Skip to content

Latest commit

 

History

History
137 lines (82 loc) · 5.01 KB

edit_history.md

File metadata and controls

137 lines (82 loc) · 5.01 KB
Raw

20231008 Summary:

  1. Wrap up code in a module called MESTI
  2. Rename mesti.jl to mesti_main.jl
  3. Add testset.

20230723 Summary:

  1. Add more comments
  2. Add the off-diagonal terms of the epsilon feature.

20230525 Summary:

  1. Let the opts.symmetrize_K feature available for mesti2s() and mesti() for 2D TM cases.
  2. Fix the bug in mesti2s().

mesti.jl

  • Make the opts.symmetrize_K feature available for 2D TM cases.

mesti2s.jl

  • Make the opts.symmetrize_K feature available for 2D TM cases.
  • Make the multiplication of nu occur during the wrap-up stage instead of during the construction of B and C for 2D TM cases.
  • Fix the bug for the implicit expansion in 2D TM cases: S = S .* prefactor -> S = S .* reshape(prefactor,1,:)

setup_longitudinal.jl

  • Make the permutation that switches one propagating channel with one having a complex-conjugated transverse profile available in 2D TM cases

20230517 Summary:

  • Set the default value of opts.use_single_precision_MUMP to be true.

20230504 Summary:

  1. Rename opts.use_keep_401_1 as opts.parallel_dependency_graph.

  2. Add opts.use_single_precision_MUMPS. This enables to use the single-precision MUMPS. We simply convert the double precision to the single precision for the matrices A and B (or K for the APF case).

  3. Extend the element data type of matrices A, B, and C: including SparseMatrixCSC{Complex{Int64}, Int64}

    mesti_matrix_solver.jl

  • Add use_single_precision_MUMP

  • Extend the element data type of matrices A, B, and C: including SparseMatrixCSC{Complex{Int64}, Int64}

    mesti.jl

  • Add use_single_precision_MUMPS

    mesti2s.jl

  • Add use_single_precision_MUMPS

20230503 Summary:

  1. Add branches for the computation of 2D TM fields Ex(y,z), where only epsilon_xx(y,z) is used. Always use the dimension of epsilon_xx to determine if 2D TM computation is required.

  2. Keep as many common lines as possible.

  3. Sessions not implemented:

    1. Self-energy
    2. RGF method
    3. codes related to symmetrize_K in mesti2s

    matrix_build_fdfd_matrix.jl

  • Add extra allowed types for input arguments:

    • epsilon_xx: Matrix.
    • epsilon_yy, epsilon_zz, xBC, xPML: Nothing

  • Add 2D TM checking term use_2D_TM to determine which branch to go if there are differences between 2D and 3D codes.

    • use_2D_TM = true if dims(epsilon_xx) = 2

  • Add 2D TM branches; No change at 3D branches

  • Add a new function handle for 2D TM fields (with "nothing" specified for users) after the main function handle.

    setup_longitudinal.jl

  • Add extra allowed types for input arguments:

    • kxdx_all: Nothing

  • Add extra allowed types for output arguments:

    • side.kxdx_prop: Nothing

  • Add 2D TM checking term use_2D_TM to determine which branch to go if there are differences between 2D and 3D codes.

    • use_2D_TM = true if kxdx_all = nothing && kLambda_x = nothing && ind_zero_kx = nothing

  • Add 2D TM branches; No change at 3D branches

    mesti_build_channels.jl

  • Add extra allowed types for input arguments:

    • nx_Ex, nx_Ey, xBC, ny_Ey, n0: Nothing

  • Add extra allowed types for output arguments:

    • kxdx_all, kxdx_prop: Nothing

  • Add 2D TM checking term use_2D_TM to determine which branch to go if there are differences between 2D and 3D codes.

    • use_2D_TM = true if nx_Ex == nothing && nx_Ey == nothing && ny_Ex != nothing && ny_Ey == nothing
    • use_2D_TM = true if dims(syst.epsilon_xx) = 2

  • Add 2D TM branches; No change at 3D branches

  • Add a new function handle for 2D TM fields (with "nothing" specified for users) after the two main function handles.

    mesti_matrix_solver.jl

  • Edit line 352 by changing "transpose" to its equivalent "permutedims", since the output of the former one has the form Transpose{SparseMatrixCSC{ComplexF64, Int64}} instead of SparseMatrixCSC{ComplexF64, Int64} and the codes throw an error because matrix C does not support this type.

  • Exchange provide_rhs! and set_icntl! at lines 705-706 and 739-739.

    mesti.jl

  • Add extra allowed types for input arguments:

    • Syst.epsilon_xx: Matrix
    • Syst.epsilon_yy, Syst.epsilon_zz, Syst.xBC, Syst.kx_B: Nothing

  • Add 2D TM checking term use_2D_TM to determine which branch to go if there are differences between 2D and 3D codes.

    • use_2D_TM = true if dims(epsilon_xx) = 2

  • Add 2D TM branches, mainly at adding PMLs, building B and C; No change at 3D branches.

  • Edit one line at building matrix C section:

    from C_ii[ii] = [C_ii[ii]; [spzeros(M_ii, nxyz_before) transpose(reshape(data, nxyz_data, M_ii)) spzeros(M_ii, nxyz_after)]]

    to C_ii[ii] = permutedims([transpose(C_ii[ii]) [spzeros(nxyz_before, M_ii); reshape(data, nxyz_data, M_ii); spzeros(nxyz_after, M_ii)]],(2,1))

    Since the later concatenating way (the same as building matrix B) seems to be faster than the former one.

    mesti2s.jl

  • Add 2D TM checking term use_2D_TM to determine which branch to go if there are differences between 2D and 3D codes.

    • use_2D_TM = true if dims(epsilon_xx) = 2

  • Add 2D TM branches; No change at 3D branches