Skip to content

Releases: lammps/lammps

Feature release 19 November 2024

19 Nov 19:17
patch_19Nov2024
e6ed911
Compare
Choose a tag to compare
Pre-release

Changes since the 29 August 2024 release:

  • Add support for running FFTs with NVPL from Nvidia for Grace CPUs for plain FFTs and KOKKOS (Richard Berger, LANL) PR #4294
  • Update support for using MKL FFTs for CPUs and add support for Intel GPUs when compiling LAMMPS with KOKKOS for SYCL (Chris Knight, ANL) PR #4313
  • Implement kspace_modify collective yes for KOKKOS (Nick Hagerty, ORNL with help from Stan Moore, SNL) PR #4143
  • Refactor support for ReaxFF and QEq in KOKKOS to integer avoid overflows for very large systems (Evan Weinberg, Nvidia and Stan Moore, SNL) PR #4207, PR #4318
  • Bugfixes, updates and ports for GRANULAR and related styles in KOKKOS (Chris Knight, ANL and Suman Chakraborty, Purdue) PR #4346
  • Ports of several fixes to KOKKOS (Mitch Murphy, alphataubio) PR #4296
  • Update of the bundled Kokkos library to version 4.4.1 (Stan Moore, SNL and the Kokkos developers) PR #4322
  • Update of the bundled libfmt library to version 11.0.2 (Axel Kohlmeyer and the libfmt developers) PR #4345
  • Add lammps_extract_atom_size() to the library interface and all modules depending on it. This allows to accurately size arrays in returned in NumPy format and in the Fortran module. (Axel Kohlmeyer, Temple U) PR #4304
  • Add support to region plane for having the reference point read from variables (Evangelos Voyiatsis, NovaMechanics Ltd.) PR #4299
  • New coulomb pair style and QEq fix implementing CTIP as an alternative to could/streits (Gabriel Plummer ,NASA Ames Research Center) PR #4181
  • New fix qtpie/reaxff that implements the QTPIE charge equilibration algorithm for ReaxFF (Navraj S Lalli, Imperial College London and Efstratios M Kritikos, CalTech) PR #4312
  • Add extract() support to many force styles (Evangelos Voyiatsis, NovaMechanics Ltd.) PR #4333, PR #4356
  • Small updates to the BPM package (Joel Clemmer, SNL) PR # 4359
  • New GitHub workflows testing for variable-length arrays, C++23 compatibility, and Linux unit tests with BIGBIG. Major speedup for Windows runner by enabling Ccache support. (Axel Kohlmeyer, Temple U) PR #4304
  • Add GitHub workflows for "quick", "kokkos", and "full" regression tests using the scripts in the tools folder (Trung Nguyen, U Chicago and Axel Kohlmeyer, Temple U) PR #4308, PR #4352, and PR #4362
  • Expanded KOKKOS support for force-style unit tests (Mitch Murphy, alphataubio) PR #4190
  • Many small bug fixes, minor code improvements and refactoring, and updates for coding style, build system, documentation, and formatting (multiple authors) PR #4299, PR #4314, PR #4319, PR #4321, PR #4324, PR #4326, PR #4330, PR #4331, PR #4339, PR #4343, PR #4348, PR #4352, PR #4353, PR #4354, PR #4357, PR #4358, PR #4370, PR #4371, PR #4374, PR #4375

Backward compatibility notes:

  • The bugfix change to the MS-MEAM in the meam pair style in PR #4331 changes energies and forces for systems using L12 lattice settings (like the example included in LAMMPS, or the force style unit test). The reference data for these has been updated.

Update 1 for Stable release 29 August 2024

01 Oct 12:37
stable_29Aug2024_update1
e410a28
Compare
Choose a tag to compare

This is an update to the 29 August 2024 stable release that fixes bugs that have been reported since then and updates some libraries, dependencies, tools, and build system support.

