opm-simulators/tests/test_dilu.cpp
Tobias Meyer Andersen 5f6c97ff3b add OpenMP parallelized version of DILU.
Implement graphcoloring to expose rows in level sets that that can be
executed in parallel during the sparse triangular solves.
Add copy of A matrix that is reordered to ensure continuous memory reads
when traversing the matrix in level set order.
TODO: add number of threads available as constructor argument in DILU
2023-11-21 15:41:53 +01:00

851 lines
22 KiB
C++

/*
Copyright 2022-2023 SINTEF AS
This file is part of the Open Porous Media project (OPM).
OPM is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
OPM is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with OPM. If not, see <http://www.gnu.org/licenses/>.
*/
#define BOOST_TEST_MODULE TestSeqDILU
#include <config.h>
#include <opm/simulators/linalg/DILU.hpp>
#include <boost/mpl/list.hpp>
#include <boost/test/unit_test.hpp>
#include <dune/common/fmatrix.hh>
#include <dune/istl/bcrsmatrix.hh>
#include <dune/istl/preconditioners.hh>
using NumericTypes = boost::mpl::list<double, float>;
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUDiagIsCorrect2x2NoZeros, T, NumericTypes)
{
/*
Tests that the dilu decomposition mathces the expected result
for a 2x2 matrix with no zero blocks, with block size 2x2.
A
| | 3 1| | 1 0| |
| | 2 1| | 0 1| |
| |
| | 2 0| |-1 0| |
| | 0 2| | 0 -1| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 4;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(0);
row.insert(1);
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[0][1][0][0] = 1.0;
A[0][1][1][1] = 1.0;
A[1][0][0][0] = 2.0;
A[1][0][1][1] = 2.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_00_inv U_01
auto D_11 = A[1][1] - A[1][0] * D_00_inv * A[0][1];
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
auto Dinv = seqdilu.getDiagonal();
// diagonal stores inverse
auto D_00_dilu = Dinv[0];
D_00_dilu.invert();
auto D_11_dilu = Dinv[1];
D_11_dilu.invert();
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(D_00_dilu[i][j], D_00[i][j], 1e-7);
BOOST_CHECK_CLOSE(D_11_dilu[i][j], D_11[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUDiagIsCorrect2x2, T, NumericTypes)
{
/*
Tests that the dilu decomposition mathces the expected result
for a 2x2 matrix, with block size 2x2.
A
| | 3 1| | 1 0| |
| | 2 1| | 0 1| |
| |
| | 0 0| |-1 0| |
| | 0 0| | 0 -1| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 3;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(row.index());
if (row.index() == 0) {
row.insert(row.index() + 1);
}
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[0][1][0][0] = 1.0;
A[0][1][1][1] = 1.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_00_inv U_01 = A_11
auto D_11 = A[1][1];
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
auto Dinv = seqdilu.getDiagonal();
// diagonal stores inverse
auto D_00_dilu = Dinv[0];
D_00_dilu.invert();
auto D_11_dilu = Dinv[1];
D_11_dilu.invert();
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(D_00_dilu[i][j], D_00[i][j], 1e-7);
BOOST_CHECK_CLOSE(D_11_dilu[i][j], D_11[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUApplyIsCorrectNoZeros, T, NumericTypes)
{
/*
Tests that applying the dilu preconditioner mathces the expected result
for a 2x2 matrix with no zero blocks, with block size 2x2.
