devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

* the Upper->UpperTriangular change

Browse Source 
* finally get ei_add_test right
This commit is contained in:
Benoit Jacob
2008-12-20 13:36:12 +00:00
parent 21ab65e4b3
commit 9e00d94543
33 changed files with 181 additions and 179 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
Eigen/src/QR/HessenbergDecomposition.h
View File

@@ -243,7 +243,7 @@ HessenbergDecomposition<MatrixType>::matrixH(void) const
int n = m_matrix.rows();
MatrixType matH = m_matrix;
if (n>2)
matH.corner(BottomLeft,n-2, n-2).template part<Lower>().setZero();
matH.corner(BottomLeft,n-2, n-2).template part<LowerTriangular>().setZero();
return matH;
}

4
Eigen/src/QR/QR.h
View File

@@ -60,11 +60,11 @@ template<typename MatrixType> class QR
bool isFullRank() const { return ei_isMuchSmallerThan(m_hCoeffs.cwise().abs().minCoeff(), Scalar(1)); }
/** \returns a read-only expression of the matrix R of the actual the QR decomposition */
const Part<NestByValue<MatrixRBlockType>, Upper>
const Part<NestByValue<MatrixRBlockType>, UpperTriangular>
matrixR(void) const
{
int cols = m_qr.cols();
return MatrixRBlockType(m_qr, 0, 0, cols, cols).nestByValue().template part<Upper>();
return MatrixRBlockType(m_qr, 0, 0, cols, cols).nestByValue().template part<UpperTriangular>();
}
MatrixType matrixQ(void) const;

8
Eigen/src/QR/SelfAdjointEigenSolver.h
View File

@@ -254,22 +254,22 @@ compute(const MatrixType& matA, const MatrixType& matB, bool computeEigenvectors
cholB.matrixL().solveTriangularInPlace(matC);
// FIXME since we currently do not support A * inv(L'), let's do (inv(L) A')' :
matC = matC.adjoint().eval();
cholB.matrixL().template marked<Lower>().solveTriangularInPlace(matC);
cholB.matrixL().template marked<LowerTriangular>().solveTriangularInPlace(matC);
matC = matC.adjoint().eval();
// this version works too:
// matC = matC.transpose();
// cholB.matrixL().conjugate().template marked<Lower>().solveTriangularInPlace(matC);
// cholB.matrixL().conjugate().template marked<LowerTriangular>().solveTriangularInPlace(matC);
// matC = matC.transpose();
// FIXME: this should work: (currently it only does for small matrices)
// Transpose<MatrixType> trMatC(matC);
// cholB.matrixL().conjugate().eval().template marked<Lower>().solveTriangularInPlace(trMatC);
// cholB.matrixL().conjugate().eval().template marked<LowerTriangular>().solveTriangularInPlace(trMatC);
compute(matC, computeEigenvectors);
if (computeEigenvectors)
{
// transform back the eigen vectors: evecs = inv(U) * evecs
cholB.matrixL().adjoint().template marked<Upper>().solveTriangularInPlace(m_eivec);
cholB.matrixL().adjoint().template marked<UpperTriangular>().solveTriangularInPlace(m_eivec);
for (int i=0; i<m_eivec.cols(); ++i)
m_eivec.col(i) = m_eivec.col(i).normalized();
}

12
Eigen/src/QR/Tridiagonalization.h
View File

@@ -167,8 +167,8 @@ Tridiagonalization<MatrixType>::matrixT(void) const
matT.corner(TopRight,n-1, n-1).diagonal() = subDiagonal().conjugate();
if (n>2)
{
matT.corner(TopRight,n-2, n-2).template part<Upper>().setZero();
matT.corner(BottomLeft,n-2, n-2).template part<Lower>().setZero();
matT.corner(TopRight,n-2, n-2).template part<UpperTriangular>().setZero();
matT.corner(BottomLeft,n-2, n-2).template part<LowerTriangular>().setZero();
}
return matT;
}
@@ -223,17 +223,17 @@ void Tridiagonalization<MatrixType>::_compute(MatrixType& matA, CoeffVectorType&
/* This is the initial algorithm which minimize operation counts and maximize
* the use of Eigen's expression. Unfortunately, the first matrix-vector product
* using Part<Lower|Selfadjoint> is very very slow */
* using Part<LowerTriangular|Selfadjoint> is very very slow */
#ifdef EIGEN_NEVER_DEFINED
// matrix - vector product
hCoeffs.end(n-i-1) = (matA.corner(BottomRight,n-i-1,n-i-1).template part<Lower|SelfAdjoint>()
hCoeffs.end(n-i-1) = (matA.corner(BottomRight,n-i-1,n-i-1).template part<LowerTriangular|SelfAdjoint>()
* (h * matA.col(i).end(n-i-1))).lazy();
// simple axpy
hCoeffs.end(n-i-1) += (h * Scalar(-0.5) * matA.col(i).end(n-i-1).dot(hCoeffs.end(n-i-1)))
* matA.col(i).end(n-i-1);
// rank-2 update
//Block<MatrixType,Dynamic,1> B(matA,i+1,i,n-i-1,1);
matA.corner(BottomRight,n-i-1,n-i-1).template part<Lower>() -=
matA.corner(BottomRight,n-i-1,n-i-1).template part<LowerTriangular>() -=
(matA.col(i).end(n-i-1) * hCoeffs.end(n-i-1).adjoint()).lazy()
+ (hCoeffs.end(n-i-1) * matA.col(i).end(n-i-1).adjoint()).lazy();
#endif
@@ -256,7 +256,7 @@ void Tridiagonalization<MatrixType>::_compute(MatrixType& matA, CoeffVectorType&
Block<MatrixType,Dynamic,4>(matA,b+4,b,n-b-4,4).adjoint() * Block<MatrixType,Dynamic,1>(matA,b+4,i,n-b-4,1);
// the 4x4 block diagonal:
Block<CoeffVectorType,4,1>(hCoeffs, b, 0, 4,1) +=
(Block<MatrixType,4,4>(matA,b,b,4,4).template part<Lower|SelfAdjoint>()
(Block<MatrixType,4,4>(matA,b,b,4,4).template part<LowerTriangular|SelfAdjoint>()
* (h * Block<MatrixType,4,1>(matA,b,i,4,1))).lazy();
}
#endif