Abstract
This paper targets the autonomic management of dynamically partially reconfigurable hardware architectures based on FPGAs. Such hardware-level autonomic computing has been less often studied than at software-level. We consider control techniques to model the considered behaviours of the computing system and derive a controller for the control objective enforcement. Discrete Control modelled with Labelled Transition Systems is employed in this paper. Such models are amenable to Discrete Controller Synthesis algorithms that can automatically generate a controller enforcing the correct behaviours of a controlled system. A general modelling framework is proposed for the control of FPGA based computing systems. We consider system application described as task graphs and FPGA as a set of reconfigurable areas that can be dynamically partially reconfigured to execute tasks. We encode the computation of an autonomic manager as a DCS problem w.r.t. multiple constraints and objectives e.g., mutual exclusion of resource uses, power cost minimization. We validate our models and manager computations by using the BZR language and an experimental demonstrator implemented on a Xilinx FPGA platform.