What's SkeTo Project?
"SkeTo Project" (Skeletons in Tokyo) is a research project working on skeletal parallelism (or algorithmic skeletons).
We are studying parallel skeletons based on the constructive algorithmics. The main contributions of our project are as follows:
- Parallel Skeletons for Several Data Structures: The constructive algorithms provides a general handling of several data structures. Based on this, we are studying parallel skeletons for lists, trees, and matrixes.
- Optimization Mechanism for Skeletal Programs: The programs with many skeletons may lead undesirable overheads. We are studying an optimization mechanism of the skeletal programs, based on the program transformation techniques.
This project was partially supported by the PRESTO program by Japan Science and Technology Agency (JST).
Follow the link for more details (Sorry, under-construction).
Results
We have proposed several fundamental parallel skeletons for data structures lists, trees, matrices based on the BMF (Bird-Meertens Formalism), which provides a concise interface of skeletons. We have also shown efficient implementations of the skeletons.
In terms of the optimization mechanism, we have developed a fusion transformation of the parallel skeletons on lists. This optimization mechanism is powerful and easy to implement.
We implemented our ideas as a parallel skeleton library. The current version of our parallel skeleton library includes several parallel skeletons for lists, trees, and matrices. The optimization mechanism is now under implementation, and will be included our library soon. Our parallel skeleton library is available from the Download page.
Researchers
- Yoshiki AKASHI*1
- Kento EMOTO*2
- Zhenjiang HU*2
- Hideya IWASAKI*1
- Kazuhiko KAKEHI*2
- Kiminori MATSUZAKI*2
- Masato TAKEICHI*2
*1
Information Processing Laboratory,
Department of Computer Science,
The University of Electro-Communications
*2
Information Processing Laboratory,
Department of Mathematical Informatics,
Graduate School of Information Science and Technology,
The University of Tokyo
Publications
- Kento Emoto, Zhenjiang Hu, Kazuhiko Kakehi, and Masato Takeichi:
A Compositional Framework for Developing Parallel Programs on Two Dimensional Arrays.
Technical Report METR2005-09, Department of Mathematical Informatics, University of Tokyo, 2005. - Kiminori Matsuzaki, Zhenjiang Hu, Kazuhiko Kakehi, and Masato Takeichi:
Systematic Derivation of Tree Contraction Algorithms.
To appear in Parallel Processing Letters, 2004. - Kiminori Matsuzaki, Kazuhiko Kakehi, Hideya Iwasaki, Zhenjiang Hu, and Yoshiki Akashi:
A Fusion-Embedded Skeleton Library.
In Proceedings of Annual European Conference on Parallel Processing (Euro-Par 2004), 31th August - 3rd September, 2004, Lecture Notes in Computer Science 3149, pp. 644-653, Springer Verlag. - Kiminori Matsuzaki, Kazuhiko Kakehi, Zhenjiang Hu, and Masato Takeichi:
Systematic Derivation of Tree Contraction Algorithms.
In Proceedings of the 4th International Workshop on "Constructive Methods for Parallel Programming" (CMPP 2004), 14th July, 2004, Technical Report of Westfalische wilhelms-univeersitat munster, pp. 109-123. - Hideya Iwasaki, and Zhenjiang Hu:
A New Parallel Skeleton for General Accumulative Computations.
International Journal of Parallel Programming, 32 (5): 389-414, October 2004. - Kazuhiko Kakehi, Zhenjiang Hu, and Masato Takeichi:
List Homomorphism with Accumulation.
4th International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD'03), Lubeck, Germany, October 16-18, 2003, pp. 250-259. - Kiminori Matsuzaki, Zhenjiang Hu, and Masato Takeichi:
Parallelization with Tree Skeletons.
International Conference on Parallel and Distributed Computing (Euro-Par 2003), Klagenfurt, Austria, 26th - 29th August 2003, Lecture Notes in Computer Science 2790, Springer Verlag, pp. 789-798. An extended version appears as Technical Report METR 2003-21, Department of Mathematical Informatics, University of Tokyo, 2003.