A x = b
| | 3 1| | 1 0| | | |1| | | |2| |
| | 2 1| | 0 1| | | |2| | | |1| |
| | x | | = | |
| | 2 0| |-1 0| | | |1| | | |3| |
| | 0 2| | 0 -1| | | |1| | | |4| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 4;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(0);
row.insert(1);
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[0][1][0][0] = 1.0;
A[0][1][1][1] = 1.0;
A[1][0][0][0] = 2.0;
A[1][0][1][1] = 2.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
Vector x(2);
x[0][0] = 1.0;
x[0][1] = 2.0;
x[1][0] = 1.0;
x[1][1] = 1.0;
Vector b(2);
b[0][0] = 2.0;
b[0][1] = 1.0;
b[1][0] = 3.0;
b[1][1] = 4.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11= A_11 - L_10 D_00_inv U_01
auto D_11 = A[1][1] - A[1][0] * D_00_inv * A[0][1];
auto D_11_inv = D_11;
D_11_inv.invert();
// define: z = M^-1(b - Ax)
// where: M = (D + L_A) A^-1 (D + U_A)
// lower triangular solve: (E + L) y = b - Ax
// y_0 = D_00_inv*[b_0 - (A_00*x_0 + A_01*x_1)]
Vector y(2);
auto rhs = b[0];
A[0][0].mmv(x[0], rhs);
A[0][1].mmv(x[1], rhs);
D_00_inv.mv(rhs, y[0]);
// y_1 = D_11_inv*(b_1 - (A_10*x_0 + A_11*x_1) - A_10*y_0)
rhs = b[1];
A[1][0].mmv(x[0], rhs);
A[1][1].mmv(x[1], rhs);
A[1][0].mmv(y[0], rhs);
D_11_inv.mv(rhs, y[1]);
// upper triangular solve: (E + U) z = Ey
// z_1 = y_1
Vector z(2);
z[1] = y[1];
// z_0 = y_0 - D_00_inv*A_01*z_1
z[0] = y[0];
auto temp = D_00_inv * A[0][1];
temp.mmv(z[1], z[0]);
// x_k+1 = x_k + z
Vector new_x = x;
new_x += z;
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
seqdilu.apply(x, b);
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(x[i][j], new_x[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUApplyIsCorrect1, T, NumericTypes)
{
/*
Tests that applying the dilu preconditioner mathces the expected result
for a 2x2 matrix, with block size 2x2.
A x = b
| | 3 1| | 1 0| | | |1| | | |2| |
| | 2 1| | 0 1| | | |2| | | |1| |
| | x | | = | |
| | 0 0| |-1 0| | | |1| | | |3| |
| | 0 0| | 0 -1| | | |1| | | |4| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 3;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(row.index());
if (row.index() == 0) {
row.insert(row.index() + 1);
}
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[0][1][0][0] = 1.0;
A[0][1][1][1] = 1.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
Vector x(2);
x[0][0] = 1.0;
x[0][1] = 2.0;
x[1][0] = 1.0;
x[1][1] = 1.0;
Vector b(2);
b[0][0] = 2.0;
b[0][1] = 1.0;
b[1][0] = 3.0;
b[1][1] = 4.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_0_inv U_01 = A_11
auto D_11 = A[1][1];
auto D_11_inv = D_11;
D_11_inv.invert();
// define: z = M^-1(b - Ax)
// where: M = (D + L_A) A^-1 (D + U_A)
// lower triangular solve: (E + L) y = b - Ax
// y_0 = D_00_inv*[b_0 - (A_00*x_0 + A_01*x_1)]
Vector y(2);
auto rhs = b[0];
A[0][0].mmv(x[0], rhs);
A[0][1].mmv(x[1], rhs);
D_00_inv.mv(rhs, y[0]);
// y_1 = D_11_inv*(b_1 - (A_10*x_0 + A_11*x_1) - A_10*y_0) = D_11_inv*(b_1 - A_11*x_1)
rhs = b[1];
A[1][1].mmv(x[1], rhs);
D_11_inv.mv(rhs, y[1]);
// upper triangular solve: (E + U) z = Ey
// z_1 = y_1
Vector z(2);
z[1] = y[1];
// z_0 = y_0 - D_00_inv*A_01*z_1
z[0] = y[0];
auto temp = D_00_inv * A[0][1];
temp.mmv(z[1], z[0]);
// x_k+1 = x_k + z
Vector new_x = x;
new_x += z;
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
seqdilu.apply(x, b);
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(x[i][j], new_x[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUApplyIsCorrect2, T, NumericTypes)
{
/*
Tests that applying the dilu preconditioner mathces the expected result
for a 2x2 matrix, with block size 2x2.
