Changelog

All notable changes to PNL will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

[1.15.0] - 2024-01-03

Changed

Remove the full tensor representation of PnlBasis and only use the sparse tensor (#9).

Fixed

Fix unpacking of sparse matrices.

[1.14.0] - 2023-12-20

Added

Apply optional non-linear pre-treatment to the data before renormalisation in PnlBasis (#8).

Changed

Remove internal C translation of Blas & Lapack.
Improve basis memory management (#7).

Fixed

Fix packing/unpacking of PnlBasis objects.

[1.13.1] - 2023-11-04

Fixed

pnl_mat_fprint was printing to stdout instead of the file descriptor.

[1.13.0] - 2023-11-04

Added

Add missing documentation on local bases.
Add tests for basis tensor constructors.

Changed

For the sake of consistency, the constructors pnl_basis_create_local and pnl_basis_create_local_regular are renamed to pnl_basis_local_create and pnl_basis_local_create_regular.
The function pnl_basis_i is not marked inline anymore.
In functions to compute a basis derivative, first check that Df is not NULL.
Prevent standard PnlBasis constructors to be used to create a local basis (type PNL_BASIS_LOCAL).
Refactor PnlBasis tensor constructors to drop the recursive and copy approach.

[1.12.1] - 2023-10-25

Fixed

Add missing definitions for Windows in pnl.def.

[1.12.0] - 2023-10-25

Added

Add local bases (#5)

[1.11.0] - 2023-01-17

Changed

The macros RETRIEVE and CREATE are prefixed with PNL_RAND. This is a breaking change.
Add PNL_ prefix to FALSE, TRUE, OK and FAIL. This is a breaking change.

Added

Add pnl_mat_print_csv to print a matrix in a csv format.

[1.10.4] - 2019-10-21

Fixed

Add missing libwinpthread-1.dll to post_build event

[1.10.3] - 2019-09-16

Fixed

Add missing libwinpthread-1.dll to Windows binaries

[1.10.2] - 2019-09-13

Fixed

Fix libgfortran version for Windows binary release
Fix README typos

[1.10.1] - 2019-09-11

Fixed

Add missing exports to pnl.def

[1.10.0] - 2019-09-11

Fixed

Fix strcasecmp not available on MSVC
Fix PNL_RANGE_CHECK_OFF in Release mode

Added

Add new functions for sparse matrices
Add pnl_vect_set_subblock
Add FFT api with work spaces

[1.9.6] - 2018-06-06

Fixed

Fix memory corruption in LECUYER random number generator.
Set mem_size field to 0 in _free methods.

[1.9.5] - 2018-06-04

Fixed

Fix compilation error with Glibc 2.27 (missing symbols in math.h)

[1.9.4] - 2018-05-22

Fixed

Fix wrong release number

[1.9.3] - 2018-05-18

Fixed

Fix compilation error with Glibc 2.27 (missing symbols in math.h)
Fix initialization of Mersenne Twister random generator with random seed.

[1.9.2] - 2017-12-24

Fixed

Fix access to global variables under Windows. This was causing a segfault in random_test.

Added

Add BUILD_SHARED_LIBS as a CMake option to set the type of library to build. Default value in ON. When set to OFF, PNL is built as a static library.

[1.9.1] - 2017-12-13

Added

Add option PNL_ENABLE_TESTS=OFF to CMake to disable tests. Default value is sON.

Changed

Update the CMakeuser.incl mechanism.
Update some installation path to be more consistent with Linux standards:
- CMakeuser.incl is installed to <prefix>/share/pnl,
- PnlConfig.cmake is installed to <prefix>/lib/cmake/pnl/.

Fixed

Fix minor bugs in pnl_sp_mat_isequal and mtherr.
Fix some compilation warnings in LP_Solve.

[1.9.0] - 2017-11-30

Added

Read sparse matrices from file.
Update read functions to deal with comments.
Add floating point comparison functions for vectors and matrices.

[1.8.0] - 2017-10-04

Added

Add a linear programming routine based on the Simplex implementation of LP_Solve.

