MPSoC （multi-processor systems on-chip） are going to be the leading hardware platform featured in embedded systems, The complexity of multimedia applications implemented on MPSoC and the requirements of users compli...MPSoC （multi-processor systems on-chip） are going to be the leading hardware platform featured in embedded systems, The complexity of multimedia applications implemented on MPSoC and the requirements of users complicate the designer＇s job to conceive and deliver efficient systems in the shortest time. MPSoC dedicated to specific embedded applications are facing constraints of performance and cost. In this context, in order to improve the performance of these systems, we propose a methodology for dynamic task migration in dedicated MPSoC. This methodology aims to improve the performance of execution and also reduction of energy consumption. We propose in this paper, an estimation of the real-time energy （line）. This approach is structured on four steps： （1） modeling software; （2） mapping and execution; （3） performance evaluation; （4） migration. We have using the Soclib tool to determine performance estimation of migrating software task to hardware component. Experiments on MJPEG decoder are made to illustrate the efficiency of our approach in performance estimation.展开更多
文摘MPSoC （multi-processor systems on-chip） are going to be the leading hardware platform featured in embedded systems, The complexity of multimedia applications implemented on MPSoC and the requirements of users complicate the designer＇s job to conceive and deliver efficient systems in the shortest time. MPSoC dedicated to specific embedded applications are facing constraints of performance and cost. In this context, in order to improve the performance of these systems, we propose a methodology for dynamic task migration in dedicated MPSoC. This methodology aims to improve the performance of execution and also reduction of energy consumption. We propose in this paper, an estimation of the real-time energy （line）. This approach is structured on four steps： （1） modeling software; （2） mapping and execution; （3） performance evaluation; （4） migration. We have using the Soclib tool to determine performance estimation of migrating software task to hardware component. Experiments on MJPEG decoder are made to illustrate the efficiency of our approach in performance estimation.
