Public development project of the LAMMPS MD software package

Last update: Dec 30, 2022

Overview

This is the LAMMPS software package.

LAMMPS stands for Large-scale Atomic/Molecular Massively Parallel
Simulator.

Copyright (2003) Sandia Corporation.  Under the terms of Contract
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
certain rights in this software.  This software is distributed under
the GNU General Public License.

----------------------------------------------------------------------

LAMMPS is a classical molecular dynamics simulation code designed to
run efficiently on parallel computers.  It was developed at Sandia
National Laboratories, a US Department of Energy facility, with
funding from the DOE.  It is an open-source code, distributed freely
under the terms of the GNU Public License (GPL).

The primary author of the code is Steve Plimpton, who can be emailed
at [email protected].  The LAMMPS WWW Site at lammps.sandia.gov has
more information about the code and its uses.

The LAMMPS distribution includes the following files and directories:

README                     this file
LICENSE                    the GNU General Public License (GPL)
bench                      benchmark problems
cmake                      CMake build files
doc                        documentation
examples                   simple test problems
fortran                    Fortran wrapper for LAMMPS
lib                        additional provided or external libraries
potentials                 interatomic potential files
python                     Python wrappers for LAMMPS
src                        source files
tools                      pre- and post-processing tools

Point your browser at any of these files to get started:

https://lammps.sandia.gov/doc/Manual.html         LAMMPS manual
https://lammps.sandia.gov/doc/Intro.html          hi-level introduction
https://lammps.sandia.gov/doc/Build.html          how to build LAMMPS
https://lammps.sandia.gov/doc/Run_head.html       how to run LAMMPS
https://lammps.sandia.gov/doc/Commands_all.html   Table of available commands
https://lammps.sandia.gov/doc/Library.html        LAMMPS library interfaces
https://lammps.sandia.gov/doc/Modify.html         how to modify and extend LAMMPS
https://lammps.sandia.gov/doc/Developer.html      LAMMPS developer info

You can also create these doc pages locally:

% cd doc
% make html                # creates HTML pages in doc/html
% make pdf                 # creates Manual.pdf

Comments

cmake: major refactor
Use more modern cmake paradigms.

[x] Remote old style include_directories() calls

[x] Remove old style add_definitions() calls

[x] Create imported target for

[x] NetCDF

[x] QUIP

[x] turn linalg, mpi_stubs in ${PKG_LIB} in object libraries - Nope, CMake-3.10 doesn't have full support for OBJECT libs yet

[x] Create exported target for lammps library

[x] Drop pkg-config files? - Nope

[x] Update old style (without PRIVATE/ PUBLIC keyword) target_link_libraries() calls

cmake ready_for_merge
opened by junghans 106
Develop native support for KIM simulator models
Summary

This PR will create built-in support for KIM Simulator Models

Author(s)

Ryan S. Elliott (U. Minnesota) Ellad B. Tadmor (U. Minnesota) Axel Kohlmeyer (Temple U)

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Implementation Notes

Most of the new commands work by essentially generating LAMMPS commands, either individually or from a file and feeding them to the LAMMPS instance. For clarity, they will generated begin/end markers to show the generated LAMMPS commands like in the following example log output segments:

kim_init SW_StillingerWeber_1985_Si__MO_405512056662_005 real #=== BEGIN kim-init ========================================== units real #=== END kim-init ============================================

kim_interactions Si #=== BEGIN kim_interactions ================================== pair_style kim SW_StillingerWeber_1985_Si__MO_405512056662_005 pair_coeff * * Si #=== END kim_interactions ====================================
kim_package update
opened by ellio167 90
ML potentials with proper orthogonal descriptors
Summary

This pull request implements the fitting and calculation of a new machine-learning interatomic potential (ML-IAP) based on proper orthogonal descriptors (POD). The implementation is located in the folder src/ML-POD, while the documentation is located in the folder doc/src. The new features of this pull request include:

Users can train POD potentials inside LAMMPS.

Users can perform MD simulations with the trained POD potentials.

Users can combine POD and SNAP to create hybrid POD-SNAP potentials.

Users can make linear and quadratic POD, POD-SNAP potentials.

Related Issue(s)

This pull request does not address any open GitHub issue.

Author(s)

Ngoc Cuong Nguyen, Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, 02139, MA. Email: [email protected] and [email protected]

Andrew Rohskopf, Sandia National Laboratories. Email: [email protected].

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

This pull request should not break backward compatibility for inputs.

Implementation Notes

In this pull request, three new C++ classes were created, including mlpod, pair_mlpod, and fitpod_command, to implement the fitting command and pair style for POD. Examples are provided in the examples/mlpod folder to demonstrate the implementation and its performance.

This pull request should not affect other features in LAMMPS.

Post Submission Checklist

[x] The feature or features in this pull request is complete

[x] Licensing information is complete

[x] Corresponding author information is complete

[x] The source code follows the LAMMPS formatting guidelines

[x] Suitable new documentation files and/or updates to the existing docs are included

[x] The added/updated documentation is integrated and tested with the documentation build system

[x] The feature has been verified to work with the conventional build system

[x] The feature has been verified to work with the CMake based build system

[x] Suitable tests have been added to the unittest tree.

[x] A package specific README file has been included or updated

[x] One or more example input decks are included

Further Information, Files, and Links

The original code of this pull request is hosted at https://github.com/cesmix-mit. The Github repository https://github.com/cesmix-mit/pod_examples contains additional examples for this pull request.

More information about the authors can be found on their personal websites https://www.mit.edu/~cuongng/ and https://rohskopf.github.io/. Technical details about the methodology implemented in this pull request can be found in the paper https://arxiv.org/abs/2209.02362.

The research of Dr. Nguyen is supported by Center for the Exascale Simulation of Materials in Extreme Environments https://computing.mit.edu/cesmix/.
enhancement
opened by exapde 57
A shifted Tersoff Potential
Summary

This potential is designed for simulations of closely matched van der Waals heterostructures. For instance, let's consider the case of a system with few-layers graphene atop a thick hexagonal boron nitride (h-BN) substrate simulated using periodic boundary conditions. The experimental lattice mismatch of ~1.8% between graphene and h-BN is not well captured by the equilibrium lattice constants of available potentials, thus a small in-plane strain will be introduced in the system when building a periodic supercell. To minimize the effect of strain on simulation results, the tersoff/shift style is proposed that allows adjusting the equilibrium bond length of one of the two materials (e.g., h-BN). Validitation, benchmark tests and applications of the tersoff/shift style can be found in ACS Nano 13, 7603 (2019) and J. Chem. Theory Comput. 16, 666 (2020).