A x = b
| | 3 1| | 0 0| | | |1| | | |2| |
| | 2 1| | 0 0| | | |2| | | |1| |
| | x | | = | |
| | 2 0| |-1 0| | | |1| | | |3| |
| | 0 2| | 0 -1| | | |1| | | |4| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 3;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(row.index());
if (row.index() == 1) {
row.insert(row.index() - 1);
}
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[1][1][0][0] = 2.0;
A[1][1][1][1] = 2.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
Vector x(2);
x[0][0] = 1.0;
x[0][1] = 2.0;
x[1][0] = 1.0;
x[1][1] = 1.0;
Vector b(2);
b[0][0] = 2.0;
b[0][1] = 1.0;
b[1][0] = 3.0;
b[1][1] = 4.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_0_inv U_01 = A_11
auto D_11 = A[1][1];
auto D_11_inv = D_11;
D_11_inv.invert();
// define: z = M^-1(b - Ax)
// where: M = (D + L_A) A^-1 (D + U_A)
// lower triangular solve: (E + L) y = b - Ax
// y_0 = D_00_inv*[b_0 - (A_00*x_0 + A_01*x_1)] = D_00_inv*[b_0 - A_00*x_0]
Vector y(2);
auto rhs = b[0];
A[0][0].mmv(x[0], rhs);
D_00_inv.mv(rhs, y[0]);
// y_1 = D_11_inv*(b_1 - (A_10*x_0 + A_11*x_1) - A_10*y_0)
rhs = b[1];
A[1][1].mmv(x[1], rhs);
D_11_inv.mv(rhs, y[1]);
// upper triangular solve: (E + U) z = Ey
// z_1 = y_1
Vector z(2);
z[1] = y[1];
// z_0 = y_0 - D_00_inv*A_01*z_1 = y_0
z[0] = y[0];
// x_k+1 = x_k + z
Vector new_x = x;
new_x += z;
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
seqdilu.apply(x, b);
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(x[i][j], new_x[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUDiagIsCorrect3x3, T, NumericTypes)
{
/*
Tests that the dilu decomposition mathces the expected result
for a 3x3 matrix, with block size 3x3.
A
| | 3 1 2| | 0 0 0| | 0 0 0| |
| | 2 3 1| | 0 0 0| | 0 0 0| |
| | 2 1 0| | 0 0 0| | 0 0 0| |
| |
| | 0 0 0| | 1 0 1| | 1 0 2| |
| | 0 0 0| | 4 1 0| | 0 1 1| |
| | 0 0 0| | 3 1 3| | 0 1 3| |
| |
| | 0 0 0| | 1 0 2| | 1 3 2| |
| | 0 0 0| | 0 1 4| | 2 1 3| |
| | 0 0 0| | 5 1 1| | 3 1 2| |
*/
const int N = 3;
constexpr int bz = 3;
const int nonZeroes = 5;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
if (row.index() == 0) {
row.insert(row.index());
} else if (row.index() == 1) {
row.insert(row.index());
row.insert(row.index() + 1);
} else if (row.index() == 2) {
row.insert(row.index() - 1);
row.insert(row.index());
}
}
A[0][0][0][0] = 3.0;
A[1][1][0][0] = 1.0;