Fixed

Fix a few bugs.

[1.7.5] - 2017-01-18

Changed

Improve CMake's integration for Visual Studio.
The binary version pnl-win64-1.7.5 is compatible with Visual Studio up to VS2013 but not with VS2015. If you use VS2015, download the source files and compile them yourself.

Fixed

Fix minor bugs.

[1.7.4] - 2016-03-21

Added

Provide a PnlConfig.cmake file to use CMake find_package command in config mode.
Enable in-place MPI reductions of PNLObjects.

Changed

Improve the efficiency of creating and evaluating PnlBasis objects.

Fixed

Fix compiling errors in the examples under MSVC.

[1.7.3] - 2016-02-15

Added

Add non tensor functions to PnlBasis objects.
Add Complex error functions based on the Faddeeva package.

[1.7.2] - 2015-09-10

Added

A CMake module to include the library in other projects.
The non central chi squared distribution to the random number generation toolbox.

Changed

Improves and fixes a bug in the pnl_basis module.

[1.7.1] - 2014-07-10

Fixed

Minor fixes in the examples
Improvement of the inline facility detection in CMake. This fixes a bug of the 1.7.0 under Windows.

[1.7.0] - 2014-04-28

Added

Random number generation. All methods know how to handle QMC generators. Three new distributions have been added: log-normal, inverse gaussian and asymmetric double exponential distributions.
Linear Algebra. Computations of the logarithm and exponential of complex matrices. Sparse matrix object.
Optimization. Newton's algorithm with Armijo line search.

Changed

Three new distributions are added: log-normal, inverse gaussian and asymmetric double exponential distributions.
The evaluation of a multivariate polynomial is greatly improved thanks to the use of sparse storage of these objects. This new implementation relies on sparse matrices.
The internal structure of the top level object is modified to keep track of the number references on to a given object, which has enabled us to improve by a great deal the memory management of lists and arrays.

[1.6.0] - 2013-03-22

Added

Add a few functions for rounding (some are part of C99).

Changed

Improve Blas & Lapack detection on Mac OS X.

Fixed

Fix a minor bug in two dimensional FFT.

[1.5.2] - 2013-01-23

Changed

The compilation chain has moved from the Autotools to CMake.

[1.5.1] - 2012-10-16

Added

MPI_Reduce binding for PnlVect and PnlMat objects.
2D FFT.
The user can define new function bases and register them for further usage as native bases.
The PnlObject structure has three new function members: clone, copy, new. It is mainly useful for inheritance purposes.

[1.5.0] - 2012-03-19

Added

Linear solver using a QR decomposition with column pivoting
Inverse of permutations and its application in place
Approximations for inverse hyperbolic functions (part of C99 but missing under Visual)

Changed

Most macros are renamed with a pnl_ prefix to avoid name clashes

[1.4.1] - 2011-10-11

Added

New functions to extract data from Hmatrices as vectors or matrices
New PnlArray type: array of PnlObjects. It is often a good alternative to using Hmatrices.

Changed

Redesign of how Hmatrices are stored in memory

Fixed

A bug in pnl_mat_{lower,upper}_syslin has been fixed

[1.4.0] - 2011-09-09

Added

Runge Kutta Fehler 45 integrator for n dimensional ODEs
LU decomposition of tridiagonal matrices
Cholesky factorization with complete pivoting for positive semidefinite matrices
New threadsafe Sobol generators (from John Burkardt)

Changed

Design of a new unit test framework
New organisation of the manual
Update Lapack to version 3.2.1
Integration of the Mersenne Twister Dynamic Creator version 0.6.1

[1.3.3] - 2011-03-10

Added

Generation of Bessel random variables
A generator for parallel computing (MPI) is added : Dynamic created Mersenne Twister
Sparse polynomial bases based on hyperbolic sets of indices. Bases can be centered and normalized.
Save/Load interface based on MPI Pack/Unpack

Fixed

Random number generators become thread-safe.

[1.2.0] - 2010-07-15

Added

Bindings for MPI

Changed

Update of random number generators to work on 64-bit machines