Module linear 

Deterministic small linear-algebra kernels.

These routines keep scalar operation order explicit for parity-sensitive GNSS callers. When pivot tie-breaking or accumulation order matters, the variant name states the policy instead of hiding it in a local copy.

Structs

FlatCholeskySolveScratch

Reusable buffers for the owned Cholesky (square-root) solve (solve_flat_normal_square_root_into): the lower-triangular factor L (row-major n x n), the forward-substitution vector z, and the solution x. Held across solves so a steady-state iteration does not allocate.

FlatLinearScratch

FlatNormalSolveScratch

Enums

LinearError

Functions

det4_cofactor

dot4

invert_3x3_adjugate

invert_4x4_cofactor

invert_flat_first_tie_into

invert_matrix_first_tie

invert_matrix_last_tie

invert_symmetric_pd

mat4_vec4

matmul

matrix_sub

minor3_of_4

normal_equations_weighted

normal_matrix_4_unweighted_row_outer

normal_matrix_4_weighted_column_outer

solve_augmented_flat_first_tie_in_place

solve_flat_normal_first_tie

solve_flat_normal_first_tie_into

solve_flat_normal_square_root_into

Solve the symmetric positive-definite information system Λ x = η by an owned deterministic Cholesky (square-root) factorization Λ = L Lᵀ, then forward substitution L z = η and back substitution Lᵀ x = z. lambda is the row-major n x n information matrix, eta the length-n information vector.

solve_linear_first_tie

solve_linear_last_tie

solve_matrix_flat_first_tie_into

solve_matrix_last_tie

transpose