Related Issue(s)

Author(s)

Wengen Ouyang, Tel Aviv University, Email: w.g.o[email protected] Davide Mandelli, Istituto Italiano di Tecnologia

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

The backward compatibility is kept

Implementation Notes

The tersoff/shift style computes the energy E of a system of atoms, whose formula is the same as the original Tersoff potential. The only modification is that the original equilibrium bond length r_0 of the system is shifted to r_0-\delta, where values of \delta < 0 will result in elongation of the bond length, while values of \delta > 0 will result in shrinking of the bond length.

Post Submission Checklist

[x] The feature or features in this pull request is complete

[x] Licensing information is complete

[x] Corresponding author information is complete

[x] The source code follows the LAMMPS formatting guidelines

[x] Suitable new documentation files and/or updates to the existing docs are included

[x] The added/updated documentation is integrated and tested with the documentation build system

[x] The feature has been verified to work with the conventional build system

[x] The feature has been verified to work with the CMake based build system

[x] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[x] One or more example input decks are included

Further Information, Files, and Links

Validitation, benchmark tests and applications of the tersoff/shift style can be found in https://pubs.acs.org/doi/10.1021/acsnano.9b00645 and https://pubs.acs.org/doi/10.1021/acs.jctc.9b00908.
user_misc_package
opened by oywg11 47
ML_PACE with extrapolation grade / active learning
Summary

Extension of ML-PACE to compute extrapolation grade for active learning applications

Author(s) Yury Lysogorskiy ([email protected])

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

no breaking of backward compatibility is expected

Implementation Notes

New pair style pace/al

Every gamma_grade_eval_freq steps call more expensive ACEBEvaluator using B-basis in YAML format. The relation between potential.yaml and potential.yace currently being used in ML-PACE is shown here.

ACEBEvaluator computes extrapolation grade (following MaxVol algorithm similar to Shapeev's MTP potentials) : current_atom_gamma_grade is computed for each atom here . This value is stored into per-atom array extrapolation_grade_gamma[i] here. The maximum per-atom extrapolation grade is aggregated over MPI processes here and if this max_gamma_grade_per_structure exceeds some thresholds gamma_lower_bound and gamma_upper_bound - dump this value and dump the current structure, and stop (if exceed upper bound).

The per-atom array extrapolation_grade_gamma[i] and per-structure scalar max_gamma_grade_per_structure are accessible to new compute class ComputePaceAtom here and here. This compute style is always created by pair-style pace/al here, and try to "register" in LAMMPS list of computes here. If compute style is invoked in different timestep from gamma_grade_eval_freq frequency, then additional computation of pair_style::compute is invoked here and here.

Dump class to optionally store extrapolative structures is created here but not "registered" into lammps lists.

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links

the ACEBEvaluator and related functionality are represented in ML-PACE/ace, whereas current ML-PACE library sources are in ML-PACE/ace-evaluator
Also, CMakeLists.txt file is added here
pace_package
opened by yury-lysogorskiy 45
Add faster method for multi neighboring
Summary

This pull request contains a slightly modified implementation of a new polydisperse neighboring method developed by Tom Shire, Kevin Stratford, and Kevin Hanley (references below). In the current multi implementation, there is a single binlist with a bin size set by the smallest interaction length. A separate stencil is then constructed for each atom type. For each itype, the stencil extends to the maximum interaction distance with other particles. In the new implementation, a separate binlist is used for each atom type and ~ntype^2 stencils are then constructed for itype-jtype pairs. Stencils extend up to the itype-jtype interaction distance.

This approach greatly speeds up the construction of neighbor lists and may also allow fewer ghost atoms to be communicated. If Newton is on and there are only half neighborlists, one only needs to loop through small particles to look for large neighbors to find cross-interactions. Therefore, small ghost atoms only need to be communicated a distance set by the small-small interaction. In all cases tested, the new method improves performance over the old multi method by a factor of 3+. Therefore, we renamed the old method multi/old and replaced multi with this new method.

In addition to upgrading multi, this pull request also cleans up redundancies in the nstencil classes. Previously, most stencil classes had a full, half+newton, and half+newtoff variety. However, the full and half+newtoff classes were redundant as both created what one might consider a fully "spherical" stencil. This contrasts to half+newton where stencils only included bins to the upper right creating a "hemi-spherical" stencil. To remove this redundancy, all half+newtoff stencils were removed and half+newton stencils were renamed half stencils. The logic in neighbor->choose_stencil() was accordingly updated.

Related Issue(s)

NA

Author(s)

Tom Shire (U. Glasgow) Kevin Stratford (U. Edinburgh) Kevin Hanley (U. Edinburgh) Joel Clemmer (SNL - [email protected])

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

No foreseen issues, input scripts using the old multi style will automatically use the new multi style

Implementation Notes

This PR is still in progress. Remaining tasks include:

[x] Allow users to create custom categories by grouping different types to avoid ntype^2 overhead

[x] Add check in comm_init() to confirm a full neighborlist isn't used with hierachical communication (small-to-large lookups)

[x] Resize bins and atom2bin arrays (# atoms in group vs nall)

[ ] Check compatibility of new stencils with Kokkos

[x] Add examples

[x] Update documentation when arguments are finalized (reference new paper)

[x] Add size_one methods to granular pairstyles

[x] Add option to specify collections by cutoff intervals

[x] Add option to specify non-sequential collections of types

[x] Revamp multi neighbor and comm to use a new, separate collection array of size nlocal+nghost

Extra items (probably a different PR):

Update fix pour to work with >1 atom types

Update fix pour/adapt to work with multi if atom radii can change

Post Submission Checklist

[x] The feature or features in this pull request is complete

[x] Licensing information is complete

[x] Corresponding author information is complete

[x] The source code follows the LAMMPS formatting guidelines

[x] Suitable new documentation files and/or updates to the existing docs are included

[x] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[x] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[x] One or more example input decks are included

Further Information, Files, and Links

Relevant links: Journal article for old method Journal article for new method Technical document for new method Original implementation of new method
neighborlist_feature
opened by jtclemm 41
Fix active
Purpose

This adds a simple fix that applies a constant force along a vector dependent on the atoms' current orientation quaternion. I think this could have been achieved by using addforce and per-vector variables but adding this stuff to a fix seems like a cleaner solution and allows for more easy extension of more advanced features later.

Author(s)

Stefan Paquay @ Brandeis University.

Backward Compatibility

Yes; just adds new feature.

Implementation Notes

This fix applies a constant force to an atom along the "body-frame" z-axis. Therefore, if the atom rotates (due to fix langevin for instance) the force also rotates with it. One check I did is to make sure that, when using fix viscous with a constant friction, the velocity of non-interacting particles converges to the right velocity (given by force magnitude / friction coefficient).

