Cache-oblivious MPI all-to-all communications based on Morton order.
The work of COMPI is pulished in TPDS'18 and PPoPP'17. See the paper for details. To cite our work:
@ARTICLE{8091010,
author={Li, Shigang and Zhang, Yunquan and Hoefler, Torsten},
journal={IEEE Transactions on Parallel and Distributed Systems},
title={Cache-Oblivious MPI All-to-All Communications Based on Morton Order},
year={2018},
volume={29},
number={3},
pages={542-555},
doi={10.1109/TPDS.2017.2768413}
}
@article{li2017poster,
title={POSTER: cache-oblivious MPI all-to-all communications on many-core architectures},
author={Li, Shigang and Zhang, Yunquan and Hoefler, Torsten},
journal={ACM SIGPLAN Notices},
volume={52},
number={8},
pages={445--446},
year={2017},
publisher={ACM New York, NY, USA}
}
See LICENSE.