The following individual changes are included:

  • fix syntax error when updating compiler flags for the CrayClang compiler
  • make atom style oxdna compatible with comm_modify vel yes
  • add missing extvector setting to fix saed/vtk
  • correct handling of error checks with wild card arguments in several fix commands
  • better handling of systems with a large number of atoms in KOKKOS
  • deallocate views of views in serial in KOKKOS
  • avoid MPI stalling due to collective operations when handling errors in several MC fixes
  • fix bugs with allocating and initializing restricted neighbor list in ML-IAP
  • fix cut-n-paste bug in fix adapt for angles
  • fix memory allocation issue in pair style quip
  • avoid uninitialized data access in pair style reaxff when using tabulation
  • avoid uninitialized data access for per-type masses
  • correct logic for custom per-atom arrays and return correct constant when looking for data type
  • must set up a label map for atom types to read xyz files with strings as atom types
  • fix a few bugs in fix pour
  • fix a bug in fix wall/gran causing segfaults
  • fix a bug in MS-MEAM for L12 lattices

Backward compatibility notes:

  • The change to the MS-MEAM in the meam pair style changes energies and forces for systems using L12 lattice settings (like the example included in LAMMPS, or the force style unit test). The reference data for these has been updated.

Stable release 29 August 2024

30 Aug 01:19
stable_29Aug2024
570c9d1
Compare
Choose a tag to compare

Below is a list of major changes since the last stable release 2 August 2023

This stable release adds about 3800 commits since the previous stable release. This includes many contributions from beyond the core LAMMPS developers, but also continues the internal refactoring of the code base to make it more consistent and reliable yet also take more advantage of C++ features where they help to simplify the code and easier to modify, maintain, and contribute to.
Since the last stable release the LAMMPS core team has grown, so there are now more eyes and hands taking care of LAMMPS.

General Changes

  • Continued refactoring of the core LAMMPS code and many packages to increase code reuse, simplify new additions, improve consistency, and benefit from C++11 features. Several convenience functions and classes were added in the process.
  • Continued improving of error message to be more specific and provide more hints about what went wrong.
  • Enhancements and extensions of the Programmer' Guide section of the manual.
  • Significant enhancements to LAMMPS-GUI and support for packaging of LAMMPS-GUI and LAMMPS on a variety of platforms.
  • Expanded CTest based internal testing.
  • CMake build requires at least CMake version 3.16.
  • The -DLAMMPS_EXCEPTIONS setting is now always active.
  • Compiling the KOKKOS package requires a C++17 compatible compiler.

Updates and new commands or styles or packages (for more details see the individual release announcements):

  • New RHEO package reproducing hydrodynamics and elastic objects (i.e. an improved version of SPH that can work in combination with BPM)
  • New ML-UF3 package with pair style uf3 implementing the ultra-fast forcefield (uf3) machine learned interatomic potential. A KOKKOS version is included as well.
  • New fix wall/flow command which provides flow boundary conditions
  • New fix deform/pressure command for expanded deformation controls
  • New angle and dihedral styles cosine/squared/restricted for use with the MARTINI force field
  • New pair style pedone for the non-Coulomb part of the Pedone (PMMCS) potential
  • New pair style hybrid/molecular which allows different pair styles (or pair_coeffs) for inter- and intra-molecular interactions
  • New compute pace command for evaluating ACE descriptors which functions similar to compute snap
  • New compute to identify under-coordinated particles (rattlers)
  • New fix to measure the nonaffine displacement of particles
  • New compute slcsa/atom for supervised learning structural analysis
  • New per-atom compute to calculate local atom type compositions
  • New fix press/langevin command for a langevin based barostat
  • New pair style nb3b/screened which computes a screened angle interactions compared to nb3b/harmonic
  • Ports of several pair, bond, angle, dihedral, improper, and fix styles to KOKKOS
  • Bundled Kokkos library is updated for version 4.3.1
  • Enhancements and bug fixes for the SPH package
  • Updates to the COLVARS package
  • Updates to the ML-POD package
  • Updates for the ELECTRODE package
  • Enhancements for running LAMMPS with very large atomic or molecular systems across a very large number of processors
  • Significantly expanded support for using typelabels
  • Add functionality to build the PLUMED as a plugin so it can be provided in binary distributions, e.g. on Windows
  • Add support for using the heFFTe library to do FFTs
  • Updates for multiple INTERLAYER potential files and bugfixes for a couple of pair styles in the package
  • Add Born matrix support to more potentials
  • Inclusion of angle and dihedral contribution to compuite stress/mop and stress/mop/profile
  • Enhancements to fix pimd/langevin
  • Add option to do anti-aliasing for better image quality with dump image and dump movie
  • Update bundled libfmt code to version 10.1.1

