Benchmark suite with known workloads

[AveryClapp]

Feature Description

Reference benchmark suite for validating cache simulator accuracy and performance.

Benchmarks

1. SPEC CPU 2017 (subset)
   • 500.perlbench_r
   • 502.gcc_r
   • 520.omnetpp_r
2. PARSEC 3.0
   • blackscholes
   • canneal
   • streamcluster
3. Custom microbenchmarks
   • Sequential access (100% L1 hit)
   • Random access (high L3 miss)
   • False sharing (high coherence traffic)
   • Strided access (prefetcher stress test)

Goals

• Validation: Compare against hardware performance counters
• Performance: Track analysis speed over time
• Regression testing: Ensure accuracy doesn't degrade

Implementation

• Docker container with all benchmarks
• Automated nightly runs
• Store historical results
• Alert on regressions (>5% difference)

Files

• benchmarks/ (new directory)
• benchmarks/run-all.sh
• benchmarks/validate.py (compare with perf)

Expected Results

Benchmark	Our Hit Rate	perf stat	Delta
sequential	99.8%	99.9%	-0.1%
random	45.2%	44.8%	+0.4%