%0 Conference Proceedings %T Polyhedral Dataflow Programming: a Case Study %+ CITI Centre of Innovation in Telecommunications and Integration of services (CITI) %+ Laboratoire de l'Informatique du Parallélisme (LIP) %+ Université Claude Bernard Lyon 1 (UCBL) %+ CASH - Compilation and Analysis, Software and Hardware (CASH) %+ Commissariat à l'énergie atomique et aux énergies alternatives (CEA) %+ CEA Tech en régions (CEA-TECH-Reg) %A Fontaine, Romain %A Gonnord, Laure %A Morel, Lionel %Z We are grateful to people at Kalray for allowing us to experiment with their ΣC compiler. Some of the experiments presented in this paper were carried out using the Grid’5000 testbed, supported by a scientific interest group hosted by Inriaand including CNRS, RENATER and several Universities as well as other organizations (see https://www.grid5000.fr).This work was also partially funded by the French National Agency of Research in theCODAS Project (ANR-17-CE23-0004-01) %< avec comité de lecture %B SBAC-PAD 2018 - 30th International Symposium on Computer Architecture and High-Performance Computing %C Lyon, France %I IEEE %P 1-9 %8 2018-09-24 %D 2018 %K load-balancing %K runtime system %K compilation %K dataflow programming %K parallelism %Z Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC]Conference papers %X Dataflow languages expose the application's potential parallelism naturally and have thus been studied and developed for the past thirty years as a solution for harnessing the increasing hardware parallelism. However, when generating code for parallel processors, current dataflow compilers only take into consideration the overall dataflow network of the application. This leaves out the potential parallelism that could be extracted from the internals of agents, typically when those include loop nests, for instance, but also potential application of intra-agent pipelining, or task splitting and rescheduling. In this work, we study the benefits of jointly using polyhedral compilation with dataflow languages. More precisely, we propose to expend the parallelization of dataflow programs by taking into account the parallelism exposed by loop nests describing the internal behavior of the program's agents. This approach is validated through the development of a prototype toolchain based on an extended version of the ΣC language. We demonstrate the benefit of this approach and the potentiality of further improvements on relevant case studies. %G English %2 https://cea.hal.science/cea-01855997v2/document %2 https://cea.hal.science/cea-01855997v2/file/finalversion_hal_sbacpad18_FGM.pdf %L cea-01855997 %U https://cea.hal.science/cea-01855997 %~ ENS-LYON %~ CEA %~ CNRS %~ INRIA %~ UNIV-LYON1 %~ INSA-LYON %~ INRIA-RHA %~ INRIA_TEST %~ GRID5000 %~ TESTALAIN1 %~ DRT %~ INRIA2 %~ INRIA-RENGRE %~ CITI %~ INSA-GROUPE %~ UDL %~ UNIV-LYON %~ SILECS %~ ANR