Removed features

  • The lammps-shell is no longer included. Use LAMMPS-GUI in its place.
  • The (outdated) i-PI distribution was removed from tools folder. It can be downloaded using pip instead.
  • The "reax/c" alias to provide backward compatibility for "reaxff" commands and keywords has been (finally) removed. Now using "reaxff" is required.
  • The obsolete MSCG package was removed
  • The unmaintained and unreliable MPIIO package was removed

Stable release candidate August 29 2024

30 Aug 00:52
patch_29Aug2024
05e836f
Compare
Choose a tag to compare
Pre-release

This release primarily contains bug fixes for the 27 June 2023 feature release to make this a stable release. But there are a few new features as listed below. Changes since the 27 June 2024 release:

  • New RHEO package reproducing hydrodynamics and elastic objects (i.e. an improved version of SPH that can work in combination with BPM) (Joel Clemmer, SNL, Thomas C. O’Connor and Eric Palermo, CMU) PR #4214
  • Port angle, dihedral, and improper hybrid styles to KOKKOS (Meg McCarthy and Stan Moore, SNL) PR #4211
  • Port angle style spica and pair style lj/spica/coul/long to KOKKOS (Mitch Murphy) PR #4183
  • New regression test tool to eventually supersede the scripts in the lammps-testing repository (Trung Nguyen, U Chicaco) PR #4228
  • Enhancements and bug fixes for the SPH package (Joel Clemmer, SNL) PR #4243
  • Updates to the COLVARS package (Giacomo Fiorin and the Colvars Developers) PR #4261
  • Add support to build the PLUMED package as a plugin (Axel Kohlmeyer, Temple U) PR #4209
  • Remove LAMMPS shell tool which has been superseded by LAMMPS-GUI (Axel Kohlmeyer, Temple U) PR #4238
  • Various enhancements and bugfixes to LAMMPS-GUI, new lammps_extract_pair() library interface function (Axel Kohlmeyer, Temple U) PR #4235
  • More support for using typelabels in popular LAMMPS features (Jake Gissinger, Stevens Institute of Technology and Axel Kohlmeyer, Temple U) PR #4240, PR #4259, PR #4262
  • Many small bug fixes, minor code improvements and refactoring, and updates for coding style, build system, documentation, and formatting (multiple authors) PR #4212, PR #4215, PR #4218, PR #4224, PR #4227, PR #4230, PR #4232, PR #4236, PR #4237, PR #4239, PR #4242, PR #4247, PR #4248, PR #4250, PR #4252, PR #4255, PR #4256, PR #4263, PR #4270, PR #4273, PR #4275, PR #4276, PR #4278, PR #4279, PR #4283, PR #4285, PR #4287, PR #4289, PR #4290, PR #4291, PR #4292, PR #4293, PR #4295, PR #4298, PR #4301

Backward compatibility notes:

  • the LAMMPS shell is no longer available. Use LAMMPS-GUI instead
  • old fix colvars state files may not be readable by the current version of COLVARS due to changes internal to COLVARS

Update 4 for Stable release 2 August 2023

28 Aug 21:28
stable_2Aug2023_update4
0cb7242
Compare
Choose a tag to compare

This is the fourth and final update to the 2 August 2023 stable release that fixes bugs that have been reported since then and updates some libraries, dependencies, tools, and build system support.

