Two-level Partitioning of Image Processing Algorithms for the Parallel Map-oriented Machine

The partitioning of image processing algorithms with a novel hardware/software co-designframework (CoDe-X) is presented in this paper, where a new Xputer-architecture (parallel Map-oriented Machine) is used as universal accelerator based on a reconfigurable datapath hardware for speeding-up image processing applications. CoDe-X accepts C-programs and carries out both, the profilingdriven host/accelerator partitioning for performance optimization, and the resource-driven sequential/structural partitioning of the accelerator source code to optimize the utilization of its reconfigurable datapath resources.

• Type: Conference
• Year: 1996
• Where: CODES
• Authors: Reiner W. Hartenstein, Jürgen Becker, Rainer Kress
• Keywords: CODES 1996, Reconfigurable Datapath, Reconfigurable Datapath Hardware, Reconfigurable Datapath Resources, Software Engineering