Skip to content

Latest commit

 

History

History
444 lines (358 loc) · 21.1 KB

CHANGELOG.md

File metadata and controls

444 lines (358 loc) · 21.1 KB
Raw

Change Log

Version 1.1.12 - 2018-01-20

Added

  • The function pcmsolver_fstring_to_carray was added to the Fortran bindings for the library. As the name suggests, this function "translates" a Fortran string into a C char array. This supersedes and replaces the pcmsolver_f2c_string function.
  • An updated version of the pcmsolver_new API function has been introduced. The full name of the function is pcmsolver_new_v1112 (to avoid API breakage) and accepts as additional argument the name of the parsed input file generated by pcmsolver.py. This allows to decoupled parsing and writing out the parsed file, in case directory-specific operations have to be performed by the host program.

Changed

  • Dispatching of Green’s functions in the factory is purely label-based. The input reader generates a string that holds the type of the Green’s function, the strategy for calculating its normal derivatives and the dielectric profile (if needed): TYPE_DERIVATIVE_PROFILE. The Green’s functions are subscribed to the factory using this label, which is then also the one used for their creation. Iteration over lists of types to dispatch the correct template parameters is thus completely avoided. This removes the dependency on the metafunctions in ForId.hpp and the dependency on Boost.MPL, the Boost metaprogramming library.
  • Dispatching of quadrature rules for the numerical boundary integral operator has been rewritten, removing the dependency on Boost.MPL.
  • The input wrapping data struct-s have been modified to include the string defining the type of object and to remove now unused information.
  • Documentation building is fully handled via sphinx-build: CMake will no longer generate a doc build target.
  • Simplified .travis.yml and got rid of Conda to handle multiple Python versions.
  • Moved python scripts from bin/ to bin/pcmsolver/ and introduced bin/pcmsolver/__init__.py.
  • The parse_pcm_input function returns the parsed input as a string without saving it to file. The old behavior can be recovered by setting the write_out argument to True.
  • The Catch unit test framework has been updated to its latest version still supporting C++03 v1.11.0
  • Travis continuous integration is no longer run on Mac infrastructure.

Fixed

  • Documentation building on ReadTheDocs is fully functional again, thanks @arnfinn 🎉 The build had been failing for a while since docs were generated for all files, including documentation files from previous build. Besides, source code doxygen blocks were not extracted when inside namespaces.
  • Compiler warning in the Sphere class due to a redeclaration of operator<<.
  • The plot_cavity.py script is now Python 3 compatible.

Removed

  • FortranCUtils header and source files were removed. The pcmsolver_f2c_string and pcmsolver_c2f_string functions are thus gone. For the former, use the replacement pcmsolver_fstring_to_carray function provided in the Fortran bindings to the library.
  • Some unused files have been removed:
    • Interpolation.hpp
    • Interpolation.cpp
    • Vector2.hpp
    • Vector3.hpp
    • ForId.hpp
    • DerivativeUtils.hpp
  • The cavityType(), solverType(), greenInsideType(), greenOutsideType() and integratorType() functions in the Input object have been removed. This information is now wrapped into the corresponding input-wrapping struct-s: CavityData, SolverData, GreenData and BIOperatorData.

Deprecated

  • We are in the process of removing the dependency on the Boost libraries. The only dependency on Boost allowed in the 1.Y.Z release series will be for replacing C++11 functionality that is missing from the C++98 standard. For such code, we will require a pure C++11 counterpart.

Version 1.1.11 - 2017-10-25

Added

  • A Python script, using Matplotlib, to plot the cavity. The script can also color-map the finite elements according to the values of a surface function.
  • The input learnt to parse the additional ChargeDistribution section. It is possible to specify a classical charge distribution of point multipoles. This can be an additional source of electrostatic potential for the calculation of the ASC.
  • Restored compilation for g++ < v5.1.
  • Ninja can be used as a generator. Notice that at least CMake 3.7.2 and the Kitware-maintained version of Ninja are required to successfully compile.

Changed

  • Use #pragma once instead of #ifndef, #define, #endif to guard against multiple inclusion of header files.
  • The uppercased contents of the .pcm input file are written to a temporary file, instead of overwriting the user provided file. The temporary file is removed after it has been parsed. Fixes #91 as noted by @ilfreddy.

