In this Chapter we report various results to support our claim that function / architecture optimization and co-design is indeed an improvement over other current day co-design approaches. We compare the improved and original co-design process using the Polis co-design tool; the results that compare the Polis methodology to other approaches reported by Balarin et. al. in  (such as the synchronous composition done in Esterel ) carries by transitivity since we use Polis here as the baseline1. We will, unless otherwise stated, present results for the (A)FFG Tree Form which, as we described in earlier Chapters, provides for the best quality output. The reader should also assume that all the aforementioned optimization techniques have been applied except Relaxed Operation Motion (ROM), unless otherwise stated. We will dedicate later in the Chapter several result tables and analyses for the latter, and also compare the Tree vs. the Shared DAG form in terms of output quality and running times, but report in the main result tables what the toolset default2 co-design flow obtains. The results have been collected by exercising several representative applications with typical input, and validating by trace comparison3 with the golden trace; both obtained from the application designer.
Unable to display preview. Download preview PDF.