﻿Version 1.1 (2022-09-23)

* Introduced the additional possibility to require the boundary conditions to
  be accurately fulfilled (instead of only approximately via the least-squares
  approach). The proofs should be provided along the lines of the existing
  theory.

* The type int has been replaced by LSCMint which in turn is mapped to the
  integer type used in lapack. This is done in order to simplify the
  portability of the code.

* minor bug fixes.

Version 2.0 (2023-06-09)

* Complete rewriting of the collocation conditions (Yn) by using general
  quadrature rules following some ideas of chebfun (https://www.chebfun.org)

* Introduced a new routine for computing a matrix that can be used for
  defining accurate initial conditions. More precisely, a matrix is provided
  that has the same nullspace as the canonical projector.
  
* Added an initial value solver. It is heavily based on the routine for
  computing accurate initial conditions to be used as transfer conditions.

Version 2.0.1 (2023-11-06)

* Modified for use with AMD AOCL on their CPUs

Version 2.0.2 (2024-05-08)

* Added a new routine for computing a smooth QR decomposition: rank-revealing
  QR-factorization by Christian M. Bischof and Gregorio Quintana-Orti,
  Algorithm 782 of TOMS.

* Bug fixes and minor improvements

Version 2.0.3 (2024-09-10)

* Included a new fast estimator of singular values based on UTV decompositions.
  The algorithm is published here: P.G. Martinsson, G. Quintana-Orti, and N. Heavner:
  ACM Transaction on Math. Software, 45(2019)1, 4:1-4:26.
  
* Bug fixes and minor improvements
