A lock contention prediction pipeline.
cd mutrace ./bootstrap.sh make sudo make install cd ..
- Profile the application using
mutrace
mutrace application_name
- This will produce a series of files as thread.[tid].log
- Process mutrace trace files.
python data_preprocess.py --input-dir [path/to/directory/containing/mutrace/files] --output-dir [path/to/directory/for/processed/traces]
- Generate models using the train_online.py script. Models are generated per thread, per lock, per event (subject to change).