Expand description

Provides pure-Rust implementations of linear algebra routines for ndarray without depending on external LAPACK/BLAS libraries.

Eliminating BLAS dependencies

If this crate is being used as a BLAS-less replacement for ndarray-linalg, make sure to remove ndarray-linalg from the entire dependency tree of your crate. This is because ndarray-linalg, even as a transitive dependency, forces ndarray to be built with the blas feature, which forces all matrix multiplications to rely on a BLAS backend. This leads to linker errors if no BLAS backend is specified.


Compact bidiagonal decomposition for matrices

Cholesky decomposition on symmetric positive definite matrices.

Eigendecomposition for symmetric square matrices

Locally Optimal Block Preconditioned Conjugate Gradient (LOBPCG) is a matrix-free method for finding the large (or smallest) eigenvalues and the corresponding eigenvectors of a symmetric eigenvalue problem

Norm of vectors

QR decomposition of rectangular matrices.

Compact singular-value decomposition of matrices

Traits for creating and manipulating triangular matrices

Tridiagonal decomposition of a symmetric matrix


Find largest or smallest eigenvalues

Type Definitions