Matrices.LU_solve2(LU, pivots, B);
This function call returns the solution X of the linear systems of equations
P*L*U*X = B;
where P is a permutation matrix (implicitly
defined by vector pivots
),
L is a lower triangular matrix with unit
diagonal elements (lower trapezoidal if m > n), and
U is an upper triangular matrix (upper trapezoidal if m < n).
The matrices of this decomposition are computed with function
Matrices.LU that
returns arguments LU
and pivots
used as input arguments of Matrices.LU_solve2
.
With Matrices.LU
and Matrices.LU_solve2
it is possible to efficiently solve linear systems
with different right hand side matrices. If a linear system of equations with
just one right hand side matrix shall be solved, it is
more convenient to just use the function
Matrices.solve2.
If a unique solution X does not exist (since the LU decomposition is singular), an exception is raised.
The LU factorization is computed with the LAPACK function "dgetrf", i.e., by Gaussian elimination using partial pivoting with row interchanges. Vector "pivots" are the pivot indices, i.e., for 1 ≤ i ≤ min(m,n), row i of matrix A was interchanged with row pivots[i].
Real A[3,3] = [1,2,3; 3,4,5; 2,1,4]; Real B1[3] = [10, 20; 22, 44; 12, 24]; Real B2[3] = [ 7, 14; 13, 26; 10, 20]; Real LU[3,3]; Integer pivots[3]; Real X1[3,2]; Real X2[3,2]; algorithm (LU, pivots) := Matrices.LU(A); X1 := Matrices.LU_solve2(LU, pivots, B1); /* X1 = [3, 6; 2, 4; 1, 2] */ X2 := Matrices.LU_solve2(LU, pivots, B2); /* X2 = [1, 2; 0, 0; 2, 4] */
function LU_solve2 extends Modelica.Icons.Function; input Real LU[:, size(LU, 1)] "L,U factors of Matrices.LU(..) for a square matrix"; input Integer pivots[size(LU, 1)] "Pivots indices of Matrices.LU(..)"; input Real B[size(LU, 1), :] "Right hand side matrix of P*L*U*X=B"; output Real X[size(B, 1), size(B, 2)] "Solution matrix such that P*L*U*X = B"; end LU_solve2;