Adaptivity Support for MPSoCs Based on Process Migration in Polyhedral Process Networks

摘要

System adaptivity is becoming an important featureof modern embedded multiprocessor systems. To achieve thegoal of system adaptivity when executing Polyhedral ProcessNetworks (PPNs) on a generic tiled Network-on-Chip (NoC)MPSoC platform, we propose an approach to enable the run-timemigration of processes among the available platform resources. Inour approach, process migration is allowed by a middleware layerwhich comprises two main components. The first componentconcerns the inter-tile data communication between processes. We develop and evaluate a number of different communicationapproaches which implement the semantics of the PPN modelof computation on a generic NoC platform. The presentedcommunication approaches do not depend on the mappingof processes and have been implemented on a Network-on-Chip multiprocessor platform prototyped on an FPGA. Theircomparison in terms of the introduced overhead is presented intwo case studies with different communication characteristics. The second middleware component allows the actual run-timemigration of PPN processes. To this end, we propose and evaluatea process migration mechanism which leverages the PPN modelof computation to guarantee a predictable and efficient migrationprocedure. The efficiency and applicability of the proposedmigration mechanism is shown in a real-life case study.