The following individual changes are included:

  • fix a thread initialization issue in LAMMPS-GUI
  • add workaround for fix not computed at compatible time with write_dump for timestep counts > 2 billion
  • fix off-by-one bug in f2c string conversion in Fortran interface module
  • fix bug in electron radius update for fix eff Nose-Hoover thermostat variants
  • fix memory leaks in gather*concat() functions of the library interface
  • avoid segfaults when changing timestep before run but after fix definition for Nose-Hoover fixes and fix shake/rattle
  • update XDR support in dump xtc with code version that is GPL compatible
  • Simplify detection and output of Windows versions
  • fix bug in GPU/CPU overlap in KOKKOS
  • Must always return nfaces == 0 when there are fewer than 3 vertices for rounded polyhedrons in 3d
  • fix atom map bug in fix pour and fix deposit
  • fix bugs in argument parsing for fix ave/correlate and fix ave/correlate/long
  • fix multiple bugs in fix ipi
  • fix bug in converting strings to 64-bit integers on Windows
  • fix logic bug in verbose output for NEB
  • fix bug when requesting occasional neighbor lists where the list would not always be rebuilt properly
  • Do not reference cuFFT when compiling the GPU package with CUDA
  • fix Kokkos shrinkwrap bug when having groups without atoms
  • fix logic bug in compute spin
  • fix hbondchk bug in ReaxFF/OpenMP
  • fix bug in the SPH package where velocities are only partially remapped by a deformation fix (see PR #4243 for details)
  • fix bug when reading incomplete label maps from restart files (see PR #4252 for details)
  • allow using compute spin for groups other than "all"
  • make vector style variables compatible with chunks
  • symmetrize cutoff for pair style ylz
  • make "mass" property in PyLammps class behave similar to "rmass"
  • force using sphinx 7.2 and older so we remain compatible with breathe
  • add vectorstyle variable check to fix ave/histo
  • do not output data from fix reaxff/bonds during setup() except for the very first time
  • avoid infinite loop with cg minimizer and small forces
  • support that cmdargs argument is not modified when creating LAMMPS object and thus can be reused
  • fix logic bug with newton bond off for compute stress/cartesian
  • may install runtime dll dependencies for LAMMPS library only when building with -DBUILD_SHARED_LIBS=on
  • remove spurious dependency of GPU package on cuFFT (which is currently not used)
  • fix compilation issue with recent versions of libAtoms/QUIP
  • lock GitHub actions on macOS to macOS 13 as macOS 14 images are incompatible and fail.
  • set up and install python packages into a virtual environment for GitHub macOS runner to avoid issues with modifying system files
  • add compatibility with plumed 2.9.x
  • make LAMMPS python module compatible with mpi4py version 4.x.x
  • avoid issues with write_dump re-using and existing dump ID when a previous command crashed
  • improve error handling when parsing the MEAM user parameter file: catch value tokenizer exceptions record line numbers and report filename and line number with the error
  • increase MAXELT in pair style meam to 8 (same as upstream)

Feature release 27 June 2024

27 Jun 04:00
patch_27Jun2024
a74500f
Compare
Choose a tag to compare
Pre-release

Changes since the 17 April 2024 release:

