Different results from python GetPFT and scuff-scatter
Closed this issue · 2 comments
Hi Homer,
I found the new interface in libscuffSolver very useful and easy to handle. I am starting to migrate some of my models from command line to python, and I found a problem with the output of GetPFT which differs from that of scuff-scatter. The scattered power is the same, but the absorbed power and rest of magnitudes differ. I have been browsing the source code, but I haven't found any remarkable difference between the different calls.
In the attached zip file, there is a small example with a simple mesh file, the scuffgeo file, Args and a short python script.
Any help is welcome.
José.
I think I have finally found the problems. The first one is that CachedIF in scuffSolver.cc always gets a copy of a PointSource. The second one is that the copy method of PlaneWave and PointSource does not copy the attributes of the base class IncField.
- In
scuffSolver.cc:684the copy of the source toCachedIFtakes place:
else if (IF)
{ G->AssembleRHSVector(Omega, IF, KN);
if (CachedIF){
delete CachedIF;
}
PlaneWave *PW = (PlaneWave *) IF;
if (PW) CachedIF=new PlaneWave(*PW);
PointSource *PS = (PointSource *) IF;
if (PS) CachedIF=new PointSource(*PS);
}
The two if clauses are always true and CachedIF ends up with a copy of PS.
- The copy method of
PlaneWaveinPlaneWave.cc:50:
PlaneWave::PlaneWave(const PlaneWave &PW)
{
memcpy(E0, PW.E0, 3*sizeof(cdouble));
memcpy(nHat, PW.nHat, 3*sizeof(double));
SetRegionLabel(PW.RegionLabel);
}
does not include a copy of the IncField attributes. Therefore, Omega is equal to -1 +0j. The same problem happens for PointSource.
It would be nice to find a nice solution for both problems. I can only make a dirty patch at the moment.
José.