Re: [ff3d-users] possibility to define x = convect([U,V,W],dt,Z); outside a solve command ?

[Stephane Del Pino]

Le mercredi 11 juillet 2007, Benoit_Desjardins a écrit :
> Hello Stephane
Hello Benoit.

> thanks for your help on Robin boundary conditions !
>
> There is another question I would like to ask you :
>
> As far as I remember, it was possible in freefem2D to define an
> auxiliary variable on a mesh M
>
> x = convect([U,V,W],dt,Z);
>
> outside of a "solve" command.
> It does not seem to be the case in ff3d (I get a "Segmentation fault").
The segfault is not normal. This is a bug. Can you send me the example?
How ever, it is possible to define a function:
        function c = convect([U,V,W],dt,Z);
this defines the function c to be the expression convect([U,V,W],dt,Z)
> My idea was to save CP time by pre-computing "x" and use it in several
> "solve" loops.
In that case, you must store it in a fem function, the previous example stores 
only the function (convection) but not the values :
        femfunction c(Z) = convect([U,V,W],dt,Z);
This evaluates the function at degrees of freedom. It might be better to solve 
a L^2 projection to compute c, since the convection will be evaluated a 
quadrature vertices.

Best regards,
Stéphane.