A high level synthesis exploration framework with iterative design space partitioning

Xydis, S; Pekmestzi, K; Soudris, D; Economakos, G

dc.contributor.author	Xydis, S	en
dc.contributor.author	Pekmestzi, K	en
dc.contributor.author	Soudris, D	en
dc.contributor.author	Economakos, G	en
dc.date.accessioned	2014-03-01T02:01:39Z
dc.date.available	2014-03-01T02:01:39Z
dc.date.issued	2011	en
dc.identifier.issn	18761100	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/29220
dc.subject.other	Data paths	en
dc.subject.other	Design spaces	en
dc.subject.other	Efficient designs	en
dc.subject.other	Exploration strategies	en
dc.subject.other	Gradient based	en
dc.subject.other	High Level Synthesis	en
dc.subject.other	High quality	en
dc.subject.other	Iterative design	en
dc.subject.other	Pruning techniques	en
dc.subject.other	Quality design	en
dc.subject.other	Runtimes	en
dc.subject.other	Solution space	en
dc.subject.other	Partitions (building)	en
dc.subject.other	Space research	en
dc.subject.other	Design	en
dc.title	A high level synthesis exploration framework with iterative design space partitioning	en
heal.type	journalArticle	en
heal.identifier.primary	10.1007/978-94-007-1488-5_7	en
heal.identifier.secondary	http://dx.doi.org/10.1007/978-94-007-1488-5_7	en
heal.publicationDate	2011	en
heal.abstract	This chapter introduces a methodology for fast and efficient Design Space Exploration during High Level Synthesis. Motivated by the fact that higher quality design solutions are delivered when a larger number of parameters are explored, we study an augmented instance of the design space considering the combined impact of loop-unrolling, operation chaining and resource allocation onto the final datapath. We propose an iterative design space partitioning exploration strategy based on the synergy of an exhaustive traversal together with an introduced heuristic one. The introduced heuristic is based on a gradient-based pruning technique which efficiently evaluates large portions of the solution space in a quick manner. We show that the proposed exploration approach delivers high quality results, with considerable reductions of the exploration's runtime in respect to the fully exhaustive approach. © 2011 Springer Science+Business Media B.V.	en
heal.journalName	Lecture Notes in Electrical Engineering	en
dc.identifier.doi	10.1007/978-94-007-1488-5_7	en
dc.identifier.volume	105 LNEE	en
dc.identifier.spage	117	en
dc.identifier.epage	131	en