  • Add generalized algorithm for replicate command to correctly replicate molecules across periodic boundaries box (Jacob Gissinger, Stevens Institute of Technology) PR #3118
  • Add overlap keyword to create_atoms command which searches for overlaps on an atom-by-atom basis when inserting molecules (Jacob Gissinger, Stevens Institute of Technology) PR #3810
  • Add coefficient of restitution based damping in granular models (Dhairya Vyas, Northwestern University) PR #4068
  • Make compute stress/mop and stress/mop/profile compatible with 2D systems (Evangelos Voyiatzis, Novamechanics Ltd) #4159
  • Update Kokkos library version in LAMMPS to v4.3.1 (Stan Moore, SNL, and Kokkos developers) PR #4162
  • Add special variable functions sort() and rsort() for sorting vectors by value (Axel Kohlmeyer, Temple U) PR #4171
  • Add support for extracting a few more properties and the Atom::map() function to the library interface (Axel Kohlmeyer, Temple U) PR #4174
  • Add KOKKOS package support for hybrid bonds (Stan Moore, SNL) PR #4167
  • Updates to ML-POD package and port of pair style pod to KOKKOS (Ngoc Cuong Nguyen, MIT) PR #4168
  • New option for the CG-DNA package to read potential parameters from a potential file and support for "real" units in addition to "lj" (Kierran Falloon and Oliver Henrich, University of Strathclyde) PR #4169
  • New pair style dpd/coul/slater/long including GPU package version (Eddy Barraud, IFPEN/Sorbonne) PR #4188
  • New ML-UF3 package with pair style uf3 implementing the ultra-fast forcefield (uf3) machine learned interatomic potential. A KOKKOS version is included as well. Ajinkya C Hire (University of Florida), Hendrik Kraß (University of Constance), Matthias Rupp (Luxembourg Institute of Science and Technology) PR #4110
    Richard Hennig (University of Florida)
  • New pair style hybrid/molecular which allows different pair styles (or pair_coeffs) for inter- and intra-molecular interactions (Axel Kohlmeyer, Temple U) PR #4191
  • Add KOKKOS package version of pair_style soft (Stan Moore, SNL) PR #4184
  • Add support for type labels to a whole lot of commands and styles (Jake Gissinger, Stevens Institute of Technology) PR #4145
  • Many small bug fixes, minor code improvements and refactoring, and updates for coding style, build system, documentation, and formatting (multiple authors) PR #3824, PR #4050, PR #4142, PR #4144, PR #4146, PR #4148, PR #4150, PR #4152, PR #4157, PR #4161, PR #4164, PR #4166, PR #4172, PR #4176, PR #4179, PR #4185, PR #4189, PR #4192, PR #4193, PR #4195, PR #4197, PR #4200, PR #4201, PR #4202, PR #4203, PR #4204

Backward compatibility notes:

  • When using fix qeq/reaxff and fix qeq/shielded, more thorough checks are applied that QEq is not used on atoms in the fix group without parameters being provided. This will primarily apply to simulations using ReaxFF as sub-style of pair style hybrid.
  • The i-PI distribution was removed from tools folder. It can be downloaded using pip instead

Feature release 17 April 2024

17 Apr 20:50
patch_17Apr2024
6836bca
Compare
Choose a tag to compare
Pre-release

Changes since the 7 February 2024 release:

  • New rebomos pair style for MoS2 which was provided as external source for over a decade (multiple authors see PR) PR #4087
  • New fix wall/flow command which provides flow boundary conditions (Vladislav Galigerov, Higher School of Economics and Daniil Pavlov, The Moscow Institute of Physics and Technology) PR #4069
  • New fix deform/pressure command for expanded deformation controls specifically aimed at mesoscopic and discrete element models (Joel Clemmer, SNL and Kevin Hanley, U. Edinburgh and Jose Salomon, Imperial College) PR #4017
  • New angle and dihedral styles cosine/squared/restricted for use with the MARTINI force field (Evangelos Voyiatzis) PR #4113
  • New pair style pedone for the non-Coulomb part of the Pedone (PMMCS) potential (Axel Kohlmeyer, Temple U) PR #4129
  • Small updates for the ELECTRODE package (new qtotal, eta keywords, check for mobile atoms) (Ludwig Ahrense-Iwers, Robert Meissner, TU Hamburg, Shern Tee, Griffith U.) PR #4095
  • Small updates to fix pimd/langevin to improve the support for the pimd method (Yifan Li, Princeton) PR #4107
  • Improve sorting performance for Kokkos with GPUs and address (Stan Moore, SNL), PR #4127
  • Update bundled Kokkos library to version 4.3.00 (Stan Moore, SNL and the Kokkos developers) PR #4096, PR #4124
  • Add option to package command that selects building the atom map on the CPU instead of the GPU device. This can lead to significant speedups for runs of large molecular systems (billions of molecules) with a very large number of MPI ranks (> 1000) during initialization (up to 100x at 4096 ranks) and during the run (up to 2x with 4096 ranks) (Stan Moore, SNL) PR #4079
  • Add support for general triclinic simulation boxes and systems (Steve Plimpton) PR #3984
  • Add support for ACE descriptors to the ML-IAP package (James Goff, SNL) PR #4078
  • Add support for a conical shape to fix indent (Evangelos Voyiatzis, NovaMechanics Ltd) PR #3989
  • Many small bug fixes, minor code improvements and refactoring, and updates for coding style, build system, documentation, and formatting (multiple authors) PR #4074, PR #4080, PR #4082, PR #4084, PR #4085, PR #4086, PR #4088, PR #4089, PR #4090, PR #4091, PR #4094, PR #4097, PR #4098, PR #4099, PR #4108, PR #4101, PR #4102, PR #4103, PR #4105, PR #4111, PR #4116, PR #4120, PR #4121, PR #4122, PR #4123, PR #4125, PR #4126, PR #4130, PR #4134, PR #4135, PR #4139

