scientific article; zbMATH DE number 434534

From MaRDI portal
Publication:3139315





The effectiveness of spectral methods for the numerical solution of partial differential equations is somehow downgraded within parallel environments, due to the global communication load needed. The paper examines this issue by implementing a three-dimensional implicit spectral method on an Intel hypercube. More precisely, a multigrid-based implicit solution for the time-dependent, variable coefficient Helmholtz equation \[ u_ t=\nabla a(x,y,z)\cdot\nabla u- b(x,y,z)u+ f(x,y,z,t), \] on three- dimensional tensor product grids is computed. (This is a problem which arises from the Usawa formulation of the incompressible Navier Stokes equations in ocean circulation problems.) Utilization of about \(50\%\) was achieved on a 32 node iPSC/860 hypercube, for a \(64\times 64\times 64\) Fourier-spectral grid; finer grids yield higher utilizations. Chebyshev-spectral grids get more difficulties, as plane-relaxation based multigrids are required. However, by using a semicoarsening multigrid algorithm and by relaxing all multigrid levels concurrently, relatively high utilizations were also achieved, typically \(60\%\) on moderate sized problems. This justifies the authors' claim that spectral methods remain attractive on the current generation of distributed memory computers. For the entire collection see [Zbl 0772.00024].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3139315)