Dear All Can someone suggest me for how to solve the 3D heat equation implicitly using ADI please. In brief, I would like to solve the 3D heat equation with a constant diffusion using finite differences. In 1D, I am using the Thomas algorithm in serial. However, since the 3D problem is large, I need to parallelise it. Any pointers will be appreciated. thanks Sanjay
On Tue, Mar 6, 2012 at 9:10 AM, Kharche, Sanjay < [email protected]> wrote:
Dear All
Can someone suggest me for how to solve the 3D heat equation implicitly using ADI please.
In brief, I would like to solve the 3D heat equation with a constant diffusion using finite differences. In 1D, I am using the Thomas algorithm in serial. However, since the 3D problem is large, I need to parallelise it. Any pointers will be appreciated.
ADI is from the last century. Use PETSc's Geometric Multigrid. If you look at SNES ex50 and the tutorial runs for it, you can see how this works. Matt
thanks Sanjay
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
On Tue, Mar 6, 2012 at 09:38, Matthew Knepley <[email protected]> wrote:
ADI is from the last century. Use PETSc's Geometric Multigrid. If you look at SNES ex50 and the tutorial runs for it, you can see how this works.
ex5.c is a better example if you are solving a scalar problem. Sanjay, ADI has poor memory memory characteristics, does not scale well in parallel, and is very restrictive. As Matt says, multigrid is a better and more general approach, it will converge faster and scale better.
participants (3)
-
Jed Brown -
Kharche, Sanjay -
Matthew Knepley