Post Submission Checklist

Please check the fields below as they are completed

[X] The feature or features in this pull request is complete

[X] Suitable new documentation files and/or updates to the existing docs are included

[X] One or more example input decks are included

[X] The source code follows the LAMMPS formatting guidelines

enhancement documentation user_misc_package work_in_progress
opened by Pakketeretet2 41
Support for distributed grids
Summary

Refactor the GridComm class into Grid2d and Grid3d classes which partition and communicate distributed grid data across and between processors for 2d and 3d simulations. "Distributed" means the grid cells or points and the data they store are spread across processors. Each processor owns the points within its subdomain. Note that Grid2d/3d do not store any per-grid data. The data is stored by the calling classes which use Grid2d/3d.

Existing classes that used distributed grids now use these grid classes: PPPM, MSM, fix ttm/grid, pair amoeba/hippo.

Added additional infrastructure to allow computes and fixes to create and output multiple grids (per compute or fix) and multiple data fields on each grid. Added a few new commands to work with distributed per-grid data:

compute property/grid fix ave/grid dump grid dump grid/vtk

Fix ave/grid can either time average per-atom data within each grid cell, or time-average per-grid data directly. The former use case is similar to fix ave/chunk for regular 2d or 3d bins, except that that all processors own a copy of all the bins. So it is less scalable than the new fix ave/grid.

Also added options to dump image and dump_modify to color grid cells in output viz images by grid data.

The motivation for this PR is to standardize how per-grid data is used within LAMMPS and add infrastructure that makes it easier for developers to create their own pair styles or other models which need to use grids in conjunction with particles.

New commands being thought about for future PRs: compute fft/grid fix grid (internal command for use by other styles) others to suggest ?

See the "Howto grid" doc page for an overview of per-grid (distributed) data.

Related Issue(s)

Implements and thus closes #2269

Author(s)

Steve Plimpton, SNL

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

Implementation Notes

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links
enhancement ready_for_merge refactor
opened by sjplimp 40
Add compute born/matrix command to compute elastic stress
Summary

This is a first implementation of the Born term for stress-stress elastic constant computation (see: here or here).

Related Issue(s)

N/A

Author(s) Germain Clavier - Technilogische Universiteit Eindhoven, The Netherlands - [email protected] - github.com/Bibobu

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

None known so far.

Implementation Notes

The method needs coefficients derived from the potential and a configuration dependent part. Since analytical derivation of the potential is needed for the potential coefficient, a specific function needs to be coded for every potential component, if possible. A function signature has been added in each base class for simplicity. Only the pair_lj_cut potential has been modified with a proper method. The compute_born is supposed to make the computation of the configuration part and call the energy coefficient computation when needed for every component.

compute_born.cpp and compute_born.h are added to the src directory. For now method concerning bonds, angles and dihedral are commented waiting for test cases. Improper method has not been coded yet.

pair.h, bond.h, angle.h, dihedral.h and improper.h have an int born_enable declaration and void born method for proper namespace referencing. The new integer value is set to 0 in the respective .cpp files.

pair_lj_cut.h and pair_lj_cut.cpp where modified to include a functional born method. Test on Argon crystal gave comparable results with this article although the implementation was done in a different code: C11=2.27 GPa; C12=1.20 GPa; C44=1.39 GPa. Input script for the test case can be added if needed.

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links
enhancement
opened by Bibobu 40
Support for tiled decompositions in PPPM
Summary

Add support to PPPM for load-balanced decompositions produced by balance and fix balance rcb, i.e. tiled decompositions. The majority of changes for this are in the GridComm class which now stores a regular or tiled decomposition of the FFT grids used by PPPM and does forward and reverse communication of the charge and efield values on those grids. This change should support both orthogonal and triclinic domains and all the options that the PPPM class supports. By the time it is ready to merge all variant PPPM classes, including accelerator versions, should be supported.

Related Issues

N/A

Author(s)

Steve

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

Non-tiled decompositions should work as before.

Implementation Notes

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links
enhancement kspace_package test_runs test_for_regression ready_for_merge
opened by sjplimp 40
Contribute USER-CAC package for multi-scale modeling
Summary

Provides an implementation of the CAC method as presented in: Chen, Y. Reformulation of microscopic balance equations for multiscale materials modeling. The Journal of Chemical Physics 130, 134706, (2009).

The implementation is designed for scalability with both uniform and multi-resolution models.

Author(s)

Adrian Diaz (University of Florida) Email: [email protected]

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

Clear of Conflict (at least without accounting for unforeseen bugs)

Implementation Notes

Several styles are implemented to carry forth the calculations required for the CAC method to be solved in parallel; this includes atom, comm, fix, dump, min, and compute styles. In addition several parent interfaces are added to for operation such as neighbor.cpp.

Correctness was verified with several example CAC and pure MD simulations (to make sure changes didn't break the original LAMMPS code execution). A couple of these examples can be found in the examples folder for USER-CAC.

Post Submission Checklist

Please check the fields below as they are completed after the pull request has been submitted. Delete lines that don't apply

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links
documentation needs_work
opened by Adrian-Diaz 37
Collected small fixes and changes
Summary

The pull request combines multiple small changes and fixes

Related Issue(s)

N/A

Author(s)

Axel Kohlmeyer, Temple U

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

N/A

Implementation Notes

The following individual changes are included:

small cleanup in the mliappy code for Kokkos support to address reports from static analysis with CodeQL

always build the C++-only "mesocnt" styles even if no Fortran is available and mesont/tpm cannot be built.

handle dependency of bond style mesont on bond style harmonic

conditionally add CONFIGURE_DEPENDS keyword to CMake globbing. We do this by defining the variable ${CONFIGURE_DEPENDS} to either by the keyword or empty for CMake prior to 3.12. This will make it easy to turn it always on when we increase the minimum CMake version required. With this change CMake now automatically recognizes when files are added to or removed from the a source folder and re-runs CMake to update the build rules.

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[x] Licensing information is complete

[x] Corresponding author information is complete

[x] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[x] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

python_module cmake refactor
opened by akohlmey 0
[BUG] _LJ/Cut/Intel Incompatable with Ellipsoids

Summary I am trying to run an Intel accelerated simulation of lammps using a hybrid gayberne lj/coul/cut/long potential. After much troubleshooting, I’ve found that the combination of ellipsoids and the LJ/cut/intel potential eventually leads to a particle (seemingly spontaneously) gaining enormous pair energy or thermal energy. The same code with just an LJ/cut (no intel) pair potential runs fine.

LAMMPS Version and Platform LAMMPS (22 Dec 2022) Red Hat Enterprise Linux Server 7.9 (Maipo) Dual 32core Intel Ice Lake 2.6GHz Node and LAMMPS (29 Oct 2020) CentOS Linux 7 (Core)

Expected Behavior When using LJ/cut, the system runs to completion with appropriate thermostatting.

Actual Behavior When using LJ/cut/intel, the system will eventually (typically in < 2e8 steps) gain large amounts of pair energy and/or kinetic energy (>1e6 in lj units), causing one of the atoms to eventually get lost and the simulation to fail. I have tested langevin damping parameters of 1 & 1000, and several of random seeds with the lj/cut/intel always failing, and the lj/cut never failing. The erratic behavior occurs much sooner (~100,000 steps) for larger systems (1000 atoms).

