view +sbp/InterpOpsMC.m @ 1198:2924b3a9b921 feature/d2_compatible

Add OpSet for fully compatible D2Variable, created from regular D2Variable by replacing d1 by first row of D1. Formal reduction by one order of accuracy at the boundary point.
author Martin Almquist <malmquist@stanford.edu>
date Fri, 16 Aug 2019 14:30:28 -0700
parents 27ce3f653aa7
children
line wrap: on
line source

% Interpolation operators by Mattsson and Carpenter (MC), see
% Mattsson and Carpenter,
% "Stable and Accurate Interpolatino Operators for High-Order Multiblock Finite DIfference Methods",
% https://epubs.siam.org/doi/pdf/10.1137/090750068
%
% Let ^* denote the adjoint. These operators satsify
%
% Iuv2 = Iv2u^*
%
% Both Iu2v and Iv2u have p:th order accuracy, if the interior stencil is
% of order 2p.
%
% This approach leads to a reduction of the convergence rate by one order for
% PDEs with 2nd derivatives in space, as compared to conforming interfaces.
% To obtain full convergence rate, use the order-preserving (OP) operators in
% InterpOpsOP.m
classdef InterpOpsMC < sbp.InterpOps
    properties

        % Structs of interpolation operators, fields .good and .bad
        % Here .good and .bad are the same, but this makes them fit in the
        % OP (order-preserving) framework.
        Iu2v
        Iv2u
    end

    methods
        % m_u, m_v         --   number of grid points along the interface
        % order_u, order_v --   order of accuracy in the different blocks
        function obj = InterpOpsMC(m_u, m_v, order_u, order_v)

            assert(order_u == order_v,...
                    'InterpOpsMC: Different orders of accuracy not available');

            switch order_u
            case 2
                intOpSet = @sbp.implementations.intOpMC_orders_2to2_ratio2to1;
            case 4
                intOpSet = @sbp.implementations.intOpMC_orders_4to4_ratio2to1;
            case 6
                intOpSet = @sbp.implementations.intOpMC_orders_6to6_ratio2to1;
            case 8
                intOpSet = @sbp.implementations.intOpMC_orders_8to8_ratio2to1;
            otherwise
                error('InterpOpsMC: Order of accuracy %d not available.', order_u);
            end

            Iu2v = struct;
            Iv2u = struct;

            if (m_u-1)/(m_v-1) == 2
                % Block u is fine, v is coarse
                m_C = m_v;
                [Iv2u.good, Iu2v.bad] = intOpSet(m_C);
                Iv2u.bad = Iv2u.good;
                Iu2v.good = Iu2v.bad;

            elseif (m_v-1)/(m_u-1) == 2
                % Block v is fine, u is coarse
                m_C = m_u;
                [Iu2v.good, Iv2u.bad] = intOpSet(m_C);
                Iu2v.bad = Iu2v.good;
                Iv2u.good = Iv2u.bad;
            else
                error('InterpOpsMC: Interpolation operators for grid ratio %f have not yet been constructed', (m_u-1)/(m_v-1));
            end

            obj.Iu2v = Iu2v;
            obj.Iv2u = Iv2u;

        end

        function str = string(obj)
            str = [class(obj)];
        end

    end
end