Applying an XC6200 to Real-Time Image Processing


mentum over the past few years.1 A customcomputing machine (CCM) consists of a host processor such as a microprocessor connected to programmable hardware that implements the computationally complex part of a program. The concept arose from the fact that in microprocessor implementations, most computationally complex applications spend 90% of their execution time on only 10% of their code.2 Because hardware always outperforms software, CCM implementations should offer superior performance to microprocessor implementations. The FPGA has emerged as the natural platform for CCMs due to its reprogrammability. Recently, dedicated FPGA architectures targeted at custom computing—for example, the XC6200 FPGA series from Xilinx3—have become available. Like other FPGAs, these devices are programmable, but they also contain a microprocessor interface and can be partially reconfigured while operating. CCM technology is currently in use. For example, Virtual Computing Corporation offers a PCI-interfaced, FPGA-based customcomputing board complete with development software. At first glance, custom-computing solutions seem attractive, but the technology poses some problems. First, in the customcomputing industry, unlike the microprocessor industry, hardware development has outstripped software development. Implementing a hardware program (an algorithm) on a CCM is still largely an FPGA design task. Designers must go through numerous iterations, using VHDL-based design tools and FPGA place-and-route tools, to achieve a suitable solution. Researchers are tackling this problem in two ways. They are developing software tools that allow fast technology mapping—for example, tools developed at the ETH Computer Systems Institute in Zurich.4 Researchers are also developing FPGA libraries. Another issue we need to investigate is the effects of applying algorithmic and architectural mapping techniques used primarily in VLSI mapping for the efficient implementation of FPGA circuits. Of particular importance is the use of these techniques to produce small, fast circuit implementations that could operate concurrently in a custom-computing environment. Such mappings tend to be most successful for implementing algorithms with high degrees of parallelism. Typically, these are the algorithms implemented inefficiently on microprocessors. With these issues in mind, we designed a highly efficient XC6200 implementation of a two-dimensional discrete cosine transform (2D DCT) circuit that performs the equivaApplying an XC6200 to RealTime Image Processing XC6200

DOI: 10.1109/54.655180

Extracted Key Phrases

12 Figures and Tables

Cite this paper

@article{Woods1998ApplyingAX, title={Applying an XC6200 to Real-Time Image Processing}, author={Roger F. Woods and David W. Trainor and Jean-Paul Heron}, journal={IEEE Design & Test of Computers}, year={1998}, volume={15}, pages={30-38} }