Steps to Reproduce I run an "sbatch lammpssubmit2.txt" command. It runs inp.txt which reads the datafile data.txt, and outputs out.txt

Further Information, Files, and Links data.txt inp.txt lammpssubmit2.txt out.txt
bug user_intel_package

opened by friday20 0

[BUG] Can't build CUDA KOKKOS with CUDA 12 and cmake

Summary

Trying to use cmake for LAMMPS compilation with KOKKOS fails if CUDA 12 is installed.

LAMMPS Version and Platform

Current develop

Expected Behavior

Cmake should correctly prepare files for compilation.

Actual Behavior

An error appears: nvcc fatal : Value 'sm_35' is not defined for option 'gpu-architecture'

Followed by errors from the cmake itself:

CMake Error at /tmp/lammps/lib/kokkos/cmake/kokkos_compiler_id.cmake:12 (STRING):
  STRING sub-command REPLACE requires at least four arguments.
Call Stack (most recent call first):
  /tmp/lammps/lib/kokkos/cmake/kokkos_compiler_id.cmake:53 (kokkos_internal_have_compiler_nvcc)
  /tmp/lammps/lib/kokkos/cmake/kokkos_tribits.cmake:215 (INCLUDE)
  /tmp/lammps/lib/kokkos/CMakeLists.txt:188 (KOKKOS_SETUP_BUILD_ENVIRONMENT)


-- SERIAL backend is being turned on to ensure there is at least one Host space. To change this, you must enable another host execution space and configure with -DKokkos_ENABLE_SERIAL=OFF or change CMakeCache.txt
-- Using -std=c++14 for C++14 standard as feature
CMake Error at /tmp/lammps/lib/kokkos/cmake/kokkos_test_cxx_std.cmake:132 (MESSAGE):
  Invalid compiler for CUDA.  The compiler must be nvcc_wrapper or Clang or
  NVC++ or use kokkos_launch_compiler, but compiler ID was GNU
Call Stack (most recent call first):
  /tmp/lammps/lib/kokkos/cmake/kokkos_tribits.cmake:242 (INCLUDE)
  /tmp/lammps/lib/kokkos/CMakeLists.txt:188 (KOKKOS_SETUP_BUILD_ENVIRONMENT

Steps to Reproduce

Install CUDA 12.0, and run following lines:

git clone https://github.com/lammps/lammps
cd lammps
mkdir build
cd build
cmake -D PKG_KOKKOS=yes -D Kokkos_ENABLE_CUDA=yes ../cmake

The issue also appears if a GPU architecture is specified, e.g. -D Kokkos_ARCH_AMPERE86=yes is added.

Further Information, Files, and Links

The issue arises from the nvcc_wrapper script which uses sm_35 as a default architecture and it was removed in CUDA 12: https://github.com/lammps/lammps/blob/f84b64bc869003320ccc3b83e8cfe428d9b87f0f/lib/kokkos/bin/nvcc_wrapper#L15-L16

This problem is solved if the line with sm_35 is commented and the next one uncommented, but I think it should documented somewhere.

bug

opened by mkanski 7

[BUG] _Fix rigid/npt does not keep bond lengths and angles rigid in SWM4-ndp water molecules

Summary

Fix rigid/npt does not keep bond lengths and angle rigid of the SWM4-ndp water molecules.

LAMMPS Version and Platform

LAMMPS (3 Nov 2022) Operating system: CentOS Linux release 7.9.2009 (Core) Hardware/CPU: Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz

Expected Behavior

Fix rigid/npt keeps bond lengths and angles constant, or with very small variations (machine precision).

Actual Behavior

Bond lengths (O-M and O-H) vary with at least up to 0.3 Å over time (several ns). H-O-H angle also varies by 0.3 degrees in much shorter time.

Steps to Reproduce

Run in.example.txt using data.swm4-ndp.txt as data (system) file.

Further Information, Files, and Links

Input file, data file and output files attached.

in.example.txt log.lammps.txt OH_bond.dump.txt OM_bond.dump.txt data.swm4-ndp.txt HOH_angle.dump.txt
bug

opened by gullbrekken 1
Updates to the DIELECTRIC package
Summary

This PR includes maintenance changes to the pair and kspace dielectric styles to ensure consistency with non-dielectric counterparts.

Related Issue(s)

N/A

Author(s)

Trung Nguyen (UChicago)

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

Should maintain backward compatibility.

Implementation Notes

Previously the pair and kspace styles use the atom charges (atom->q) scaled by local epsilon's. This PR renames the variables so that atom->q has consistent meaning with other packages as unscaled values. The pair and kspace dielectric styles now operate on the scaled values instead. There are also changes to the pppm/dielectric to match the elong value given by the scaled charges with the elong value given by regular pppm with a global dielectric constant. This is a nontrivial issue as electrostatic energy is not linearly dependent upon charge density.

The documentation stays unchanged with the changes in this PR, unless suggested otherwise.

Post Submission Checklist

[x] The feature or features in this pull request is complete

[x] Licensing information is complete

[x] Corresponding author information is complete

[x] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[x] The feature has been verified to work with the conventional build system

[x] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

Further Information, Files, and Links
maintenance dielectric_package
opened by ndtrung81 5
Fix init arguments and some join stuff for Kokkos 4
Summary

Update LAMMPS to work with upcoming Kokkos 4.0 release

Author(s)

Christian Trott [email protected]

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility

Should be backward compatible, including to pre Kokkos 4 releases of Kokkos.

Implementation Notes

Did a bunch of compilation and run tests with bench and examples directory

Post Submission Checklist

[ ] The feature or features in this pull request is complete

[ ] Licensing information is complete

[ ] Corresponding author information is complete

[ ] The source code follows the LAMMPS formatting guidelines

[ ] Suitable new documentation files and/or updates to the existing docs are included

[ ] The added/updated documentation is integrated and tested with the documentation build system

[ ] The feature has been verified to work with the conventional build system

[ ] The feature has been verified to work with the CMake based build system

[ ] Suitable tests have been added to the unittest tree.

