An hardware/software partitioning algorithm for custom computing machines

Abstract

In this paper an Hardware/Software partitioning algorithm is presented. Appropriate cost and performance estimation functions were developed, as well, as techniques for their automated calculation. The partitioning algorithm that explores the parallelism in acyclic code regions is part of a larger tool kit specific for custom computing machines. The tool kit includes a parallelising compiler, an hardware/software partitioning program, as well as, a set of programs for performance estimation and system implementation. It speeds up the computationally intensive tasks using a FPGA based processing platform to augment the functionality of the processor with new operations and parallel capacities. An example was used to demonstrate the proposed partitioning techniques.