Running Benchmarks

PGSI supports both quick sanity runs and full-scale comparison suites.

List Available Targets

pgsi-analyzer benchmark list --algorithms --benchmarks-dir my-benchmarks
pgsi-analyzer benchmark list --methods --benchmarks-dir my-benchmarks

Run a Focused Comparison

Use this during setup and debugging:

pgsi-analyzer benchmark run \
  --algorithms hanoi \
  --methods cpython pypy \
  --runs 5 \
  --benchmarks-dir my-benchmarks \
  --output results_quick

Run a Full Suite

pgsi-analyzer benchmark run \
  --algorithms all \
  --methods all \
  --runs 30 \
  --benchmarks-dir my-benchmarks \
  --output results_full

Use Per-Algorithm Run Overrides

Some algorithms are slower by design. Tune per algorithm:

pgsi-analyzer benchmark run \
  --algorithms hanoi fasta binary-trees \
  --methods cpython pypy \
  --runs 10 \
  --algorithm-runs hanoi=30 fasta=8 binary-trees=15 \
  --benchmarks-dir my-benchmarks

Control Carbon/GreenScore Weights

You can adjust weighting and carbon intensity:

pgsi-analyzer benchmark run \
  --algorithms all \
  --methods all \
  --runs 20 \
  --alpha 0.4 \
  --beta 0.4 \
  --gamma 0.2 \
  --carbon-intensity 0.000475 \
  --benchmarks-dir my-benchmarks

Input and Output Expectations

Input benchmark project should include pgsi_registry.json and method folders.
Output includes: - energy_combined.csv - time_combined.csv - carbon_footprint.csv - GreenScore.csv - audit_report.json