A[1][2][0][0] = 1.0;
A[0][0][0][1] = 1.0;
A[1][1][0][1] = 0.0;
A[1][2][0][1] = 0.0;
A[0][0][0][2] = 2.0;
A[1][1][0][2] = 1.0;
A[1][2][0][2] = 2.0;
A[0][0][1][0] = 2.0;
A[1][1][1][0] = 4.0;
A[1][2][1][0] = 0.0;
A[0][0][1][1] = 3.0;
A[1][1][1][1] = 1.0;
A[1][2][1][1] = 1.0;
A[0][0][1][2] = 1.0;
A[1][1][1][2] = 0.0;
A[1][2][1][2] = 1.0;
A[0][0][2][0] = 2.0;
A[1][1][2][0] = 3.0;
A[1][2][2][0] = 0.0;
A[0][0][2][1] = 1.0;
A[1][1][2][1] = 1.0;
A[1][2][2][1] = 1.0;
A[0][0][2][2] = 0.0;
A[1][1][2][2] = 3.0;
A[1][2][2][2] = 3.0;
A[2][1][0][0] = 1.0;
A[2][2][0][0] = 1.0;
A[2][1][0][1] = 0.0;
A[2][2][0][1] = 3.0;
A[2][1][0][2] = 2.0;
A[2][2][0][2] = 2.0;
A[2][1][1][0] = 0.0;
A[2][2][1][0] = 2.0;
A[2][1][1][1] = 1.0;
A[2][2][1][1] = 1.0;
A[2][1][1][2] = 4.0;
A[2][2][1][2] = 3.0;
A[2][1][2][0] = 5.0;
A[2][2][2][0] = 3.0;
A[2][1][2][1] = 1.0;
A[2][2][2][1] = 1.0;
A[2][1][2][2] = 1.0;
A[2][2][2][2] = 2.0;
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_00_inv U_01 = A_11
auto D_11 = A[1][1];
auto D_11_inv = D_11;
D_11_inv.invert();
// D_22 = A_22 - A_20 D_00_inv A_02 - A_21 D_11_inv A_12 = A_22 - A_21 D_11_inv A_12
auto D_22 = A[2][2] - A[2][1] * D_11_inv * A[1][2];
auto D_22_inv = D_22;
D_22_inv.invert();
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
auto Dinv = seqdilu.getDiagonal();
// diagonal stores inverse
auto D_00_dilu = Dinv[0];
D_00_dilu.invert();
auto D_11_dilu = Dinv[1];
D_11_dilu.invert();
auto D_22_dilu = Dinv[2];
D_22_dilu.invert();
for (int i = 0; i < 3; ++i) {
for (int j = 0; j < 3; ++j) {
BOOST_CHECK_CLOSE(D_00_dilu[i][j], D_00[i][j], 1e-7);
BOOST_CHECK_CLOSE(D_11_dilu[i][j], D_11[i][j], 1e-7);
BOOST_CHECK_CLOSE(D_22_dilu[i][j], D_22[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUApplyIsCorrect3, T, NumericTypes)
{
/*
Tests that applying the dilu preconditioner mathces the expected result
for a 3x3 matrix, with block size 3x3.
A x = b
| | 3 1 2| | 0 0 0| | 0 0 0| | | |1| | | |2| |
| | 2 3 1| | 0 0 0| | 0 0 0| | | |2| | | |1| |
| | 2 1 0| | 0 0 0| | 0 0 0| | | |3| | | |2| |
| | | | | |
| | 0 0 0| | 1 0 1| | 1 0 2| | | |1| | | |2| |
| | 0 0 0| | 4 1 0| | 0 1 1| | x | |3| | = | |3| |
| | 0 0 0| | 3 1 3| | 0 1 3| | | |2| | | |2| |
| | | | | |
| | 0 0 0| | 1 0 2| | 1 3 2| | | |1| | | |0| |
| | 0 0 0| | 0 1 4| | 2 1 3| | | |0| | | |2| |
| | 0 0 0| | 5 1 1| | 3 1 2| | | |2| | | |1| |
