CISE Instrumentation: Acquisition of a Multidisciplinary Research in Computing

9729809 Manoj Clemson大学CISE研究仪器：获得用于计算多学科研究的多处理器该研究仪器奖旨在获得一个多处理器系统，该系统将支持以下研究项目：-指令级并行处理器，-共享存储计算机基准的自动生成，-结构声学中的有限元计算，-磁轴承非线性控制器的性能验证，以及-流固耦合系统的非线性振动。每个项目都涉及大型模拟器的开发和计算密集型模拟的使用，这些模拟需要强大的专用处理器、大RAM容量和大I/O容量。对ILP处理的研究包括开发分散的执行模型，并通过详细的并行模拟器来研究它们。在自动基准生成方面的研究包括开发从简明的算法规范自动生成多处理器基准程序的技术，这有助于在单个系统的基础上识别性能最好的算法。有限元计算的研究是通过开发和实现高阶时空有限元方法以及在多处理机上模拟它们的适当数据结构来预测来自大型结构物的波传播。非线性控制器的研究包括建立改进的主动磁轴承非线性动态模型，并在多处理机上进行动态仿真验证。非线性振动的研究主要集中在利用多处理机建立模型和计算密集型技术来解决纺织和高性能水下缆索结构应用中的复杂、非线性、流体-结构系统的平衡和线性化振动问题。

9729809 Franklin, Manoj Clemson University CISE Research Instrumentation: Acquisition of a Multiprocessor for Multidisciplinary Research in Computing This research instrumentation award serves to acquire a multiprocessor system which will enable the following research projects:- Instruction-Level Parallel Processors,- Automatic Generation of Benchmarks for Shared Memory Computers, - Finite Element Computations in Structural Acoustics,- Performance Validation of Nonlinear Controllers for Magnetic Bearings, and - Nonlinear Vibrations of Coupled Fluid-Structure Systems. Each of these projects involves the development of large simulators and the use of compute-intensive simulations that require powerful dedicated processors, large RAM capacity, and large I/O capacity. Research in ILP processing involves developing decentralized execution models and studying them by detailed, parallel simulators. Research in automatic benchmark generation involves developing techniques to automatically generate multiprocessor benchmark programs from concise algorithm specifications, which aids in identifying the top-performing algorithm on an individual system basis. Research on finite element computations deals with predicting wave propagation from large structures by developing and implementing high-order space-time finite element methods and appropriate data structures for simulating them on multiprocessors. Research on nonlinear controllers involves the development of improved nonlinear dynamic models for active magnetic bearing, and their validation via dynamic simulation on multiprocessors. Research on nonlinear vibrations focuses on using multiprocessors to develop models and computationally intensive techniques to solve the equilibrium and linearized vibration of complex, nonlinear, fluid-structure systems used in textiles and high performance underwater cable structures applications.