Fixed

  • A bug in the initialization of a restart cavity from old .npz files. Currently, the .npz file saves sphere center, arcs and vertices of each finite element. This information is in fact needed to plot the cavity using the Python script in tools. Older .npz files did not contain this information and could not be read in. The additional information is read in as arrays of zeros in case it is not present on file.

Version 1.1.10 - 2017-03-27

Changed

  • Updated the cloc.pl script to version 1.72
  • Simplified the internal structure of the Meddle and Input objects.
  • Export dependency on Zlib for the static libraries. Thanks @loriab for the pull request fixing a build problem within Psi4

Version 1.1.9 - 2017-02-16

Changed

  • PCMSolver is now exported as a proper CMake target See PR #38 for details. Thanks @loriab for the work.
  • The Python scripts shipped with the library are now Python 2 and Python 3 compatible.
  • Factory is no longer implemented as a Singleton.
  • The Catch unit test framework has been updated to its latest version v1.7.2
  • Updated the version of Eigen bundled with the code. The minimum required version of Eigen is still 3.3.0, but we ship Eigen 3.3.2

Fixed

Version 1.1.8 - 2017-02-06

Added

  • Namespaces for all of the internal code have been introduced. The top-level namespace is pcm. At finer levels the namespaces have the same names as the respective subdirectories. Read the programmers' documentation for further details on the use of namespaces in the project. Related to issue #34 on GitHub and #60 on GitLab.
  • The top-level, convenience header BoundaryIntegralOperator.hpp includes all subclasses and utility headers in the bi_operators subdirectory. Related to issue #34 on GitHub and #60 on GitLab.
  • The top-level, convenience header Cavity.hpp includes all subclasses and utility headers in the cavity subdirectory. Related to issue #34 on GitHub and #60 on GitLab.
  • The top-level, convenience header Green.hpp includes all subclasses and utility headers in the green subdirectory. Related to issue #34 on GitHub and #60 on GitLab.
  • The top-level, convenience header Solver.hpp includes all subclasses and utility headers in the solver subdirectory. Related to issue #34 on GitHub and #60 on GitLab.

Changed

  • The abstract base class for the boundary integral operator integrators has been renamed IBoundaryIntegralOperator. The relevant factory is bootstrapped upon creation of the Meddle object, i.e. at library initialization. Related to issue #34 on GitHub and #60 on GitLab.
  • The abstract base class for the cavities has been renamed ICavity. The relevant factory is bootstrapped upon creation of the Meddle object, i.e. at library initialization. Related to issue #34 on GitHub and #60 on GitLab.
  • The abstract base class for the solvers has been renamed ISolver. The relevant factory is bootstrapped upon creation of the Meddle object, i.e. at library initialization. Related to issue #34 on GitHub and #60 on GitLab.
  • The typedef for numerical differentiation in the Green's function classes has been renamed Stencil to avoid name clashes with the Numerical boundary integral operator type.

Fixed

  • A bug in the selection of the extended diagnostics flags for the GNU C++ compiler. These flags are now enabled only for versions >= 5.1.0 and when the C++11 standard is enable. Fixes issue #36 on GitHub and #62 on GitLab.
  • A bug in the initialization of the factory for the cavity classes was fixed. The bug manifested only in the static library libpcm.a Fixes issue #34 on GitHub and #60 on GitLab.

