A Reconfigurable Data Path Processor for Space Applications Gregory W. Donohoe K. Joseph Hass Institute for Advanced Microelectronics Pen-Shu Yeh Goddard Space Flight Center This paper presents the Reconfigrable Data Path Processor, a highly parallel, configurable processor being developed for high-throughput data processing aboard spacecraft. Current and planned spacecraft systems will require more complex on-board data processing. High data rate sensors, such as multi- and hyper-spectral imagers and space-based radar, require high- throughput processing on large quantities of streaming data. Space-qualified data processors lag many generations behind those available on Earth, however, and existing processors may not meet the demands of new missions. Although several research programs are underway to develop space-qualified Power PC's and Digital Signal Processors (DSPs), current designs do no translate easily into radiation-tolerant processes. Recent research in reconfigurable computing offers a new approach to high-throughput processing, using reconfigurable computer architectures that rewire themselves according to the needs of the problem at hand. Most reconfigurable computing systems are based on commercial programmable logic devices such as Field Programmable Gate Arrays (FPGAs). Current FPGAs are very powerful, with approximately one million equiv- alent gates, and they allow designs concepts to be implemented quickly. FPGAs are not optimal for most reconfigurable computing applications, however: their very fine-grained configurability exacts a heavy price in interconnects and configuration logic, and on support software. Complex FPGAs are not available in radiation-tolerant form. The Institute for Advanced Microelectronics at the University of New Mexico is teaming with the NASA Goddard Space Flight Center to develop a new reconfigurable processor for space applications. Called the Reconfigurable Data Path Processor (RDPP), it is targeted to a radiation-tolerant CMOS technology. The RDPP occupies a middle ground between fine-grained, FPGA-based reconfigurable computers, and coarse-grained general-purpose processors and DSPs. The RDPP will employ sixteen reconfigurable processing elements with a flexible interconnect network in a highly parallel architecture. Typical applications will be on-board sensor readout correction and point tracking. This paper presents the RDPP concept and design approach, including architecture, computational model, and support software, along with some preliminary results.