Kernel scheduling techniques for efficient solution space exploration in reconfigurable computing

摘要

Run-time reconfigurable systems have emerged as a flexible way of facing with a single device a wide range of applications. The adequate exploitation of the available hardware resources requires the development of design tools that support the design process. Task scheduling becomes a very critical issue in achieving the high performance that DSP and multimedia applications demand. This paper addresses task scheduling in reconfigurable computing for these kinds of applications. First, we discuss the main issues involved in the problem. Then, we propose a new methodology with some bounding and pruning techniques in order to produce an efficient exploration of the design space. This work has been developed for Morphosys, a coarse-grain multi-context reconfigurable architecture, and it can be successfully applied to systems with similar features.