Home Browse Files

Content

QR Factorization (Complex Variant)

In this example we again compute the \(QR\) factorization and use it for solving a system of linear equations. However, in this example we do not setup matrix \(Q\) explicitly.

Example Code

flens/examples/lapack-complex-unmqr.cc
#include <cxxstd/iostream.h>
#include <flens/flens.cxx>

using namespace std;
using namespace flens;

int
main()
{
    typedef complex<double>             Complex;
    const Complex                       I(0,1);


    GeMatrix<FullStorage<Complex> >     A(4,4);
    DenseVector<Array<Complex> >        b(4);
    DenseVector<Array<Complex> >        tau;
    //DenseVector<Array<Complex> >      work;

    A =  2,   3,  -1,   0,
        -6,  -5,   0,   2,
         2,  -5,   6,  -6,
         4,   6,   2,  -3;
    A *= I;

    b = 20,
       -33,
       -43,
        49;
    b *= I;

    cout << "A = " << A << endl;
    cout << "b = " << b << endl;

    lapack::qrf(A, tau);
    //lapack::qrf(A, tau, work);


    lapack::unmqr(Left, ConjTrans, A, tau, b);
    // lapack::ungqr(A, tau, b, work);

    const auto R = A.upper();
    blas::sv(NoTrans, R, b);

    cout << "x = " << b << endl;
}

Comments on Example Code

Compute the factorization \(A = QR\). Note that the workspace gets created implicitly and temporarily. So you might not want to do this inside a loop.

    lapack::qrf(A, tau);
    //lapack::qrf(A, tau, work);

Compute \(\tilde{b} = Q^T b\). Note that we do not setup \(A\) explicitly.

    lapack::unmqr(Left, ConjTrans, A, tau, b);
    // lapack::ungqr(A, tau, b, work);

Solve \(R x = \tilde{b}\). Vector \(b\) gets overwritten with \(x\).

    const auto R = A.upper();
    blas::sv(NoTrans, R, b);

Compile

$shell> cd flens/examples                                                       
$shell> g++ -std=c++11 -Wall -I../.. -o lapack-complex-unmqr lapack-complex-unmqr.cc

Run

$shell> cd flens/examples                                                       
$shell> ./lapack-complex-unmqr                                                  
A = 
                       (0,2)                        (0,3)                      (-0,-1)                        (0,0) 
                     (-0,-6)                      (-0,-5)                        (0,0)                        (0,2) 
                       (0,2)                      (-0,-5)                        (0,6)                      (-0,-6) 
                       (0,4)                        (0,6)                        (0,2)                      (-0,-3) 
b = 
                      (0,20)                      (-0,-33)                      (-0,-43)                        (0,49) 
x = 
            (1,-2.11698e-15)               (9,1.96902e-16)               (9,-4.7151e-15)              (9,-5.27794e-15)
Home