Performance Characteristics of Fast Elliptic Solvers on Parallel Platforms
نویسندگان
چکیده
We present performance analyses of a fast three-dimensional elliptic orthogonal spline collocation (OSC) method on the Connection Machine CM-5, and various network clusters using the Parallel Virtual Machine (PVM) software package. While the OSC method is an eecient approach for solving Poisson, Helmholtz, and other separable PDE problems, it poses serious challenges for parallel architec-tures because of its high communication to computation ratio: O(N 3) data movement to O(N 3 log N) computation. Furthermore, the communication pattern is one of the most expensive: personalized all-to-all. Results show that communication time occupies roughly 5% of total execution time in native message-passing systems (such as NX and CMMD), 10% with PVM layer on MPP's, and roughly 50% for Ethernet networks.
منابع مشابه
Parallel Reuse Methodologies for Elliptic Boundary Value Problems
We describe two parallel frameworks that allow the reuse of the discretization part of sequential general elliptic PDE (partial differential equation) solvers. These parallel reuse methodologies are based on the "divide and conquer" computational paradigm. They have been integrated into the Parallel ELLPACI( problem solving environment that supports PDE computing across many hardware platforms....
متن کاملParallel ELLPACK Elliptic PDE Solvers
Parallel ELLPACK [35, 61] is a machine independent problem solving environment (PSE) that supports PDE (partial di erential equations) computing across many hardware platforms. In this paper we review parallel methodologies based on the \divide and conquer" computational paradigm and their infrastructure for solving general elliptic PDEs. Particularly, we describe those that have been implement...
متن کاملSolvers on advanced parallel architectures with application to partial differential equations and discrete optimisation
This thesis investigates techniques for the solution of partial differential equations (PDE) on advanced parallel architectures comprising central processing units (CPU) and graphics processing units (GPU). Many physical phenomena studied by scientists and engineers aremodelled with PDEs, and these are often computationally expensive to solve. This is one of the main drivers of large-scale comp...
متن کاملParallel Smoothers for Matrix-based Multigrid Methods on Unstructured Meshes Using Multicore CPUs and GPUs
Multigrid methods are efficient and fast solvers for problems typically modeled by partial differential equations of elliptic type. For problems with complex geometries and local singularities stencil-type discrete operators on equidistant Cartesian grids need to be replaced by more flexible concepts for unstructured meshes in order to properly resolve all problem-inherent specifics and for mai...
متن کاملParallel sparse LU factorization on different message passing platforms
Several message passing-based parallel solvers have been developed for general (nonsymmetric) sparse LU factorization with partial pivoting. Existing solvers were mostly deployed and evaluated on parallel computing platforms with high message passing performance (e.g., 1–10 μs in message latency and 100–1000 Mbytes/sec in message throughput) while little attention has been paid on slower platfo...
متن کامل