[ ] A package specific README file has been included or updated

[ ] One or more example input decks are included

kokkos_package
opened by crtrott 4

Releases(patch_22Dec2022)

patch_22Dec2022(Dec 22, 2022)
Changes since the 3 November 2022 release:

new package ML-POD with a pair style for machine learning potentials using "proper orthogonal descriptors" (POD) and a command for training (Ngoc Cuong Nguyen, MIT and Andrew Rohskopf, SNL) PR #3449

new command reset_atoms that is a front end for reset_atom_ids and reset_mol_ids and the new sub-command reset_atoms image to reset image flags consistently for molecule fragments (Axel Kohlmeyer, Temple U) PR #3530

new features and updates to the ELECTRODE package (Ludwig Ahrens-Iwers and Robert Meißner, Hamburg U of T and Shern Tee, U of Queensland) PR #3544

new fix sgcmc command in the MC package imported from the external USER-VCSGC package (Aidan Thompson, SNL and Axel Kohlmeyer, Temple U) PR #3556

add KOKKOS versions of fix viscous and fix dt/reset (Michal Kanksi, Jagiellonian University) PR #3522

refactored distributed grids code to be more general and add new commands that can access those grids or compute/process/output grid properties (Steve Plimpton) PR #3405

new Python classes and functions to build tables for pair style, bond style, angle style, and dihedral style table (Axel Kohlmeyer, Temple U) PR #3559

updates, bugfixes, and a few new features for fix bond/react (Jake Gissinger, NASA Langley) PR #3514

some more updates to the LAMMPS Fortran interface module (Karl Hammond, U of Missouri) PR #3547

remove style suffixes when writing binary restart files allowing to restart non-accelerated runs from accelerated runs (Axel Kohlmeyer, Temple U and Stan Moore, SNL) PR #3538

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3516, PR #3517, PR #3521, PR #3524, PR #3525, PR #3527, PR #3529, PR #3533, PR #3534, PR #3539, PR #3541, PR #3548, PR #3550, PR #3552, PR #3553, PR #3555, PR #3558, PR #3560, PR #3562, PR #3566

Backward compatibility notices:

The checks for valid floating-point and integer numbers have been improved and are more strict now. This can lead to old inputs failing because of invalid numbers

The Si.b.meam.sw.spline had to be corrected for an invalid floating-point number (see previous note. here the "e" in the exponent was missing and thus the exponent was incorrectly ignored). Most calculations, but not all, will be unaffected since this only concerned the spline coefficients at the very beginning of the table.

The syntax for the python source command has changed slightly to be more consistent with the other features of the command.

The commands reset_atom_ids and reset_mol_ids have been renamed to reset_atoms id and reset_atoms mol, respectively. When the original commands are used, a warning is printed and the commands changed to the new versions.

The box command has been removed. The error check that it was used for to disable has been changed to a warning. If the box command is used, it is ignored and a warning is printed.

Binary restarts written with KOKKOS enabled for older versions of LAMMPS may not read back correctly with this version due an inconsistency in the restart output. Those restarts must be converted into data files using the old executable to be readable.

Source code(tar.gz)
Source code(zip)
stable_23Jun2022_update2(Nov 3, 2022)
This is an update to the 23 June 2022 stable release that fixes a few bugs that have been reported since then and updates some libraries, dependencies, and build system support. There is no new functionality added.

The following individual fixes and changes are included:

restore consistent handling of dump output frequency

backport corrections to the DIELECTRIC package

update/correct documentation for building the ELECTRODE package with the traditional make procedure

avoid race condition when downloading potential files in CMake

adapt VTK support in CMake to be compatible with VTK 9.0 and later

correct/improve OpenMP detection for CMake 3.10+

relax minimum required OpenMP version for KOKKOS package (based on improved OpenMP detection)

prefer versioned cythonize command over unversioned if it matches the selected python version

update default downloaded plumed library version to 2.8.1

adjustments for correct HTML output of the manual with sphinx 5.1+

add workaround for activate inactive hyperlinks for Fortran functions in manual

correctly report mixing for CLASS2 pair styles

increase buffer for internal regex function to accommodate larger strings

correctly apply offsets to shake types in molecule files

correct outdated neighborlist issues when calling delete_atoms or create_bonds multiple times

fix buffer allocation issue for atom style dumps

avoid segmentation faults when using the clear command with atomfile variables and pair/only package flags

update broken URLs in the manual

add support to detect Windows 10 22H2 in OS version info

Source code(tar.gz)
Source code(zip)
patch_3Nov2022(Nov 3, 2022)
Changes since the 15 September 2022 patch release:

New ML-IAP package unified python interface (Steven Anaya and Nicholas Lubbers, LANL) PR #3458

Partial Kokkos support for the ML-IAP package, (Matt Bettencourt and Markus Hrywniak, NVIDIA) PR #3487

New pair style ylz (Mehdi Baghaee,SUSTech Shenzhen) PR #3491

Updates and fixes to the BPM package (Joel Clemmer, SNL) PR #3448

Updates to the ML-PACE package introducing extrapolation grade / active learning (Yury Lysogorskiy, Ruhr-Uni Bochum) PR #3315, PR #3479, PR #3490, PR #3501

Updates and corrections to the SPIN package (Julien Tranchida, CEA) RP #3461

Updates to the Fortran module (Karl Hammond, U Missouri) PR #3396, PR #3470, PR #3508

Update bundled Kokkos library to version 3.7.0 (Stan Moore, SNL and the Kokkos developers) PR #3474

Performance improvements/consistency for the KOKKOS versions of SNAP and Tersoff (Evan Weinberg, NVIDIA) PR #3465, PR #3488

GPU package initialization improvements (Mike Brown, Intel) PR #3478

Support for quartic interaction in pair style list (Tiedong Sun) PR #3473

Support requesting a full neighbor list with pair style zero (Drew Rohskopf, SNL) PR #3471

Integrate Peridynamics user guide PDF as Howto into the LAMMPS manual (Axel Kohlmeyer, Temple U) PR #3462

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3450, PR #3454, PR #3455, PR #3456, PR #3459, PR #3463, PR #3468, PR #3475, PR #3477, PR #3480, PR #3482, PR #3485, PR #3489, PR #3495, PR #3497, PR #3498, PR #3500, PR #3502, PR #3503, PR #3504, PR #3507, PR #3510

Backward compatibility notes:

the default setting for neighbor list rebuilds has been changed from delay 10 to delay 0 to guarantee correctness at some modest performance loss. This can lead to more frequent neighbor list rebuilds or neighbor list rebuilds on different timesteps and thus diverging trajectories.

Unlike other variables, atomfile style variables are now deleted with the clear command. This avoids segmentation faults due to trying to access a deleted fix store/peratom instance.

