Imports

Procs

proc laswp(a: var Tensor; pivot_indices: openArray[int32]; pivot_from: static int32): Apply A = P * A where P is a permutation matrix, represented pivot_indices of rows.

A is a matrix of shape MxN. A is permuted in-place
Source Edit
proc orgqr[T: SomeFloat](rv_q: var Tensor[T]; tau: openArray[T]; scratchspace: var seq[T]): Wrapper for LAPACK orgqr routine Generates the orthonormal Q matrix from elementary Householder reflectors

Inputs must come from a previous geqrf

rv_q: contains r_v (reflector vector) on input. A column-major vector factors of elementary reflectors

tau: Scalar factors of elementary reflectors

Outputs

rv_q: overwritten by Q

Note that while rv_q is MxN on input on output the shape is M x min(M,N)

⚠️: Output must be sliced by M, min(M,N) if M>N as the rest contains garbage

Spec: https://www.nag.co.uk/numeric/fl/nagdoc_fl24/pdf/f08/f08aff.pdf API: http://www.netlib.org/lapack/explore-html/da/dba/group__double_o_t_h_e_rcomputational_ga14b45f7374dc8654073aa06879c1c459.html
Source Edit
proc ormqr[T: SomeFloat](C: var Tensor[T]; Q: Tensor[T]; tau: openArray[T]; side, trans: static char; scratchspace: var seq[T]): Wrapper for LAPACK ormqr routine Multiply the orthonormal Q matrix from geqrf with another matrix C without materializing Q

C is a matrix of shae M, N and will be overwritten by

SIDE = 'L' SIDE = 'R'

TRANS = 'N': Q * C C * Q TRANS = 'T': QT * C C * QT
Source Edit