nalgebra-lapack: add schur decomposition.

This commit is contained in:
Sébastien Crozet 2017-08-06 19:41:12 +02:00 committed by Sébastien Crozet
parent a7bce9cf3f
commit c616c3ddef
4 changed files with 216 additions and 1 deletions

View File

@ -21,16 +21,18 @@ mod cholesky;
mod lu; mod lu;
mod qr; mod qr;
mod hessenberg; mod hessenberg;
mod schur;
use num_complex::Complex; use num_complex::Complex;
pub use self::svd::SVD; pub use self::svd::SVD;
pub use self::cholesky::{Cholesky, CholeskyScalar}; pub use self::cholesky::{Cholesky, CholeskyScalar};
pub use self::lu::{LU, LUScalar}; pub use self::lu::{LU, LUScalar};
pub use self::eigen::RealEigensystem; pub use self::eigen::Eigen;
pub use self::symmetric_eigen::SymmetricEigen; pub use self::symmetric_eigen::SymmetricEigen;
pub use self::qr::QR; pub use self::qr::QR;
pub use self::hessenberg::Hessenberg; pub use self::hessenberg::Hessenberg;
pub use self::schur::RealSchur;
trait ComplexHelper { trait ComplexHelper {

View File

@ -0,0 +1,191 @@
use num::Zero;
use num_complex::Complex;
use alga::general::Real;
use ::ComplexHelper;
use na::{Scalar, DefaultAllocator, Matrix, VectorN, MatrixN};
use na::dimension::{Dim, U1};
use na::storage::Storage;
use na::allocator::Allocator;
use lapack::fortran as interface;
/// Eigendecomposition of a real square matrix with real eigenvalues.
pub struct RealSchur<N: Scalar, D: Dim>
where DefaultAllocator: Allocator<N, D> +
Allocator<N, D, D> {
re: VectorN<N, D>,
im: VectorN<N, D>,
t: MatrixN<N, D>,
q: MatrixN<N, D>
}
impl<N: EigenScalar + Real, D: Dim> RealSchur<N, D>
where DefaultAllocator: Allocator<N, D, D> +
Allocator<N, D> {
/// Computes the eigenvalues and real Schur foorm of the matrix `m`.
///
/// Panics if the method did not converge.
pub fn new(m: MatrixN<N, D>) -> Self {
Self::try_new(m).expect("RealSchur decomposition: convergence failed.")
}
/// Computes the eigenvalues and real Schur foorm of the matrix `m`.
///
/// Returns `None` if the method did not converge.
pub fn try_new(mut m: MatrixN<N, D>) -> Option<Self> {
assert!(m.is_square(), "Unable to compute the eigenvalue decomposition of a non-square matrix.");
let (nrows, ncols) = m.data.shape();
let n = nrows.value();
let lda = n as i32;
let mut info = 0;
let mut wr = unsafe { Matrix::new_uninitialized_generic(nrows, U1) };
let mut wi = unsafe { Matrix::new_uninitialized_generic(nrows, U1) };
let mut q = unsafe { Matrix::new_uninitialized_generic(nrows, ncols) };
// Placeholders:
let mut bwork = [ 0i32 ];
let mut unused = 0;
let lwork = N::xgees_work_size(b'V', b'N', n as i32, m.as_mut_slice(), lda, &mut unused,
wr.as_mut_slice(), wi.as_mut_slice(), q.as_mut_slice(), n as i32,
&mut bwork, &mut info);
lapack_check!(info);
let mut work = unsafe { ::uninitialized_vec(lwork as usize) };
N::xgees(b'V', b'N', n as i32, m.as_mut_slice(), lda, &mut unused,
wr.as_mut_slice(), wi.as_mut_slice(), q.as_mut_slice(),
n as i32, &mut work, lwork, &mut bwork, &mut info);