Source code(tar.gz)
Source code(zip)
patch_15Sep2022(Sep 15, 2022)
Changes since the 3 August 2022 patch release:

Add basic support for symbolic types aka type labels (Jake Gissinger, NASA, Yaser Afshar, UMN, Steve Plimpton, SNL, Axel Kohlmeyer, Temple U) PR #2531

New fix pair command to access per-atom quantities computed by pair styles, also dump_modify skip option to skip writing dump frames depending on a condition (Steve Plimpton, SNL) PR #3369

Major update of the mesocnt pair style with addition of mesocnt bond and angle styles and more to complete the functionality (Philipp Kloza, U Cambridge) PR #3421

Updates and corrections for plugin handling in LAMMPS and the MDI package (Christian Negre, LANL, and Taylor Barnes, MOLSSI, and Steve Plimpton, SNL) PR #3366

Add support for using fix shake and fix rattle during minimization by replacing the constraints with strong restraints (Axel Kohlmeyer, Temple U) PR #3244

Remove some bias from the operation of fix bond/swap caused by how LAMMPS stores neighbors and bonds (Steve Plimpton, SNL) PR #3444

Add support to fix latte so it can be used in combination with fix gcmc and similar (Steve Plimpton, SNL) PR #3436

Split internal fix STORE command into fix STORE/GLOBAL and fix STORE/PERATOM (Axel Kohlmeyer, Temple U) PR #3379

Update embedded fmtlib code to version 9.1.0 (Axel Kohlmeyer, Temple U) PR #3381, PR #3418

Tune INTEL package code for better performance with the new LLVM based Intel compiler (Michael Brown, Intel) PR #3410

Various contributions from participants at the LAMMPS Code Clinic event (multiple authors).

Project 1: Improved error messages: PR #3394, PR #3403, PR #3408, PR #3428

Project 2: Refactor for new APIs: PR #3391, PR #3398, PR #3399, PR #3401

Project 4: Modernize documentation: PR #3411, PR #3427

Project 5: Add expand and new tests: PR #3400, PR #3407, PR #3438

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3378, PR #3380, PR #3382, PR #3383, PR #3384, PR #3387, PR #3388, PR #3389, PR #3390, PR #3392, PR #3393, PR #3395, PR #3406, PR #3409, PR #3414, PR #3419, PR #3420, PR #3422, PR #3425, PR #3432, PR #3433, PR #3440, PR #3442 , PR #3443

Compatibility with external tools:

TopoTools in VMD must be at least version 1.9 to be compatible with type labels. To update your VMD TopoTools plugin, you can download it from the TopoTools release page on GitHub

OVITO is going to be compatible with data files containing type labels soon

Backward compatibility notes:

external LAMMPS code that uses fix STORE will need to be updated

the update of the mesocnt styles is not backward compatible

comments in data and data files must have a blank between the last valid input and the comment character '#'

the mandatory fourth argument (NULL) for fix latte is no longer required. Its function is moved to the coulomb keyword

Source code(tar.gz)
Source code(zip)
stable_23Jun2022_update1(Aug 6, 2022)
This is an update to the 23 June 2022 stable release that fixes a few bugs that have been reported since then and updates some libraries, dependencies and the CMake support for plugins on Windows.

The following individual changes are included: The following individual fixes and changes are included:

Fix corner case with if-then-else Boolean logic where there is just a single string argument.

Also remove references to comparing strings with Boolean operaters like "<" or ">=" in the documentation.

Add missing atom->grow() to Kokkos unpack_exchange to avoid segmentation fault

Fix small memory leak in ML-SNAP

Avoid possible out-of-bounds data access in ML-RANN

Bugfixes to fix bond/react

Bugfixes to pair style mgpt

Avoid segfault in dump vtk when used without a region

Allow ramp(x,y) variable function to be used in between runs (returning x) and avoid division by zero when used with run 0.

Allow vdisplace(), swiggle(), and cwiggle() functions to be used in between runs and disallow using them with fix dt/reset.

Prevent view bounds error when building the Kokkos neigh list and a proc has no atoms

Add missing GPU <--> CPU data transfer in Kokkos minimize, gave wrong results when using comm no or comm/reverse no running on the GPU

fixes for CMake scripts used for building plugins to cross-compile with MPI on Windows

CMake scripting updates to fix multiple issues building the n2p2 library on Ubuntu 18.04LTS and 20.04LTS

add CMake script code to provide a "make package" target when cross-compiling plugins to Windows

mention dump style cfg/uef in the main dump manual page

add missing support for dump_modify colname to local dumps

add more unit tests to improve test coverage of boolean evaluations

must not search for MPI package when MPI is disabled to avoid issues with testing the LAMMPS Fortran module

update embedded Kokkos library to bugfix release version 3.6.1

update googletest library from a snapshot to release version 1.12.1

update the Viscosity howto input script example to simulate liquid argon as advertised

avoid deprecation warnings when building the manual

update python package version requirements for building the manual

Source code(tar.gz)
Source code(zip)
patch_3Aug2022(Aug 3, 2022)
Changes since the 23 June 2022 patch release:

New AMOEBA package with an implementation of the AMOEBA and HIPPO polarizable force fields from the Tinker MD code (Josh Rackers and Steve Plimpton, SNL and Trung Nguyen U Chicago) PR #2898

New compute styles sna/grid and sna/grid/local for the ML-SNAP package (Aidan Thompson, SNL and Lenz Fiedler, HZDR) PR #3305

New pair style srp/react in the MISC package (Vaibhav Palkar, Clemson) PR #3337

Enhancements to the MDI Package with expanded functionality (Steve Plimpton, SNL and Taylor Barnes, MolSSI) PR #3307, PR #3336

Rename CG-SDK package to CG-SPICA and rename all related commands, new functional form LJ12-5 (in addition to LJ12-6 and LJ12-4, and LJ9-6) (Yusuke Miyazaki, Okayama U) PR #3343

Add KOKKOS package version of the MEAM package (Naga Vydyanathan, NVIDIA and Stan Moore, SNL) PR #3328

Update bundled Kokkos library to version 3.6.1 (Stan Moore, SNL and the Kokkos developers) PR #3329

New "threebody on/off" keyword for pair style sw that allows turning off the three-body term calculation and enabling performance optimizations for that case (Jackson Elowitt, U Utah) PR #3316

Optimizations for reading large blocks of input (Steve Plimpton, SNL) PR #3323

Corrections and enhancements to how groups are processed in compute contact/atom (Joel Clemmer, SNL) PR #3339

Add "pair_modify neigh/trim yes/no" option to allow to build optimized, trimmed neighbor lists for hybrid/overlay (Stan Moore, SNL) PR #3341

Add "timestep yes/no" option to the read_dump command to simplify simulation loops where atoms are added from a dump file (Axel Kohlmeyer, Temple U) PR #3344

Improved detection of Python interpreter and libraries in CMake to be more consistent across platforms, CMake versions and better honor the documented ways to override the automatic choice (Axel Kohlmeyer, Temple U) PR #3355

Update of bundled linalg library to LAPACK version 3.10.1 plus adding ZHEEVD and dependencies and a new CMake setting -DUSE_INTERNAL_LINALG to allow preferring the bundled library over external ones (Axel Kohlmeyer, Temple U) PR #3360

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3317, PR #3319, PR #3322, PR #3327, PR #3298, PR #3330, PR #3331, PR #3340 , PR #3342, PR #3346, PR #3349, PR #3350, PR #3358, PR #3359, PR #3361, PR #3364, PR #3365, PR #3370, PR #3372, PR #3374, PR #3377

Backward compatibility notes:

the command fix mdi/aimd has been renamed to fix mdi/qm

the new pair_modify neigh/trim option defaults to yes. Use it with no in case simulations with hybrid/overlay fail and report the details in a new issue here

the CG-SDK package has been renamed CG-SPICA, all style have been renamed accordingly, but backward compatibility to the old names is included

Source code(tar.gz)
Source code(zip)
stable_23Jun2022(Jun 22, 2022)
Below is a list of major changes since the last stable release 29 September 2021

This stable release adds about 3000 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.

