OpenSource Code for Multi-graph Learning for Parallelism Discovery in Sequential Programs
This project is the implementation of our work on the parallelism discovery with multi-graph learning. Our framework leverages ASTNN-based graph neural network model and DGCNN-based graph neural network model. About the original implementation of ASTNN and DGCNN, you can find the code here:
- ASTNN: (https://github.com/zhangj111/astnn)
- DGCNN: ( https://github.com/muhanzhang/DGCNN)
- DGCNN-tensorflow: (https://github.com/hitlic/DGCNN-tensorflow)
Python3
pytorch
pycparser
Tensorflow
Pluto==0.11.4(http://pluto-compiler.sourceforge.net/)
Rose (http://rosecompiler.org/)
Clang/LLVM
We provide a dataset generator. This step assumes that you want generating your own dataset. You
may add your source code manually to utils/data/source_code
, then run:
python3 data_gen.py
Note that you need replace the polycc
and autoPar
with your own.
Train and test:
python3 model_train.py
@article{SHEN2021515,
author = {Yuanyuan Shen and Manman Peng and Shiling Wang and Qiang Wu},
title = {Towards parallelism detection of sequential programs with graph neural network},
journal = {Future Generation Computer Systems},
volume = {125},
pages = {515-525},
year = {2021},
issn = {0167-739X},
doi = {https://doi.org/10.1016/j.future.2021.07.001},
url = {https://www.sciencedirect.com/science/article/pii/S0167739X21002557}
}
@article{SHEN,
author = {Yuanyuan Shen and Manman Peng and Qiang Wu and Guoqi Xie},
title = {Multigraph Learning for Parallelism Discovery in Sequential Programs}
},
journal = {Concurrency and Computation: Practice and Experience},
year = {2023},
doi = {https://doi.org/10.1002/cpe.7648},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/cpe.7648}