Backward compatibility notes:

  • The zlo, zhi value of a box for a 2d system may no longer be exactly zero. zlo must be less than zero, zhi must be greater.
  • The dump style custom/cfg/yaml etc. keywords "temperature" and "heatflow" are no longer available directly, they must be accessed via compute property/atom now (like other atom style properties)

Update 3 for Stable release 2 August 2023

02 Mar 20:24
stable_2Aug2023_update3
46265e3
Compare
Choose a tag to compare

This is the third update to the 2 August 2023 stable release that fixes bugs that have been reported since then and updates some libraries, dependencies, tools, and build system support.

The following individual changes are included:

  • correct factor of 2 force error in angle style cosine/periodic for multiplicity 1
  • avoid 32-bit integer overflow when running simulations of large systems requiring -DLAMMPS_BIGBIG with PPPM and SHAKE
  • fix bug when using "unpack_exchange()" with KOKKOS on device
  • fix bug in fix style neb which was not always creating the rootworld communicator when needed
  • fix bug in neb command where the print verbosity was not initialized
  • fix bugs in fix style bond/react backported from PR #3905
  • fix bugs in pairstyle meam/ms for non-zero t1m parameter settings and incorrect use of the ialloy parameter.
  • fix scaling bug in compute xrd when using radians as input
  • error out if fixes gcmc, sgcmc, widom, charge regulation and pair style dscmc are used without per-type masses
  • flag that INTEL package is incompatible with run style verlet/split
  • make fix cmap compatible with output from charmm-gui which does not create complete cmap parameter files
  • gracefully handle reaxff parameter files without hydrogen bond parameters
  • update references to (no longer available) fix ave/spatial with correct alternatives
  • avoid initializing mliappy multiple times
  • allow use of QEQ fixes with GPU package (must explicitly set newton on, though)
  • use better detection and application of cuFFT library for KOKKOS with CMake builds
  • correct compilation settings for compiling ML-QUIP package correctly with CMake
  • export symbols from executable when building with a static LAMMPS library so they can be accessed by plugins

Update 1 for Feature release 7 February 2024

14 Feb 00:01
patch_7Feb2024_update1
254a849
Compare
Choose a tag to compare

This update addresses some compilation issues using CMake with the 7 February 2024 version that only happen with custom compilations for some unusual combination of LAMMPS packages (hence the issue was not detected by our automated integration testing).
The only pull request included is #4077. There are no functional changes, so the binary packages attached to the 7 February 2024 feature release do not need to be rebuilt for this update.