With this release we will remove the previously used branches master and unstable and use the current equivalents develop and release. The stable keeps its name, but since we published bugfix updates to the last stable release it will now run parallel to develop and be explicitly synchronized with it (technically speaking, the patch_23Jun2022 and the stable_23Jun2022 tags will point to different commit hashes due to the difference in their histories, but the content will be identical).

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.

New platform namespace with abstractions for platform specific operations or getting information about platforms and tools. This includes dynamic loading of shared objects and file and directory operations.

start ongoing project to change how errors and warnings are reported and documented

Improved portability, especially to Windows. It is now possible to build most of LAMMPS (including unit testing) natively on Windows with Visual Studio 2022 and MSVC++ compilers or Clang or Intel OneAPI compilers. Settings for Visual Studio are included.

Improved plugin mechanism and option to automatically load custom plugins that are either not distributed with LAMMPS or depend on libraries that have licensing terms or technical requirements preventing linking them into LAMMPS directly. Examples for how to build such plugins for the KIM and the ML-PACE package are included. For the latter also a NSIS script to build an installer for Windows.

Semi-automatic refactoring with clang-tidy: this has updated the coding style to take advantage of C++11 features and make things more explicit, so the compiler can warn more easily about problems. Also, clang-format is now applied to more files.

update force style unit tests to work with Kokkos + OpenMP

Refactor handling of thermodynamic output for better readability and add new thermo_style "yaml"

Automatically load plugins in folders listed in the LAMMPS_PLUGIN_PATH environment variable

More content and improvements for the Programmer Guide section of the Manual

Updates and new commands or styles or packages:

New fix acks2/reaxff for ACKS2 charge equilibration and updates to fix qeq/reaxff that add support for fix efield to both charge equilibration methods.

New fix mol/swap for swapping of atom types within a single, randomly selected molecule

Add support for multi-species potentials to ML-PACE package

Allow GPU package pair styles to be used with newton_pair on

New pair style sw/mod providing an empirically modified Stillinger-Weber potential suitable for compounds like MoS2

New pair style nm/cut/split and bond style fene/nm for coarse grain polymer models

Support for binary native dump files for read_dump and rerun

Updated singularity/apptainer definition files

New compute ave/sphere/atom to compute averaged properties in a sphere around atoms

Support writing to dump files at regular time interface for variable timestep simulations

Add centroid atomic stress support for shake, rattle and rigid/small fixes

Add new pair style harmonic/cut providing a repulsive-only harmonic potential

Add new fix numdiff/virial for deriving virial stress from potential energy via numerical differences

Add two new interlayer pair styles ilp/tmd and saip/metal

new option for multi-file dumps to have files that are more evenly sized to address post-processing and visualization issues for simulations of very large systems

accelerator support (including KOKKOS) and improvements for dynamical_matrix and third_order commands

re-implementation of the tools/eam_database Fortran code in Python

update setup.py and install.py for python module to use setuptools instead of deprecated distutils and build a binary wheel file

Add computes for local components of the pressure tensor in cartesian and spherical coordinates

Major update of the LATBOLTZ package

Code optimization for several KOKKOS styles

Add fixes for damping of rotational and translational energy for extended particles

Add support for an inner cutoff switching function to SNAP related commands that smoothly turns of SNAP descriptors at short range

Memory usage optimization and bugfix for mesocnt pair style

new BPM package for bonded particle models

new ELECTRODE package

updates and additions to the DIELECTRIC package

new compute fep/ta for the FEP package implementing test-area perturbation

new pair styles ilp/graphene/hbn/opt, ilp/tmd/opt, and saip/metal/opt for the OPT package with 2x speedup

new pair styles smatb and smatb/single for the SMTBQ package

new dump_style yaml command in the EXTRA-DUMP package

add Kspace support to the PLUGIN package

enhancements and additions for the MDI package. This obsoletes the MESSAGE package.

additions to the BROWNIAN package to support different rotational and translational temperatures and option to constrain rotation in a plane in 3d

additions to PyTorch support in ML-IAP packge

new compute born/matrix command to compute elastic stress

add KOKKOS package versions of the pair styles from the DPD-BASIC package

add KOKKOS package version of pair style adp

new colname option for dump_modify and thermo_modify to replace the column header strings for thermo output or in dump files

add support for writing yaml format files to fix ave/time and also support fix_modify colname similar to thermo and dump

add support for angle styles to fix adapt and add or update "extract()" methods to some angle/bond styles and make the extract arguments more consistent

add region style ellipsoid

add "aveabs" and "sumabs" options to compute reduce

add "variable" option to delete_atoms command

bugfix for DOF handling in compute temp/profile

more consistent behavior of dynamic groups through moving the invocation of the internal fix to a later time in the timestep

make time thermo property restartable and implement a reset_timestep time option to set it manually

remove support for legacy GPU hardware and CUDA toolkits in the GPU package

Update of the bundled Kokkos library to version 3.6.0

Update of the bundled Colvars library to version 2022-05-24

Add KOKKOS package version of the pace pair style from the ML-PACE package

New pair styles "sw/angle/table" and "threebody/table" for the MANYBODY package

Add support for GPU accelerated FFTs with HIP and KOKKOS

Add support for SO3 descriptors to the ML-IAP package

New "overlap" option for create_atoms random that will avoid close contacts

New "mesh" option for create_atoms to approximate meshes from STL files with atoms

Add timer style variables for measuring wall time in seconds from the input script

Backward compatibility notices:

delete_atoms porosity has been renamed to delete_atoms random and uses/supports different options

The CC.Lebedeva interlayer potential file has been updated. It now uses by default parameters for "normal" conditions. The previous default parameters are tuned for high-pressure conditions and are still available when using "C1" instead of "C" as element.

Using the PyLammps python module now requires the presence of numpy

The new version of make install-python will not install a package over an existing package as the old version did. Instead, it will attempt to uninstall the old package and then install the new version. Packages installed with distutils, like the old version of make install-python, must be manually uninstalled (they don't store a suitable manifest). This should only be needed once, since from then on uninstall should be possible.

The update to the LATBOLTZ package is a significant update that changes the syntax of the fix lb/fluid command in a way that is not backward compatible. Inputs must be updated accordingly. Also fixes lb/pc and lb/pc/rigid/sphere have been removed and fixes nve and rigid should be used in their stead.

