Benchmarks¶

Performance tracking for physities operations.

Understanding the Results¶

The benchmarks compare physities (with unit safety) against plain Python (raw floats).

Values are in iterations/second - higher is faster.

What’s Being Measured¶

Benchmark	Plain Python	Physities
Add	`100.0 + 200.0`	`Meter(100) + Meter(200)`
Multiply	`10.0 * 5.0`	`Meter(10) * Second(5)`
Divide	`100.0 / 10.0`	`Meter(100) / Second(10)`
Convert	`value / 1000`	`m.convert(Kilometer)`

Expected Overhead¶

Physities adds overhead for unit safety:

Simple operations (add, multiply): ~10-50x slower than raw floats
Conversions: ~10x slower than manual math
Unit creation: ~200 nanoseconds per instance

This overhead is the cost of:

Dimensional analysis (preventing Meter + Second)
Automatic conversion factors
Type-safe operations

When It Matters¶

For most applications, this overhead is negligible:

Scientific calculations: Usually I/O or algorithm bound, not arithmetic
Data pipelines: The safety prevents costly bugs
Simulations: Consider using NumPy arrays with physities for batch operations

The overhead becomes noticeable only in tight loops doing millions of operations per second.

Running Benchmarks Locally¶

# Install benchmark dependencies
pip install pytest-benchmark

# Run core benchmarks
pytest benchmarks/bench_core.py --benchmark-only

# Run all detailed benchmarks
pytest benchmarks/ --benchmark-only

# Compare against previous run
pytest benchmarks/ --benchmark-only --benchmark-compare