*/
const int N = 3;
constexpr int bz = 3;
const int nonZeroes = 5;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
if (row.index() == 0) {
row.insert(row.index());
} else if (row.index() == 1) {
row.insert(row.index());
row.insert(row.index() + 1);
} else if (row.index() == 2) {
row.insert(row.index() - 1);
row.insert(row.index());
}
}
A[0][0][0][0] = 3.0;
A[1][1][0][0] = 1.0;
A[1][2][0][0] = 1.0;
A[0][0][0][1] = 1.0;
A[1][1][0][1] = 0.0;
A[1][2][0][1] = 0.0;
A[0][0][0][2] = 2.0;
A[1][1][0][2] = 1.0;
A[1][2][0][2] = 2.0;
A[0][0][1][0] = 2.0;
A[1][1][1][0] = 4.0;
A[1][2][1][0] = 0.0;
A[0][0][1][1] = 3.0;
A[1][1][1][1] = 1.0;
A[1][2][1][1] = 1.0;
A[0][0][1][2] = 1.0;
A[1][1][1][2] = 0.0;
A[1][2][1][2] = 1.0;
A[0][0][2][0] = 2.0;
A[1][1][2][0] = 3.0;
A[1][2][2][0] = 0.0;
A[0][0][2][1] = 1.0;
A[1][1][2][1] = 1.0;
A[1][2][2][1] = 1.0;
A[0][0][2][2] = 0.0;
A[1][1][2][2] = 3.0;
A[1][2][2][2] = 3.0;
A[2][1][0][0] = 1.0;
A[2][2][0][0] = 1.0;
A[2][1][0][1] = 0.0;
A[2][2][0][1] = 3.0;
A[2][1][0][2] = 2.0;
A[2][2][0][2] = 2.0;
A[2][1][1][0] = 0.0;
A[2][2][1][0] = 2.0;
A[2][1][1][1] = 1.0;
A[2][2][1][1] = 1.0;
A[2][1][1][2] = 4.0;
A[2][2][1][2] = 3.0;
A[2][1][2][0] = 5.0;
A[2][2][2][0] = 3.0;
A[2][1][2][1] = 1.0;
A[2][2][2][1] = 1.0;
A[2][1][2][2] = 1.0;
A[2][2][2][2] = 2.0;
Vector x(3);
x[0][0] = 1.0;
x[1][0] = 1.0;
x[2][0] = 1.0;
x[0][1] = 2.0;
x[1][1] = 3.0;
x[2][1] = 0.0;
x[0][2] = 3.0;
x[1][2] = 2.0;
x[2][2] = 2.0;
Vector b(3);
b[0][0] = 2.0;
b[1][0] = 2.0;
b[2][0] = 0.0;
b[0][1] = 1.0;
b[1][1] = 3.0;
b[2][1] = 2.0;
b[0][2] = 2.0;
b[1][2] = 2.0;
b[2][2] = 1.0;
// D_00 = A_00
auto D_00 = A[0][0];
auto D_00_inv = D_00;
D_00_inv.invert();
// D_11 = A_11 - L_10 D_00_inv U_01
// = A_11
auto D_11 = A[1][1];
auto D_11_inv = D_11;
D_11_inv.invert();
// D_22 = A_22 - A_20 D_00_inv A_02 - A_21 D_11_inv A_12
// = A_22 - A_21 D_11_inv A_12
auto D_22 = A[2][2] - A[2][1] * D_11_inv * A[1][2];
auto D_22_inv = D_22;
D_22_inv.invert();
// define: z = M^-1(b - Ax)
// where: M = (D + L_A) A^-1 (D + U_A)
// lower triangular solve: (E + L) y = b - Ax
Vector y(3);
// y_0 = D_00_inv*[b_0 - (A_00*x_0 + A_01*x_1)]
// = D_00_inv*[b_0 - A_00*x_0]
auto rhs = b[0];
A[0][0].mmv(x[0], rhs);
D_00_inv.mv(rhs, y[0]);
// y_1 = D_11_inv*(b_1 - (A_10*x_0 + A_11*x_1 + A_12*x_2) - A_10*y_0)
// = D_11_inv*(b_1 - A_11*x_1)
