[ngj]

Good morning,

I have a small problem, which I have not been able to solve so far.

Background:
I am going to perform different operations on a small grid (say 1400 cells):

1. Evaluating the divergence of some field, i.e fvc::div(phi,T)
2. Building the fvScalarMatrix for the same, i.e. fvm::div(phi,T)

The thing is that I am going to perform these two operations approximately 90,000 times per time step/iteration, so there is room for thinking in optimising and utilising prior knowledge of the system.
The prior knowledge is that half of the faces (fixed in time and identical for all grids) have a zero flux, why the construction of fvm::div and fvc::div can probably be reduced by a factor of 2.
This reduction approximately corresponds to 15% of the execution time, so it is worthwhile to spend some time on it.

Problem:
I would still like to utilise the numerical scheme selection in fvSchemes, i.e. the user should merely give e.g.

div(phi,T) Gauss upwind phi;

From this I need to extract the information on the surface interpolation scheme and construct it, such that I can obtain the weights.
So far I have managed to get the divergence scheme information from the fvSchemes by using

ITstream& ss = mesh.divScheme("div(phi,T)");

This is were I am stranded, because I have not been able to transform this information into anything useful for the surfaceInterpolationScheme, which must be constructed as

tmp<surfaceInterpolationScheme<scalar> > surfaceInterpolationScheme::New(const fvMesh& mesh, Istream& schemeData);

The problem is specifically to transform the

ITstream ss

into the correctly formatted

Istream schemeData

Any help is greatly appreaciated.

Kind regards,

Niels

[ngj]

Good evening all,

I have found the solution, and it is rather simple:

Istream& is = mesh.divSchemes("div(phi,T)");
word ss(is);
tmp<surfaceInterpolationScheme<scalar> > sis(Foam::surfaceInterpolationScheme<scalar>::New(mesh, is));

The important part is to construct the word 'ss', because this makes Istream go one word forward in the stream, i.e. once I am constructing the surface interpolation scheme, it is not aware of the existence of the 'Gauss' part in front.

Kind regards

Niels

[Jun_93]

Hi Niels,


I have the same problem to construct limitedSurfaceInterpolationScheme class and have already been stuck to it for two three days. The Gauss token in the itstream drives me nuts. Your solution works finally! Many thanks!!!!

best regards
Jun