ThreadContention

A lock contention prediction pipeline.

Build mutrace

cd mutrace ./bootstrap.sh make sudo make install cd ..

Run pipeline

1. Profile the application using mutrace

mutrace application_name

2. This will produce a series of files as thread.[tid].log
3. Process mutrace trace files.

python data_preprocess.py --input-dir [path/to/directory/containing/mutrace/files] --output-dir [path/to/directory/for/processed/traces]

4. Generate models using the train_online.py script. Models are generated per thread, per lock, per event (subject to change).