Equation solvers for distributed-memory computers (original) (raw)
Related papers
The Massively Parallel Processor for problems in fluid dynamics
Computer Physics Communications, 1985
We describe the use of the Massively Parallel Processor for certain large-scale scientific computations. Methods and performances are discussed for matrix calculations, tridiagonal and cyclic tridiagonal systems, iterative solution of discretized elliptic equations, the driven cavity problem and the shallow water equations.
Parallel Iterative Methods in Modern Physical Applications
Lecture Notes in Computer Science, 2002
Solving large sparse linear systems is a computationallyintensive component of many important large-scale applications. We present a few experiments stemming from a number of realistic applications including magneto-hydrodynamics structural mechanics, and ultrasound modeling, which have become possible due to the advances in parallel iterative solution techniques. Among such techniques is a recently developed Parallel Algebraic Recursive Multilevel Solver (pARMS). This is a distributed-memory iterative method that adopts the general framework of distributed sparse matrices and relies on solving the resulting distributed Schur complement systems. We discuss some issues related to parallel performance for various linear systems which arise in realistic applications. In particular, we consider the effect of different parameters and algorithms on the overall performance.
Large electromagnetic problem on large scale parallel computing systems
2012 International Conference on High Performance Computing & Simulation (HPCS), 2012
This paper deals with the electromagnetic modeling of large and complex electrical structures by means of large scale parallel systems, such as Grid Computing (GC) and supercomputer. Transmission-Line Matrix (TLM) modeling method is applied to homogeneous volumes. The planar structures are modelled with the mode matching approach. The results prove the benefits of the GC and supercomputer environments to solve electrically large structures. A prediction model for computing performances on grid, based on a hybrid approach that combines a historic-based prediction and an application profile-based prediction, has been developped. The predicted values are in good agreement with the measured values.
Scientific Computations on Modern Parallel Vector Systems
2004
Computational scientists have seen a frustrating trend of stagnating application performance despite dramatic increases in the claimed peak capability of high performance computing systems. This trend has been widely attributed to the use of superscalar-based commodity components who's architectural designs offer a balance between memory performance, network capability, and execution rate that is poorly matched to the requirements of large-scale numerical computations. Recently, two innovative parallel-vector architectures have become operational: the Japanese Earth Simulator (ES) and the Cray X1. In order to quantify what these modern vector capabilities entail for the scientists that rely on modeling and simulation, it is critical to evaluate this architectural paradigm in the context of demanding computational algorithms. Our evaluation study examines four diverse scientific applications with the potential to run at ultrascale, from the areas of plasma physics, material science, astrophysics, and magnetic fusion. We compare performance between the vector-based ES and X1, with leading superscalar-based platforms: the IBM Power3/4 and the SGI Altix. Our research team was the first international group to conduct a performance evaluation study at the Earth Simulator Center; remote ES access in not available. Results demonstrate that the vector systems achieve excellent performance on our application suite -the highest of any architecture tested to date. However, vectorization of a particle-incell code highlights the potential difficulty of expressing irregularly structured algorithms as data-parallel programs.
1992
An efficient three-dimensional unstructured Euler solver has been parallelized on a Cray Y-MP C90 shared memory computer and on an Intel Touchstone Delta distributed memory computer. Both machines yield comparable performance rates. However, the availability of sophisticated software tools enabled the parallelization of EUL3D on the shared memory vector/parallel CRAY Y-MP C90 with minimal user input. On the other hand, the implementation on the distributed memory massively parallel architecture of the Intel Touchstone Delta machine is considerably more involved. As massively parallel software tools become more mature, the task of developing or porting software to such machines should diminish. It has also been shown that with today's supercomputers, and with efficient codes such as EUL3D, the aerodynamic characteristics of complex vehicles can be computed in a matter of minutes, making design use feasible
Advances in electromagnetic modelling through high performance computing
Physica C: Superconductivity, 2006
Under the DOE SciDAC project on Accelerator Science and Technology, a suite of electromagnetic codes has been under development at SLAC that are based on unstructured grids for higher accuracy, and use parallel processing to enable large-scale simulation. The new modelling capability is supported by SciDAC collaborations on meshing, solvers, refinement, optimization and visualization. These advances in computational science are described and the application of the parallel eigensolver Omega3P to the cavity design for the International Linear Collider is discussed.
A tridiagonal solver for massively parallel computers
Advances in Engineering Software, 1998
This paper describes a tridiagonal solver for solving large systems of linear equations on massively parallel computers. Assuming the use of NP processors, the original tridiagonal matrix is divided into NP portions by NP ¹ 1 separators, with each processor storing one portion and the NP ¹ 1 separators. Communications are needed only for those arithmetic operations involved with the NP ¹ 1 separators. Numerical performance of this solver in solving 38.4 million equations on 128 Intel iPSC/860 processors (Gamma) is presented, which shows a speed-up of more than 98.
Problem-solving environment for parallel computers
Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 1997
A program environment is presented that enables a cooperative design of high-technology products demanding interdisciplinary computer simulations for their optimisation. The system includes a multi-site, multi-machine, distributed data manager accessible by a data monitor. A client-server concept supports the user of massively parallel machines. The exchange of data between different executable scientific application modules is performed through the data base. Special modules to define the geometry, to decompose the domain in subdomains, to construct adaptive meshes in each subdomain, and to represent graphically the results run on workstations. Direct and iterative matrix solvers are embedded in the system. They optimally run on massively parallel machines. As a demonstration of the system. the high frequency wave generator (Gyrotron) application is presented.