Version 1.1.7 - 2016-12-01

Added

  • A pre-commit hook in .githooks/pre-commit-clang-format checking that the format of C++ header and source files conforms to the project style. The hook uses clang-format and the style mandated by the .clang-format file to check files in the tree. Commit is rejected if one or more files are non compliant. The hook generates a patch and shows the command needed to apply it. To enable the hooks you need to have a .git/hooks/pre-commit file containing this line .githooks/pre-commit NOT recommended The hook can be skipped by passing the --no-verify option to git commit
  • A pre-commit hook in .githooks/pre-commit-license-maintainer checking the license headers. The hook is configured based on the .gitattributes file. The hook will check the license headers and amend them, either by updating the year and authors information or by adding the whole header. To enable the hooks you need to have a .git/hooks/pre-commit file containing this line .githooks/pre-commit NOT recommended The hook can be skipped by passing the --no-verify option to git commit
  • An UNUSED preprocessor macro to mark arguments as unused.
  • An UNUSED_FUNCTION preprocessor macro to mark functions as unused.
  • A set of preprocessor macros with Git information (GIT_COMMIT_HASH, GIT_COMMIT_AUTHOR, GIT_COMMIT_DATE and GIT_BRANCH) is automatically generated and saved in the git_info.h header file.
  • An API function to print the contents of a surface function to the host program output.
  • An API function to get the dipole moment, relative to the origin, due to the ASC on the cavity. Both the norm and the components can be obtained.
  • .gitattributes now instructs Git to ignore binary files in diff operations. PNG files are diff-ed using EXIF information. To set this up properly, install an EXIF tool on your machine and run git config diff.exif.textconv exiftool in your local copy of the repository.

Changed

  • The Fortran bindings file has been renamed pcmsolver.f90.
  • The Green's function, solver and boundary integral operator classes have been radically redesigned. This avoids coupling between integrators and Green's function that existed in the previous design. See the Green's function code reference for a more detailed explanation.
  • BREAKING CHANGE The minimum required version of Eigen is now 3.3.0 The version bundled with the code has been accordingly updated.
  • The PCMSOLVER_ERROR macro now takes only one argument and prints out a more informative error message.
  • Switched to the latest version of Autocmake The configuration file is now YAML-based. The PyYAML module is thus required.
  • The extended diagnostic flags -Wsuggest-attribute=pure -Wsuggest-attribute=const -Wsuggest-attribute=noreturn -Wsuggest-final-types -Wsuggest-final-methods -Wsuggest-override -Wuseless-cast -Wunsafe-loop-optimizations are always set when using the GNU C++ compiler in a debug configuration.
  • The C++11 compatibility CMake macros now check for the availability of the noreturn attribute. A workaround macro, accessible via __noreturn, has been added to the Cxx11Workarounds.hpp header file.
  • BREAKING CHANGE The ouput flushing function must be passed explicitly as a function pointer to the pcmsolver_new function during library initialization. The function pointer has the signature typedef void (*HostWriter)(const char * message) thus accepting a single argument instead of the previous two.
  • GNU standard installation directories have been imposed, thanks to work by @loriab. Given a prefix, header files are now installed to include/pcmsolver, executables to bin, libraries to lib and scripting tools to share. The install prefix and the installation directories can be specified by the --prefix, --bindir, --libdir, --includedir and --datadir options to the setup.py script (or the corresponding CMake variables)

Version 1.1.6 - 2016-09-20

Added

  • A function returning a molecule object for the water molecule.

Changed

  • Cholesky decomposition is used whenever the inverse of the S matrix has to be calculated. The S matrix is self-adjoint, positive-definite and the LLT decomposition is faster than LDLT.

Fixed

  • Some inconsistencies in input reading from host and a related memory leak in the radii initialization.

Version 1.1.5 - 2016-07-19

Added

  • A radii set derived from Allinger's MM3 model can now be chosen to build the van der Waals cavity surface. Notice that the values reported in the original paper are divided by 1.2, to match the default radii set used in ADF The closest match to ADF can be obtained by using CPCM as solver, Allinger's radii and setting the scaling of radii to false.

Version 1.1.4 - 2016-07-05

Changed

  • The CPCMSolver object now stores the scaled, Hermitian, symmetry-adapted S matrix. Polarization weights are then directly computed from the incoming MEP.
  • The IEFSolver object now stores the non-Hermitian, symmetry-adapted T and R matrices. The explicit calculation of the inverse of T is thus avoided. However, two square matrices of size equal to the cavity size are stored instead of just one. To obtain the polarization weights two linear systems of equations are solved. A partially pivoted LU decomposition is used to solve the linear system(s). The strategy used in v1.1.3 suffered from a reduced numerical accuracy, due to the fact that the polarization weights were not correctly defined.

Removed

  • The hermitivitize function will only work correctly on matrices. This reverts modifications in the previous release.

