Collaborative Research: Hierarchically Parallel Algorithms for Portable and Scalable Performance

点击翻译按钮获取中文摘要

The goal of this research proposed is to develop parallel sparse system solvers for symmetric indefinite and non-symmetric linear systems such as those occurring in computational fluid dynamics, develop parallel techniques for solving time-dependent problems, and construct a rapid prototyping environment for developing and experimenting with algorithms in sparse matrix computations on shared and distributed memory multiprocessors. In particular, the following will be examined: o parallel solution schemes for solving Stokes-like linear problems arising in CFD and other augmented Lagrangian problems such as interior point methods of linear programming, o incomplete Newton methods with Krylov or row projection iterative solvers for the inner iteration, o solution of time-dependent problems using high-order parallel methods in time, o defining and developing sparse matrix primitives analogous to the highly effective BLAS-3's for dense computations, and o implementing sparse factorization and solution schemes. The goal is portable efficiency and scalability, by exploiting parallelism on multiple levels. This in turn will allow the work to be used across a wide spectrum of existing and future computer architectures, instead of simply a particular machine.