes 4 cycles
- Next 1000 cycles each complete one instruction
- CPI = 1004/1000 ≈ 1.004 (nearly perfect!)
But with stalls from hazards, dependencies, or multi-cycle instructions:
- If we have 200 stall cycles: CPI = 1204/1000 = 1.204
- IPC drops to 1/1.204 ≈ 0.831
The execution time depends on both CPI and the clock speed—a deeper pipeline might reduce cycle time but increase CPI due to more hazards.
Key Takeaway: CPI, IPC, and execution time quantify pipeline efficiency; ideal pipelines achieve CPI ≈ 1, but real-world hazards and stalls degrade performance.