Block Scheduling of Iterative Algorithms and Graph-Level Priority Scheduling in a Simulated Data-Flow Multiprocessor
Issue No. 04 - April (1993 vol. 4)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/71.219755
<p>Iterative methods for solving linear systems are discussed. Although these methods areinherently highly sequential, it is shown that much parallelism could be exploited in adata-flow system by scheduling the iterative part of the algorithms in blocks and bylooking ahead across several iterations. This approach is general and will apply to otheriterative and loop-based problems. It is also demonstrated by simulation that relyingsolely on data-driven scheduling of parallel and unrolled loops results in low resourceutilization and poor performance. A graph-level priority scheduling mechanism has beendeveloped that greatly improves resource utilization and yields higher performance.</p>
Index Termsblock scheduling; iterative algorithms; graph-level; priority scheduling; data-flowmultiprocessor; linear systems; parallelism; scheduling; iterative methods; parallelalgorithms; scheduling
P. Evripidou and J. Gaudiot, "Block Scheduling of Iterative Algorithms and Graph-Level Priority Scheduling in a Simulated Data-Flow Multiprocessor," in IEEE Transactions on Parallel & Distributed Systems, vol. 4, no. , pp. 398-413, 1993.