A tool to run memory benchmarks on CPython commits using Memray.

pip install -e .

The tool provides a CLI command memory-tracker that can be used to run benchmarks on a range of CPython commits.

# Run benchmarks on a range of commits
memory-tracker benchmark HEAD~5..HEAD

# Specify a custom CPython repository
memory-tracker benchmark /path/to/cpython HEAD~5..HEAD

# Customize build flags
memory-tracker benchmark HEAD~5..HEAD --configure-flags="--enable-optimizations --with-lto" --make-flags="-j8"

# Specify output directory
memory-tracker benchmark HEAD~5..HEAD --output-dir="./my_benchmarks"

For each commit, the tool will:

1. Checkout the commit
2. Build CPython with the specified flags
3. Create a virtual environment with the built Python
4. Install Memray
5. Run all benchmarks in the benchmarks directory
6. Generate the following files for each benchmark:
   • .bin file with raw Memray data
   • _stats.json with memory statistics
   • _flamegraph.html with an interactive flamegraph
   • sysconfig.json with system configuration information

To add a new benchmark:

1. Create a new Python file in the benchmarks directory
2. Write your benchmark code
3. The file will be automatically discovered and run

Example benchmark:

def my_benchmark():
    # Your benchmark code here
    pass

if __name__ == '__main__':
    my_benchmark()

# Create a virtual environment
python -m venv .venv
source .venv/bin/activate

# Install development dependencies
pip install -e ".[dev]"

pytest