Performance

Overview

Understanding computer performance — how to measure, analyze, and optimize it — is critical for placement interviews. This section covers Amdahl’s Law (the limits of parallel speedup), the CPU performance equation, benchmarking methodologies, and hardware performance counters.

Performance Metrics

Latency vs Throughput

graph LR
    subgraph "Latency-Optimized"
        CPU["CPU<br/>Fast single task<br/>Low latency"]
    end
    subgraph "Throughput-Optimized"
        GPU["GPU<br/>Many tasks in parallel<br/>High throughput"]
    end

• Latency-sensitive: Web requests, interactive applications, databases
• Throughput-sensitive: Batch processing, ML training, rendering

Key Concepts

The Performance Equation

CPU Time = Instruction Count × CPI × Clock Period
         = Instruction Count × CPI / Clock Rate

Amdahl’s Law

Speedup = 1 / ((1 - P) + P / N)

Limits the benefit of parallelization based on the sequential fraction.

Performance Counters

Hardware counters that measure:

• Instructions executed
• Cache misses
• Branch mispredictions
• Memory bandwidth utilization

Cross-References

• Amdahl’s Law — Parallel speedup limits
• Performance Equation — CPU time formula
• Benchmarking — Measuring performance
• Performance Counters — Hardware measurement
• Cache Performance — Cache-specific metrics

Cross References

• Amdahl’s Law
• Benchmarking
• Performance Counters
• Performance Equation