WarpDiff-py: Differential Testing of Wasm Runtime Performance with py2wasm

this project is based on WarpDiff

Usage

Step 0: Environment and Data Preparation

• Runtime Installation: Install Wasm runtimes on your local environment for testing, including Wasmer, Wasmtime, Wasm3, WasmEdge, and WAMR.
• Programming Language: Python 3.11 (must!)
• Dataset: pybenchmarks

Step 1: Test Case Compilation

# download pybenchmarks
git clone https://github.com/dundee/pybenchmarks

# install py2wasm
pip install py2wasm
# pybenchmarks dependency
pip install jinja2 gmpy2 numpy
# if needed
sudo apt install patchelf

Adjust BENCHMARK_DIR of compile_to_target_pywasm.py based on your path of pybenchmarks.

You can also adjust TARGET_DIR to change the path you want to save the wasm target.

Then, run :

python3 compile_to_target_pywasm.py

Step 2: Runtime Performance Profiling

pip install numpy func_timeout

To obtain the total running time:

python3 runtime_profiling_total_pywasm.py

You may change PROFILINGDATA_DIR

Step 3: Differential Testing on Runtime Performance Data

pip install scikit-learn pandas

python3 analyze_performance_pywasm.py

You may change PROFILINGDATA_DIR