SBIR-STTR Award

Symbolic computation is becoming more important in experimental and theoretical physics for symbolic algebraic manipulation, for signal and image processing, and for expert systems. The utility industry is also investigating the use of symbolic computation and expert systems for reactor modeling, control, and robotics applications. With the increasing demands placed on these systems, the computational requirements especially for combined symbolic/numeric processing outstrip the capabilities of current general-purpose microprocessors. In addition, the languages and operating systems utilized by Lisp and Prolog (the two languages of choice for symbolic computation) are not appropriate for real-time control. However, a new type of single-chip microprocessor, the digital signal processor (DSP), can offer an order of magnitude better performance than can be obtained with a general purpose single-chip processor at the same price. This project will determine the feasibility of implementing a high-performance real-time common Lisp software compiler and runtime system for one of these DSP chips. The project includes the results of a design study and prototype implementation. Expected performance is three to five times the speed of a workstation for symbolic computation, and ten times that for numeric computation, while offering a cost in the several thousand dollar range and embedded real-time capabilities.Anticipated Results/Potential Commercial Applications as described by the awardee: The benefit of a real-time symbolic/numeric processing system would be the ability to incorporate high-performance numeric and symbol processing capability into a single processor. This integration can improve system performance and reduce system cost by replacing several specialized and expensive processors with a single processor and less communications overhead. Such a system would be of significant benefit whenever interpretation, planning, or intelligent behavior is required along with numeric processing, such as in smart sensors, smart robotic controls, and automatic target recognition.Topic 16: High Energy Physics Data Processing and Detector Instrumentation

Symbolic computation is becoming more important in experimental and theoretical physics for symbolic algebraic manipulation, for signal and image processing, and for expert systems. The utility industry is also investigating the use of symbolic computation and expert systems for reactor modeling, control, and robotics applications. With the increasing demands placed on these systems, the computational requirements especially for combined symbolic/numeric processing outstrip the capabilities of current general-purpose microprocessors. In addition, the languages and operating systems utilized by Lisp and Prolog (the two languages of choice for symbolic computation) are not appropriate for real-time control. However, a new type of single-chip microprocessor, the digital signal processor (DSP), can offer an order of magnitude better performance than can be obtained with a general purpose single-chip processor at the same price. This project will determine the feasibility of implementing a high-performance real-time common Lisp software compiler and runtime system for one of these DSP chips. The project includes the results of a design study and prototype implementation. Expected performance is three to five times the speed of a workstation for symbolic computation, and ten times that for numeric computation, while offering a cost in the several thousand dollar range and embedded real-time capabilities.Anticipated Results/Potential Commercial Applications as described by the awardee: The benefit of a real-time symbolic/numeric processing system would be the ability to incorporate high-performance numeric and symbol processing capability into a single processor. This integration can improve system performance and reduce system cost by replacing several specialized and expensive processors with a single processor and less communications overhead. Such a system would be of significant benefit whenever interpretation, planning, or intelligent behavior is required along with numeric processing, such as in smart sensors, smart robotic controls, and automatic target recognition.Topic 16: High Energy Physics Data Processing and Detector Instrumentation