论文信息 - A Method for Parallelizing Algorithms by Vector Scheduling Functions

A Method for Parallelizing Algorithms by Vector Scheduling Functions

A technique for parallelizing multidimensional algorithms with uniform dependencies designed for the implementation on parallel architectures of smaller dimensionalities is suggested. The technique is based on multidimensional separate scheduling of the basic operations of the algorithm. A procedure for the construction of a scheduling (parallelizing) function on the basis of theoretical results is suggested.

Nickolai A. Likhoded | S. V. Bakhanovich

[1] P. Feautrier. Some Eecient Solutions to the Aane Scheduling Problem Part Ii Multidimensional Time , 1992 .

[2] Frédéric Vivien,et al. Revisiting the decomposition of Karp, Miller and Winograd , 1995, Proceedings The International Conference on Application Specific Array Processors.

[3] W. Shang,et al. On Time Mapping of Uniform Dependence Algorithms into Lower Dimensional Processor Arrays , 1992, IEEE Trans. Parallel Distributed Syst..

[4] Paul Feautrier,et al. Some efficient solutions to the affine scheduling problem. I. One-dimensional time , 1992, International Journal of Parallel Programming.

[5] Utpal Banerjee,et al. An introduction to a formal theory of dependence analysis , 1988, The Journal of Supercomputing.

[6] Zvi M. Kedem,et al. Mapping Nested Loop Algorithms into Multidimensional Systolic Arrays , 2017, IEEE Trans. Parallel Distributed Syst..

[7] Yves Robert,et al. Mapping Uniform Loop Nests Onto Distributed Memory Architectures , 1993, Parallel Comput..

[8] Nickolai A. Likhoded. Two-Dimensional Scheduling of Algorithms with Uniform Dependencies , 1999, PaCT.