PetaFlops Acoustic Simulation

PETAFLOPS声学模拟PI：Dinesh Manocha计算机科学系北卡罗来纳大学这个奖项是根据2009年美国复苏和再投资法案(公法111-5)资助的。该项目建议为声学模拟开发一个可伸缩的、千万亿次浮点计算基础设施。主要的重点将是开发大规模并行算法，这些算法可以利用图形处理单元(GPU)等多核加速器的计算能力来实现所需的性能。主要研究内容包括：(1)开发求解声波方程的高精度和低频散数值方法；(2)高效数值和几何声学传播的大规模并行算法；(3)在GPU集群上运行数百万线程的软件库，以实现基于声场分解的Petaflop声学性能；以及(4)复杂CAD模型的声学分析。这一努力的更广泛影响是跨学科的，跨越学术界和行业，包括用于多核加速器的新应用软件库。

PetaFlops Acoustic SimulationPI: Dinesh ManochaDepartment of Computer ScienceUniversity of North CarolinaThis award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)." This project proposes to develop a scalable, petaflop computational infrastructure for acoustic simulation. The main focus will be on developing massively parallel algorithms that can exploit the computational capabilities of many-core accelerators, such as graphics processing units (GPUs), to achieve the desired performance. The main research components include: (1) development of highly accurate and low dispersion numerical methods for solving acoustic wave equation;(2) massively parallel algorithms for efficient numeric and geometric acoustic propagation; (3) software libraries to run millions of threads on GPU clusters to achieve petaflop acoustic performance based on sound field decomposition; and (4) acoustic analysis of complex CAD models. The broader impacts of this effort are interdisciplinary and span across academia and industry, including new application software libraries for many-core accelerators.