Systematic Preprocessing of Data Dependent Constructs for Embedded Systems

Abstract

Data transfers and storage are dominating contributors to the area and power consumption for all modern multimedia applications. A cost-efficient realisation of these systems can be obtained by using high-level memory optimisations. This paper demonstrates that the state-of-the-art memory optimisation techniques only partly can deal with code from real-life multimedia applications. We propose a systematic preprocessing methodology that can be applied on top of the existing work. This opens more opportunities for existing memory optimisation techniques. Our methodology is complemented with a postprocessing step, which eliminates the negative effects of preprocessing and may further improve the code quality. 13 18 Our methodology has been applied on several reallife multimedia applications. Results show a decrease in the number of main memory accesses up to 45.8% compared to applying only state-of-the-art techniques.

System-scenario-based design of dynamic embedded systems

“…Although the concept of system scenarios has been applied before on top of concrete design techniques both in an ad-hoc [Chung et al 2002;Hansson et al 2007;Murali et al 2006b;Sasanka et al 2002] as well as in a systematic way 2008b;Mamagkakis et al 2007;Palkovic et al 2006;Yang et al 2002], it is possible to generalize all those scenario approaches into a common systematic methodology. This section describes such a general and still near-optimal methodology, providing generic solutions whenever available.…”

“…A huge variety of system knobs is available. Section 2 gives the example of DVFS; entirely different examples of other possible system knobs include the version of the code to run in case of an application that contains multiple versions of its source code, different compiler optimizations being applied to each of them [Palkovic et al 2006], and the configuration of processing elements (e.g., number and type of function units) in a multi-processor system [Sasanka et al 2002]. Anything that can be changed about the system during operation and that affects system cost (directly or indirectly) can be considered a system knob.…”

“…In [Palkovic et al 2005;Palkovic et al 2006], we have presented an instantiation of our scenario methodology in the form of a general toolflow that can be applied to streaming applications to reduce their energy consumption by applying the scenario concepts on top of the Data Transfer and Storage Exploration (DTSE) methodology, particularly to enable more loop transformations in the DTSE. This work targets the same type of applications as the flow presented in section 4.1.…”

“…The previous work in system scenarios is either quite ad-hoc, e.g., [Chung et al 2002;Sasanka et al 2002;Murali et al 2006b], or targets specific contexts, e.g., [Yang et al 2002;Gheorghita et al 2008b;Palkovic et al 2006;. In this paper, we propose a general systematic scenario methodology, outlining generally applicable solutions for the various steps whenever possible.…”

Storage Estimation and Design Space Exploration Methodologies for the Memory Management of Signal Processing Applications

“…The parts of the code that do not fulfill these strict conditions cannot be modeled in the geometrical model and thus cannot be transformed. To extend the exploration scope of the GLTs, different preprocessing techniques, like selective function inlining [48], pointer analysis and conversion [60,61], dynamic single assignment conversion [62], advanced copy propagation [63], hierarchical rewriting [2], and scenario creation [64][65][66], have been proposed. These techniques often require trade-offs between the freedom they allow for loop transformations and extra cost you have to pay (e.g., code size).…”