This directory stores benchmark results for tracking performance across versions.

All benchmark scripts are in the examples/ directory. Compile first with rebar3 compile.

escript examples/benchmark.erl --quick

escript examples/benchmark.erl --full

escript examples/benchmark_compare.erl

This outputs a summary table useful for comparing performance between versions.

escript examples/benchmark.erl --concurrent

escript examples/bench_channel.erl

Compares Channel API throughput at various message sizes.

escript examples/bench_reactor_modes.erl

Compares Reactor performance in worker mode vs subinterpreter mode (Python 3.12+).

rebar3 shell
> run_benchmark:run().

Compares Erlang event loop vs standard asyncio event loop performance.

To save benchmark results for comparison:

# Run and save to timestamped file
escript examples/benchmark_compare.erl > benchmark_results/$(date +%Y%m%d_%H%M%S).txt

Result files are named with timestamps: YYYYMMDD_HHMMSS.txt

Example files:

• reactor_modes_20260309_093558.txt - Reactor modes comparison results

When comparing versions, focus on:

1. Sync call latency - Time per py:call() operation
2. Cast throughput - Non-blocking py:cast() operations per second
3. Concurrent throughput - Performance under parallel load
4. Channel throughput - Messages per second at various sizes
5. Reactor modes - Worker vs subinterpreter performance

Benchmarks automatically print system info including:

• Erlang/OTP version
• Number of schedulers
• Python version
• Execution mode (free_threaded, subinterp, multi_executor)
• Max concurrent operations