Version 1.1.3 - 2016-07-03

Changed

  • The PEDRA.OUT cavity generator log now reports the initial and final lists of spheres. The final list only contains those spheres that were actually tesselated and, possibly, the added spheres.
  • For all solvers, the symmetrization needed to obtain the polarization weights happens directly on the computed charges, instead of symmetrizing the system matrices.
  • The IEFSolver object stores the unsymmetrized T^-1R matrices. A partially pivoted LU decomposition is used to compute T^-1R. A robust Cholesky decomposition is used to form the R matrix in the anisotropic IEF case.
  • The CPCMSolver object now stores the scaled, unsymmetrized S matrix. The explicit calculation and storage of its inverse is thus avoided. A robust Cholesky decomposition is used to solve the linear equation system.
  • The hermitivitize function can now correctly symmetrize vectors.

Fixed

  • A fix for the initialization of the explicit list of spheres when running the standalone executable. The bug prevented the generation of the correct Molecule object, with subsequent failure in the cavity generator.
  • A memory leak occuring in the cavity generator PEDRA was fixed. This was uncovered by @shoefener and manifested only with considerably large cavities (> 200 input spheres)

Removed

  • The function CPCMMatrix in the SolverImpl.hpp header file is no longer available.

Version 1.1.2 - 2016-05-31

Fixed

  • Signatures for strings in Fortran90 bindings. They have now the proper C interoperable type character(kind=c_char, len=1) :: label(lbl_size). For the host this means that surface function labels will have to be declared as character arrays, for example: character :: label(7) = (/'T', 'o', 't', 'M', 'E', 'P', char(0)/)

Changed

  • More informative error messages for runtime crashes caused by access to surface functions.
  • The signatures for the interface functions now accept and/or return int (c_int) instead of size_t (c_size_t). This simplifies interfacing with Fortran hosts.

Version 1.1.1 - 2016-03-10

Added

  • A runtime check to ensure that all atoms have a nonzero radius. API kills program execution if this is the case.
  • An API function to retrieve the areas/weights of the cavity finite elements. The values in the returned array are in Bohr^2. Addresses a feature request from @shoefener (Issue #13)
  • The standalone executable run_pcm is now tested within the unit tests suite. The tests cover the cases where the cavity is given implicitly, explicitly or by substitution of radii on chosen atoms.

Changed

  • Boundary integral operators classes learnt to accept a scaling factor for the diagonal elements of the approximate collocation matrices. The change is reflected in the Green's funtion classes and in the input parsing. Addresses a feature request from @shoefener (Issue #16)
  • GePolCavity learnt to print also the list of spheres used to generate the cavity.
  • Different internal handling of conversion factors from Bohr to Angstrom.
  • CMake minimum required version is 2.8.10
  • Atom, Solvent and Sphere are now PODs. The radii and solvent lists are free functions.
  • PCMSOLVER_ERROR kills program execution when an error arises but does not use C++ exceptions.
  • include-s are now specified on a per-directory basis (see programmers' manual for a more detailed explanation)
  • Default types for template paramters DerivativeTraits, IntegratorPolicy and ProfilePolicy are now given for the Green's functions classes. This reduced the verbosity in instatiating these objects significantly.

Known Issues

  • The new printer in GePolCavity might not work properly when an explicit list of spheres is provided in the input.
  • On Ubuntu 12.10, 32 bit the Intel compiler version 2013.1 produces a faulty library. It is possibly a bug in the implementation of iso_c_binding, see Issue #25

Removed

  • SurfaceFunction as a class is no longer available. We keep track of surface functions at the interface level via a label-vector map.

Version 1.1.0 - 2016-02-07

Added

  • Green's function for diffuse interfaces in spherical symmetry

Changed

  • CMake minimum required version is 2.8.8 (2016-01-08)
  • Documentation is now served here

v1.0.4 - 2015-07-22 [YANKED]

v1.0.3 - 2015-03-29 [YANKED]

v1.0.2 - 2015-03-28 [YANKED]

v1.0.1 - 2015-01-06 [YANKED]

v1.0.0 - 2014-09-30 [YANKED]