This project is designed to benchmark and analyze the performance of various methods implemented in a Scala library for array transformations. It focuses on comparing different low-level implementations of map, foreach, and other traversal operations, aiming to detect performance differences depending on call site structure and compiler behavior.
Full perfomance review you can find here.
We test performance of various methods applied to Array[Int]:
Benchmarks are executed using Makefile targets. Each target runs a specific group of benchmarks via JMH. For example:
make bench_map_1
This command will:
- run JMH benchmarks with varying array sizes,
- store the results as CSV in
analysis/data/, - adjust measurement time depending on the input size.
Analysis is performed using Python scripts inside the analysis/ folder, relying on pandas, statsmodels, and matplotlib.
To run an analysis:
cd analysis
source venv/bin/activate
python ploty.py map_1
This script reads a benchmark result file (e.g., bench_map_1.csv), fits linear regression (latency vs size), and generates plots in plots/.
The repository includes a GitHub Actions workflow for automated benchmarking:
- Workflow config:
.github/workflows/bench.yml - Benchmarks are run manually and their results are stored as workflow artifacts
You can trigger benchmarks manually or on every push via the Actions tab in the GitHub UI.
.
├── analysis/ # Python scripts and benchmark output
├── review/ # Written reports and result summaries
├── project/ # sbt project files
├── src/ # Scala JMH benchmarks
├── .github/ # GitHub Actions workflow
├── Makefile # Benchmark automation entrypoints
└── README.md
- Scala 3 and sbt
- JMH (via sbt plugin)
- Python 3
To set up the Python environment:
cd analysis
python -m venv venv
source venv/bin/activate
make update
make bench_map_1
cd analysis
python ploty.py map_1
This will:
- run the benchmark for various array sizes,
- store the results in
analysis/data/bench_map_1.csv, - and produce a performance plot in
analysis/plots/.