rhs = b[1];
A[1][1].mmv(x[1], rhs);
A[1][2].mmv(x[2], rhs);
D_11_inv.mv(rhs, y[1]);
// y_2 = D_22_inv*(b_2 - (A_20*x_0 + A_21*x_1 + A_22*x_2) - (A_20*y_0 + A_21*y_1))
// = D_22_inv*(b_2 - (A_21*x_1 + A_22*x_2) - (A_21*y_1))
rhs = b[2];
A[2][1].mmv(x[1], rhs);
A[2][2].mmv(x[2], rhs);
A[2][1].mmv(y[1], rhs);
D_22_inv.mv(rhs, y[2]);
// upper triangular solve: (E + U) z = Ey
Vector z(3);
// z_2 = y_2
z[2] = y[2];
// z_1 = y_1 - D_11_inv*A_12*z_2
z[1] = y[1];
auto temp = D_11_inv * A[1][2];
temp.mmv(z[2], z[1]);
// z_0 = y_0 - D_00_inv(A_01*z_1 + A_02*z_2)
// z_0 = y_0
z[0] = y[0];
// x_k+1 = x_k + z
Vector new_x = x;
new_x += z;
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
seqdilu.apply(x, b);
for (int i = 0; i < 3; ++i) {
for (int j = 0; j < 3; ++j) {
BOOST_CHECK_CLOSE(x[i][j], new_x[i][j], 1e-7);
}
}
}
BOOST_AUTO_TEST_CASE_TEMPLATE(SeqDILUApplyIsEqualToDuneSeqILUApply, T, NumericTypes)
{
/*
Tests that applying the DILU preconditioner is equivalent to applying a ILU preconditioner
for a block diagonal matrix.
A x = b
| | 3 1| | 0 0| | | |1| | | |2| |
| | 2 1| | 0 0| | | |2| | | |1| |
| | x | | = | |
| | 0 0| |-1 0| | | |1| | | |3| |
| | 0 0| | 0 -1| | | |1| | | |4| |
*/
const int N = 2;
constexpr int bz = 2;
const int nonZeroes = 2;
using Matrix = Dune::BCRSMatrix<Dune::FieldMatrix<double, bz, bz>>;
using Vector = Dune::BlockVector<Dune::FieldVector<double, bz>>;
Matrix A(N, N, nonZeroes, Matrix::row_wise);
for (auto row = A.createbegin(); row != A.createend(); ++row) {
row.insert(row.index());
}
A[0][0][0][0] = 3.0;
A[0][0][0][1] = 1.0;
A[0][0][1][0] = 2.0;
A[0][0][1][1] = 1.0;
A[1][1][0][0] = -1.0;
A[1][1][1][1] = -1.0;
Dune::MultithreadDILU<Matrix, Vector, Vector> seqdilu(A);
Dune::SeqILU<Matrix, Vector, Vector> seqilu(A, 1.0);
Vector dilu_x(2);
dilu_x[0][0] = 1.0;
dilu_x[0][1] = 2.0;
dilu_x[1][0] = 1.0;
dilu_x[1][1] = 1.0;
Vector dilu_b(2);
dilu_b[0][0] = 2.0;
dilu_b[0][1] = 1.0;
dilu_b[1][0] = 3.0;
dilu_b[1][1] = 4.0;
Vector ilu_x(2);
ilu_x[0][0] = 1.0;
ilu_x[0][1] = 2.0;
ilu_x[1][0] = 1.0;
ilu_x[1][1] = 1.0;
Vector ilu_b(2);
ilu_b[0][0] = 2.0;
ilu_b[0][1] = 1.0;
ilu_b[1][0] = 3.0;
ilu_b[1][1] = 4.0;
seqdilu.apply(dilu_x, dilu_b);
seqilu.apply(ilu_x, ilu_b);
for (int i = 0; i < 2; ++i) {
for (int j = 0; j < 2; ++j) {
BOOST_CHECK_CLOSE(dilu_x[i][j], ilu_x[i][j], 1e-7);
}
}
}