Comparison guide vs cachegrind/VTune/perf

[AveryClapp]

Feature Description

Comprehensive comparison document showing when to use Cache Explorer vs other tools.

Tools to Compare

1. Valgrind Cachegrind

   • Pros: Mature, accurate, widely used
   • Cons: Slow (50-100x overhead), no multi-core, text output

2. Intel VTune

   • Pros: Hardware counters, low overhead, GUI
   • Cons: Intel only, commercial, complex setup

3. Linux perf

   • Pros: Native, low overhead, kernel support
   • Cons: Limited detail, CPU-specific, no simulation

4. AMD μProf

   • Pros: Hardware counters, AMD-specific features
   • Cons: AMD only, limited platforms

Comparison Matrix

Feature	Cache Explorer	Cachegrind	VTune	perf
Source code needed	✓	✗	✗	✗
Multi-core	✓	✗	✓	✓
Simulation	✓	✓	✗	✗
Web UI	✓	✗	✓	✗
Cross-platform	✓	✓	✗	✗
Overhead	10-30x	50-100x	2-5x	<2x
Free	✓	✓	✗	✓

When to Use Cache Explorer

• ✅ Optimizing code with source access
• ✅ Comparing different algorithms
• ✅ Teaching cache behavior
• ✅ Analyzing before deployment
• ❌ Production profiling (use perf)
• ❌ Analyzing binaries (use Pin)

Files

• docs/comparison.md
• README.md (add comparison section)

Acceptance Criteria

• Fair, objective comparison
• Links to other tools
• Clear use case guidance
• Example workflows for each tool