Analysis of Adaptive Mapping of Parallelized Application on Multicore System

In previous publications, a self-recovering strategy, which is able to "re-map" dynamically application tasks on a multi-core system, was presented. Based on run-time failure aware techniques, this Self-Recovering strategy guarantees seamlessly termination and delivering the expected results, despite multiple node and link failures in a 2D mesh topology chip. Its efficiency has already been shown for several scenario (computation, communication time). In this article, new experiments with a more realistic NoC models (i.e. a NoC with a only port, four ports, multicast) are given. A Motion JPEG-2000 application is used as a case study to analyze the impact of real NoC models on our self-recovering strategy. The results shown, as expected, a deterioration in term of performance at very low computation-to-communication ratio (CCR), mainly due to the NoC bottleneck. Furthermore, different scenario are applied to enhancement the overall speedup and compared all together.