Theoretical and Experimental Study of a Parallel Algorithm Solving the Matrix Chain Product Problem
We are interested in the matrix chain product problem that may be solved by a well known dynamic programming algorithm (DPA) a so-called polyhedral algorithm i.e. structured in a (three) DO-loop nest (3-DLN). We propose a parallelization approach by choosing a fine grain task segmentation of the 3-D...
Saved in:
Published in | Proceedings of the International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA) pp. 341 - 347 |
---|---|
Main Authors | , , |
Format | Conference Proceeding |
Language | English |
Published |
Athens
The Steering Committee of The World Congress in Computer Science, Computer Engineering and Applied Computing (WorldComp)
01.01.2017
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | We are interested in the matrix chain product problem that may be solved by a well known dynamic programming algorithm (DPA) a so-called polyhedral algorithm i.e. structured in a (three) DO-loop nest (3-DLN). We propose a parallelization approach by choosing a fine grain task segmentation of the 3-DLN. We then construct a level decomposition of the task graph (LDTG) and design, for any given number of processors, two particular schedulings based on scanning the levels of the LDTG . We proceed in the first (resp. second) without (resp. with) inter-level overlapping. We concentrate first on a theoretical study that is validated through a series of simulations. An experimental study achieved on a parallel multicore machine permits to validate the contribution. |
---|