Re: [Getfem-users] Assembling question : possibility of using a vector of complex

[Renard Yves]

Véronique Pham <address@hidden> a écrit :

> Hi everyone,
>
> I would like to ask a question about the assembling using the getfem
> generic_assembly procedure.
> In the getfem_assembling.h file, on the Helmholtz equation case, the real
> and the complex case are dissociate, there are:
>
> - For the complex case:
>
> template<typename MATr, typename MATi, typename VECTr, typename VECTi>
> void asm_Helmholtz_cplx(const MATr &Mr, const MATi &Mi, const mesh_im &mim,
>      const mesh_fem &mf_u, const mesh_fem &mf_data,
>      const VECTr &K_squaredr, const VECTi &K_squaredi,
>      const mesh_region &rg=mesh_region::all_convexes()) {
>    generic_assembly assem("Kr=data$1(#2); Ki=data$2(#2);"
>                "m = comp(Base(#1).Base(#1).Base(#2)); "
>                "M$1(#1,#1)+=sym(m(:,:,i).Kr(i) - "
>                "comp(Grad(#1).Grad(#1))(:,i,:,i));"
>                "M$2(#1,#1)+=sym(m(:,:,i).Ki(i));");
>    assem.push_mi(mim);
>    assem.push_mf(mf_u);
>    assem.push_mf(mf_data);
>    assem.push_data(K_squaredr);
>    assem.push_data(K_squaredi);
>    assem.push_mat(const_cast<MATr&>(Mr));
>    assem.push_mat(const_cast<MATi&>(Mi));
>    assem.assembly(rg);
> }
>
> - For the real case:
>
> template<typename MAT, typename VECT>
> void asm_Helmholtz_real(const MAT &M, const mesh_im &mim,
>      const mesh_fem &mf_u, const mesh_fem &mf_data,
>      const VECT &K_squared,
>      const mesh_region &rg=mesh_region::all_convexes()) {
>   generic_assembly assem("K=data$1(#2);"
>                "m = comp(Base(#1).Base(#1).Base(#2)); "
>                "M$1(#1,#1)+=sym(m(:,:,i).K(i) - "
>                "comp(Grad(#1).Grad(#1))(:,i,:,i));");
>   assem.push_mi(mim);
>   assem.push_mf(mf_u);
>   assem.push_mf(mf_data);
>   assem.push_data(K_squared);
>   assem.push_mat(const_cast<MAT&>(M));
>   assem.assembly(rg);
> }
>
> Does that mean that we cannot use directly a vector of complex
> (interleaved real and imaginary part) and that we have to split it in
> two vectors of real?

The generic assembly treats indeed only real terms. It cannot treat  
directly complex values. But the complex version of asm_Helmholtz do  
itself the
separation into real and imaginary part. This is done with the use of  
gmm::real_part and gmm::imag_part which give writable references.

Yves.