PETSc interpolate a function from global to local mesh

General Discussion
1

Dear FreeFem users and developers,

I know which is possible to “recollect” te results obtained in parallel on the tocal mesh to the global (starting) one using the mpiAllReduce(u,uS,mpiCommWorld, mpiSUM) command. Is there a function whch is doing the opposite, so dictributing a global variable on the local mesh?

I need this since I am loading from a file, node by node, a variable of my problem and so it is strictly defined on the global mesh. By the way to perform the computations I need to move it in the local one.

Thank you in advance for your help.

Sincerely

Marco

2

Like this? See this example.

int[int] n2o, rest;
rest = restrict(MhLocal, MhGlobal, n2o); // local/global fespace restriction
uLocal[] = uGlobal[](rest); //restrict global solution to each local mesh
3

I’m not sure that this is exactly what I need. Just to be clearer what I am doing is something like that:

  • I am meshing my domain starting from a point cloud, knowing the load on that point cloud.
mesh Th = triangulate(fname);
  • I get the load from the point cloud file
fespace Vh(Th,P1);

Vh EE;

{
    ifstream file(fname);
    real xx,yy;
    for(int i = 0; i < EE.n; i++)
        file >> xx >> yy >> EE[][i];
}
  • Now what I would like to do is to produce a local mesh using PETSc (buildDmesh(Th)) and then having a VhPar EEpar which is the exact correspondence of the previously defined EE on the obtained local mesh

I hope now it is clearer!

4

I don’t think that any of that should cause a problem… Try:

mesh ThG = triangulate(fname);
fespace VhG(ThG,P1);
VhG EE;
{
    ifstream file(fname);
    real xx,yy;
    for(int i = 0; i < EE.n; i++)
        file >> xx >> yy >> EE[][i];
}
mesh ThPar = ThG;
int[int] n2o, rest;
macro ThN2O()n2o// EOM
buildDmesh(ThPar);
fespace VhPar(ThPar,P1);
VhPar EEpar;
rest = restrict(VhPar, VhG, n2o); // local/global fespace restriction
EEpar[] = EE[](rest); //restrict global solution to each local mesh
5

Unfortunately this option provides me a PETSc error:

image
image1918×753 44.6 KB

6

Please attach a reproducer script (without any fileread dependencies) if you’d like help debugging.

7

Even with the extremely simple condition reported below the procedure leads to the PETSc error:

load "PETSc"
macro dimension()2//EOM
include "macro_ddm.idp"


string fname = "Tungsten_5deg_FF.txt";

//mesh Th = triangulate(fname);
mesh Th = square(10,10);

plot(Th);

fespace Vh(Th,P1);

Vh EE;

EE = x^2;
/*
{
    ifstream file(fname);
    real xx,yy;
    for(int i = 0; i < EE.n; i++)
        file >> xx >> yy >> EE[][i];
}
*/
mesh ThPar = Th;

int[int] n2o, rest;

macro ThN2O()n2o //EOM

buildDmesh(ThPar);

fespace VhPar(ThPar,P1);

VhPar EEpar;

rest = restrict(VhPar,Vh,n2o);

EEpar[] = EE[](rest);

The commented part is the original one which has been substituted by a simpler mesh and EE.

Thank you very much for your help!

8

Ah yes, my mistake. Just change the name of the macro ThN2o to ThParN2o.

9

Oh that’s true! Thank you very much @cmd !

Now I’m ok!

1 Like