Spectra
C++ Library For Large Scale Eigenvalue Problems

Upgrade to 1.0.0

Spectra 1.0.0 introduces a lot of API-breaking changes, but the migration should be straightforward following the guide below.

Toolchain

Spectra 1.0.0 requires a compiler supporting the C++11 standard. Any modern C++ compiler should already have this.

Matrix Operation Classes

In most cases you do not need to change anything for the code involving built-in matrix operation classes such as DenseSymMatProd and SparseGenMatProd. However, if you have defined your own class, you need to add a public type definition named Scalar, as the example below shows. The type Scalar indicates the element type of the matrix.

// A user-defined matrix operation class
// representing the matrix A=diag(1, 2, ..., 10)
class MyDiagonalTen
{
public:
    // The line below is new
    using Scalar = double;

    int rows() const { return 10; }
    int cols() const { return 10; }
    // y_out = M * x_in
    void perform_op(const double *x_in, double *y_out) const
    {
        for(int i = 0; i < rows(); i++)
        {
            y_out[i] = x_in[i] * (i + 1);
        }
    }
};

Eigen Solvers

The biggest change happens in the eigen solvers:

  1. The first template parameter Scalar has been removed.
  2. The second template parameter, SelectionRule, has been changed to a runtime parameter selection in the compute() member function.
  3. In the constructor, matrix operation objects are now passed as references instead of pointers.
  4. All enumerations have been converted to enum classes (see the conversion table below).

Below shows the one-to-one conversion of the code reflecting the changes above:

[0.9.0] code:

// Construct matrix operation object using the wrapper class
DenseSymMatProd<double> op(M);

// Construct eigen solver object, requesting the largest three eigenvalues
SymEigsSolver< double, LARGEST_ALGE, DenseSymMatProd<double> > eigs(&op, 3, 6);

// Initialize, and compute with at most 1000 iterations
eigs.init();
int nconv = eigs.compute(1000);

// Retrieve results
Eigen::VectorXd evalues;
if(eigs.info() == SUCCESSFUL)
    evalues = eigs.eigenvalues();

[1.0.0] code:

// Construct matrix operation object using the wrapper class
DenseSymMatProd<double> op(M);

// Construct eigen solver object, requesting the largest three eigenvalues
SymEigsSolver<DenseSymMatProd<double>> eigs(op, 3, 6);

// Initialize, and compute with at most 1000 iterations
eigs.init();
int nconv = eigs.compute(SortRule::LargestAlge, 1000);

// Retrieve results
Eigen::VectorXd evalues;
if(eigs.info() == CompInfo::Successful)
    evalues = eigs.eigenvalues();

Enumeration Conversion

[0.9.0] [1.0.0]
SUCCESSFUL CompInfo::Successful
NOT_COMPUTED CompInfo::NotComputed
NOT_CONVERGING CompInfo::NotConverging
NUMERICAL_ISSUE CompInfo::NumericalIssue
LARGEST_MAGN SortRule::LargestMagn
LARGEST_REAL SortRule::LargestReal
LARGEST_IMAG SortRule::LargestImag
LARGEST_ALGE SortRule::LargestAlge
SMALLEST_MAGN SortRule::SmallestMagn
SMALLEST_REAL SortRule::SmallestReal
SMALLEST_IMAG SortRule::SmallestImag
SMALLEST_ALGE SortRule::SmallestAlge
BOTH_ENDS SortRule::BothEnds
GEIGS_CHOLESKY GEigsMode::Cholesky
GEIGS_REGULAR_INVERSE GEigsMode::RegularInverse
GEIGS_SHIFT_INVERT GEigsMode::ShiftInvert
GEIGS_BUCKLING GEigsMode::Buckling
GEIGS_CAYLEY GEigsMode::Cayley