Feature release 7 February 2024

07 Feb 23:55
patch_7Feb2024
b5a1c1a
Compare
Choose a tag to compare
Pre-release

Changes since the 21 November 2023 release:

  • New compute pace command for evaluating ACE descriptors which functions similar to compute snap (James Goff, Drew Rohskopf, Aidan Thompson SNL) PR #3869
  • New "ternary()" function for variables (Steve Plimpton) PR #3987
  • Add support for direct references to variable functions for custom properties introduced with fix properties/atom (Steve Plimpton) PR #3990
  • New compute to identify under-coordinated particles (rattlers) and a new fix to measure the nonaffine displacement of particles (Joel Clemmer, SNL) PR #3858
  • New compute reaxff/atom to get access to local bond information for ReaxFF (Richard Berger, LANL) PR #3938
  • Add feature to compute sna/atom to use a fixed number of neighbors, and a new compute slcsa/atom for supervised learning structural analysis, (Paul Lafourcade, CEA) PR #3996
  • add ZBL core-repulsion to ML-PACE, automatic conversion of supported potential file formats (Yury Lysogorskiy, RUB) PR #3985
  • Several new pair styles for the GPU package including coul/long/slater/gpu (Trung Nguyen, U Chicago) PR #4009
  • Add KOKKOS package versions of fix temp/rescale and fix temp/berendsen (Stan Moore, SNL) PR #4012
  • Add KOKKOS package versions of pair style lj/charmmfsw/coul/long and dihedral style charmmfsw, (Mitch Murphy) PR #4030
  • Add support for heFFTe library to do FFTs (Miroslav Stoyanov, ORNL) PR #3963
  • Decouple defines for the FFT libraries for regular and KOKKOS 3d-FFT support, add displaying FFT information with info and lmp -h (Nick Hagerty, ORNL and Axel Kohlmeyer, Temple U) PR #4007
  • Updates for multiple INTERLAYER potential files and bugfixes for a couple of pair styles (Wengen Ouyang, Wuhan University) PR #4025 and PR #4029
  • Update of the bundled linear algebra library to the state of LAPACK 3.12.0 (Axel Kohlmeyer, Temple U) PR #3992
  • Update of the bundled fmtlib sources to version 10.2.1 (Axel Kohlmeyer, Temple U) PR #4042
  • Update the bundled Kokkos library to version 4.2.0 (Stan Moore, SNL, and the Kokkos developers) PR #3983
  • ML-PACE pair style optimizations for KOKKOS with GPUs (Stan Moore, SNL) PR #4031
  • Use C++ templating to reduce the number of NPair and NStencil classes and improve code re-use (Joel Clemmer, SNL) PR #3430
  • Add Born matrix support to more potentials & inclusion of angle and dihedral contribution to compuite stress/mop and stress/mop/profile (Evangelos Voyiatzis, NovaMechanics Ltd.) PR #3833
  • HowTo tutorial for using Moltemplate (Otello Roscioni, MaterialX Ltd), PR #4002
  • Add support for a "Dipoles" section to molecule files (Axel Kohlmeyer, Temple U) PR #4046
  • Add fix numdiff based tests to the force style testers in the unittest tree (Axel Kohlmeyer, Temple U) PR #4041
  • Many small bug fixes, minor code improvements and refactoring, and updates for coding style, build system, documentation, and formatting (multiple authors) PR #3905, PR #3962, PR #3988, PR #3993, PR #3997, PR #3998, PR #4000, PR #4001, PR #4003, PR #4004, PR #4005, PR #4006, PR #4011, PR #4015, PR #4018, PR #4021, PR #4022, PR #4023, PR #4034, PR #4036, PR #4037, PR #4039, PR #4040, PR #4047, PR #4048, PR #4051, PR #4054, PR #4058, PR #4062, PR #4063, PR #4067, PR #4070, PR #4072

Backward compatibility notes:

  • The "reax/c" alias to provide backward compatibility for "reaxff" commands and keywords has been (finally) removed. Now using "reaxff" is required.
  • The format of binary restart files has changed in an incompatible way for cases where bond style lepton or angle style lepton are used.