论文信息 - BSP, LogP, and Oblivious Programs

BSP, LogP, and Oblivious Programs

We compare the BSP and the LogP model from a practical point of view. Using compilation instead of interpretation improves the (best known) simulations of BSP programs on LogP machines by a factor of O(log P) for oblivious programs. We show that the runtime decreases for classes of oblivious BSP programs if they are compiled into LogP programs instead of executed directly using a BSP runtime library. Measurements support the statements above.

Welf Löwe | Wolf Zimmermann | Jörn Eisenbiegler

[1] Richard Cole,et al. On Edge Coloring Bipartite Graphs , 1980, SIAM J. Comput..

[2] Ramesh Subramonian,et al. LogP: towards a realistic model of parallel computation , 1993, PPOPP '93.

[3] Welf Löwe,et al. On the optimization by redundancy using an extended LogP model , 1997, Proceedings. Advances in Parallel and Distributed Computing.

[4] Chris J. Scheiman,et al. LogGP: incorporating long messages into the LogP model—one step closer towards a realistic model for parallel computation , 1995, SPAA '95.

[5] William F. McColl,et al. Scalable Computing , 1995, Computer Science Today.

[6] Leslie G. Valiant,et al. A bridging model for parallel computation , 1990, CACM.

[7] Paul G. Spirakis,et al. BSP vs LogP , 1996, SPAA '96.

[8] Jonathan M. D. Hill,et al. Theory, Practice, and a Tool for BSP Performance Prediction , 1996, Euro-Par, Vol. II.

[9] Richard M. Karp,et al. Optimal broadcast and summation in the LogP model , 1993, SPAA '93.