lapack_check!(info);
Some(RealSchur { re: wr, im: wi, t: m, q: q })
}
/// Retrieves the unitary matrix `Q` and the upper-quasitriangular matrix `T` such that the
/// decomposed matrix equals `Q * T * Q.transpose()`.
pub fn unpack(self) -> (MatrixN<N, D>, MatrixN<N, D>) {
(self.q, self.t)
}
/// Computes the real eigenvalues of the decomposed matrix.
///
/// Return `None` if some eigenvalues are complex.
pub fn eigenvalues(&self) -> Option<VectorN<N, D>> {
if self.im.iter().all(|e| e.is_zero()) {
Some(self.re.clone())
}
else {
None
}
}
/// Computes the complex eigenvalues of the decomposed matrix.
pub fn complex_eigenvalues(&self) -> VectorN<Complex<N>, D>
where DefaultAllocator: Allocator<Complex<N>, D> {
let mut out = unsafe { VectorN::new_uninitialized_generic(self.t.data.shape().0, U1) };
for i in 0 .. out.len() {
out[i] = Complex::new(self.re[i], self.im[i])
}
out
}
}
/*
*
* Lapack functions dispatch.
*
*/
pub trait EigenScalar: Scalar {
fn xgees(jobvs: u8,
sort: u8,
// select: ???
n: i32,
a: &mut [Self],
lda: i32,
sdim: &mut i32,
wr: &mut [Self],
wi: &mut [Self],
vs: &mut [Self],
ldvs: i32,
work: &mut [Self],
lwork: i32,
bwork: &mut [i32],
info: &mut i32);
fn xgees_work_size(jobvs: u8,
sort: u8,
// select: ???
n: i32,
a: &mut [Self],
lda: i32,
sdim: &mut i32,
wr: &mut [Self],
wi: &mut [Self],
vs: &mut [Self],
ldvs: i32,
bwork: &mut [i32],
info: &mut i32)
-> i32;
}
macro_rules! real_eigensystem_scalar_impl (
($N: ty, $xgees: path) => (
impl EigenScalar for $N {
#[inline]
fn xgees(jobvs: u8,
sort: u8,
// select: ???
n: i32,
a: &mut [$N],
lda: i32,
sdim: &mut i32,
wr: &mut [$N],
wi: &mut [$N],
vs: &mut [$N],
ldvs: i32,
work: &mut [$N],
lwork: i32,
bwork: &mut [i32],
info: &mut i32) {
$xgees(jobvs, sort, None, n, a, lda, sdim, wr, wi, vs, ldvs, work, lwork, bwork, info);
}
#[inline]
fn xgees_work_size(jobvs: u8,
sort: u8,
// select: ???
n: i32,
a: &mut [$N],
lda: i32,
sdim: &mut i32,
wr: &mut [$N],
wi: &mut [$N],
vs: &mut [$N],
ldvs: i32,
bwork: &mut [i32],
info: &mut i32)
-> i32 {
let mut work = [ Zero::zero() ];
let lwork = -1 as i32;
$xgees(jobvs, sort, None, n, a, lda, sdim, wr, wi, vs, ldvs, &mut work, lwork, bwork, info);
ComplexHelper::real_part(work[0]) as i32
}
}
)
);
real_eigensystem_scalar_impl!(f32, interface::sgees);
real_eigensystem_scalar_impl!(f64, interface::dgees);

View File

@ -4,3 +4,4 @@ mod cholesky;
mod lu; mod lu;
mod qr; mod qr;
mod svd; mod svd;
mod real_schur;

View File

@ -0,0 +1,21 @@
use std::cmp;
use nl::RealSchur;
use na::{DMatrix, Matrix4};
quickcheck! {
fn schur(n: usize) -> bool {
let n = cmp::max(1, cmp::min(n, 10));
let m = DMatrix::<f64>::new_random(n, n);
let (vecs, vals) = RealSchur::new(m.clone()).unpack();
relative_eq!(&vecs * vals * vecs.transpose(), m, epsilon = 1.0e-7)
}
fn schur_static(m: Matrix4<f64>) -> bool {
let (vecs, vals) = RealSchur::new(m.clone()).unpack();
relative_eq!(vecs * vals * vecs.transpose(), m, epsilon = 1.0e-7)
}
}