block_householder_tridiag_cl()

void stan::math::internal::block_householder_tridiag_cl ( const matrix_cl< double > &  A, matrix_cl< double > &  packed, const int  r = 60  )

inline

Tridiagonalize a symmetric matrix using block Housholder algorithm.

A = Q * T

• Q^T, where T is tridiagonal and Q is orthonormal.

  Parameters

  	A	Input matrix
  [out]	packed	Packed form of the tridiagonal matrix. Elements of the resulting symmetric tridiagonal matrix T are in the diagonal and first superdiagonal. Columns bellow diagonal contain householder vectors that can be used to construct orthogonal matrix Q.
  	r	Block size. Affects only performance of the algorithm. Optimal value depends on the size of A and cache of the processor. For larger matrices or larger cache sizes a larger value is optimal.

Definition at line 29 of file tridiagonalization.hpp.