JOHNNIE W. BAKER
CURRENT RESEARCH AREAS OF INTEREST
Parallel architecture and software for the air traffic control problem
Parallel models, data parallel and associative SIMD computing, parallel algorithms
SIMD algorithms and software for sequence alignment in bioinformatics
Molecular similarity analysis, drug design, molecular engineering, structure-activity visualization
A primary focus of Baker's research is the investigation of an efficient SIMD solution for air traffic control (ATC). Unlike current and past multiprocessor solutions, this approach avoids the use of dynamic task scheduling and load balancing, distributed data bases, and other aspects of past implementations that have caused these ATC systems to be extremely complex, highly unpredictable, and unable to process all tasks prior to their deadline. The deterministic aspect of the SIMD hardware allows a much simpler and extremely efficient solution to be created which uses the accurate prediction of the running time for tasks to statically schedule all tasks and to guarantee that all deadlines will be met. An ATC prototype consisting of 8 key ATC tasks has been implemented on the CSX600 ClearSpeed SIMD accelerator. This prototype is being implemented on a multiprocessor to establish the feasibility of the two systems and to compare the performance and predictability of the two systems. This research is especially applicable to all computational intensive real-time problems with hard deadlines. Future research may involve work to see if this ATC prototype solution can be effectively supported on an NVIDIA multicore platform.
A second major focus of Baker's current research has been the development of a Multiple SIMD a computational and architectural parallel model called MASC (for Multiple ASsociative Computing) that supports multiple dynamically configurable SIMD threads that are controlled using task parallelism. An “Associative SIMD” has a few additional basic properties not always supported by a SIMD, but which simplify SIMD programming and can easily be supported in hardware. Previous work investigated the power of MASC by comparing it to other models of computation such as PRAM and mesh with multiple broadcast (MMB). Previous work also includes building a cycle-precision simulator on a sequential computer for the MASC model. This implementation involved building a run-time system for MASC that uses the manager-worker control paradigm to control the multiple SIMD threads and extending a previous language and compiler for an associative SIMD computer to a language and compiler for MASC. Several efficient algorithms for MASC have been designed, implemented, and evaluated using this MASC simulator, including QuickHull and the Floyd Warshall all-pairs shortest path algorithm. Future research will include implementing the earlier cycle-precision simulator on a parallel computer that can handle much larger simulations, creating MASC algorithms designed for large scale computing, and using these algorithms and the parallel simulator to evaluate this model’s ability to support petascale and exascale computing.
A third focus of Baker's current research is in the computational science area. In computational chemistry, he is working with Professor Chun-che Tsai in the Department of Chemistry at Kent State and graduate students to develop sequential and parallel algorithms to measure the similarity between different molecules and using this information to predict properties of potential compounds and to engineer drugs and compounds with certain desirable properties and structure. This is joint work with Professor Chun-che Tsai in the Department of Chemistry at Kent State. Additionally, he is involved in research in bioinformatics involving creating a Smith-Waterman sequence alignment algorithm for an associative SIMD computer that takes advantage of the massively parallelism and simpler programming that is naturally supported on these types of parallel computers.
EARLIER RESEARCH AREAS OF INTEREST
Banach spaces and topology