The thermodynamic output format is slightly changed. This may break parser implementations for analysis of LAMMPS log files that are not indifferent to whitespace changes and width of fields.

the MESSAGE package has been removed. its functionality is provided in a more general way by the MDI package

the GPU package using CUDA now requires CUDA toolkit version 8 or later and GPU architector 3.0 (aka Kepler or later). When compiling for OpenCL, driver support for OpenCL 1.2 is required. OpenCL can provide support for GPUs not supported by the CUDA toolkit anymore.

the bug fix to the DOF calculation in compute temp/profile changes the results (usually by a small margin)

Source code(tar.gz)
Source code(zip)
patch_23Jun2022(Jun 22, 2022)
This patch contains the final changes to update the 2 June 2022 release to the 23 June 2022 stable release.

There are no substantial new features in this patch, only a variety of small updates and bug fixes for reported issues and based on extended testing. Notable changes are:

Refactoring of the plugin examples and a new example building the ML-PACE package as plugin including creating a windows installer. (Axel Kohlmeyer, Temple U) PR #3290

Bugfix for atom style dielectric

Bugfix for dump style xtc on Windows (Issue #3292)

Bugfix for dump style yaml when running in parallel PR #3309

Change "remap semantics" for the read_restart command to always use the remap option by default and offer an "noremap" option to be able to change to the old behavior. (Issue #3078)

Bugfix for accessing local computes in Python via NumPy, support LMP_SIZE_VECTOR as alias for LMP_SIZE_ROWS with local computes PR #3309

Updates and bugfixes for the "Fortran2" and "Fortran_dftb" Fortran interfaces, PR #3312

Improvements to compute ave/sphere/atom (Stan Moore, SNL) PR #3293

Fix integer overflow bug in Kokkos version of ReaxFF (Stan Moore, SNL) PR #3295

Updates to singularity/apptainer container definition files PR #3290

GitHub actions updates PR #3299, PR #3301, PR #3303, PR #3304

Recover building binary python wheel file after upgrade to use venv instead of virtualenv. When running with an active virtual environment, the wheel will be installed into the virtual environment as to be expected instead of the system or user site-packages folder PR #3306

Various small improvements and small corrections to build system, compiler settings, and documentation. PR #3296, PR #3300

Source code(tar.gz)
Source code(zip)
patch_2Jun2022(Jun 3, 2022)
This is the first release candidate for the next stable release of LAMMPS. Please test thoroughly and report bugs immediately.

Changes since the 4 May 2022 patch release:

Update of the bundled Kokkos library to version 3.6.0 (Stan Moore, SNL and the Kokkos developers) PR #3251

Update of the bundled Colvars library to version 2022-05-24 (Giacomo Fiorin, NIH and the Colvars deverlopers) PR #3260, PR #3286

Add KOKKOS package version of the pace pair style from the ML-PACE package (Stan Moore, SNL) PR #3264, PR #3278

New pair styles "sw/angle/table" and "threebody/table" for the MANYBODY package (Christoph Scherer, Max Planck Institute for Polymer Research Mainz) PR #3279

Add support for GPU accelerated FFTs with HIP and KOKKOS (Emily Kahl, The University of Queensland) PR #3237

Add support for SO3 descriptors to the ML-IAP package (Aloïs Castellano, University of Liège) PR #3263

New "overlap" option for create_atoms random that will avoid close contacts (Eugen Rožić, Institute Ruđer Bošković, and Steve Plimpton, SNL) PR #3119

New "mesh" option for create_atoms to approximate meshes from STL files with atoms (Axel Kohlmeyer, Temple U and Joel Clemmer, SNL) PR #3258

Add timer style variables for measuring wall time in seconds from the input script (Steve Plimpton, SNL) PR #3255

Optimizations for compute orientorder/atom (Tomas Oppelstrup, LLNL and Stan Moore, SNL) PR #2988

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3250, PR #3259, PR #3262, PR #3267, PR #3268, PR #3269, PR #3271, PR #3272, PR #3275, PR #3276, PR #3277, PR #3280, PR #3281, PR #3284, PR #3285, PR #3288, PR #3289

Backward compatibility notices:

The delete_atoms porosity option has been renamed to random and generalized to support multiple sub-styles that have adjustable behavior with respect to exactness and performance of the deletion.

Source code(tar.gz)
Source code(zip)
patch_4May2022(May 5, 2022)
Changes since the 24 March 2022 patch release:

new BPM package for bonded particle models (Joel Clemmer, SNL) PR #2906

new ELECTRODE package (Ludwig Ahrens-Iwers and Robert Meißner, Hamburg University of Technology and Shern Tee, University of Queensland) PR #3194

updates and additions to the DIELECTRIC package (Trung Nguyen, U Chicago) PR #3212

new compute fep/ta for the FEP package implementing test-area perturbation (Shifeng Ke, Zhejiang University) PR #3192

new pair styles ilp/graphene/hbn/opt, ilp/tmd/opt, and saip/metal/opt for the OPT package with 2x speedup (Ping Gao and Xiaohui Duan, National Supercomputing Center in Wuxi) PR #3228, PR #3242

new pair styles smatb and smatb/single for the SMTBQ package (Daniele Rapetti - Politecnico di Torino) PR #3031

new dump_style yaml command in the EXTRA-DUMP package (Axel Kohlmeyer, Temple U) PR #3205

add Kspace support to the PLUGIN package (Jinzhe Zeng, Rutgers University) PR #3188

enhancements and additions for the MDI package. This obsoletes the MESSAGE package. (Steve Plimpton, SNL and Taylor Barnes, MolSSI) PR #2921, PR #3207, PR #3218

additions to the BROWNIAN package to support different rotational and translational temperatures and option to constrain rotation in a plane in 3d (Sam Cameron, University of Bristol) PR #3193

additions to PyTorch support in ML-IAP packge (Charlie Sievers) PR #3182

new compute born/matrix command to compute elastic stress (Germain Clavier - TU Eindhoven) PR #3093

add KOKKOS package versions of the pair styles from the DPD-BASIC package (Matt Bettencourt, Nvidia) PR #3198

add KOKKOS package version of pair style adp (Vladislav Galigerov and Vsevolod Nikolskiy, Higher School of Economics, Russia) PR #3247

KOKKOS optimizations and bugfixes: GPU optimizations for neighbor lists and ReaxFF (Stan Moore, SNL and Evan Weinberg, NVIDIA) PR #3187, PR #3195, PR #3203

new colname option for dump_modify and thermo_modify to replace the column header strings for thermo output or in dump files (Axel Kohlmeyer, Temple U) PR #3197

add support for writing yaml format files to fix ave/time and also support fix_modify colname similar to thermo and dump (Axel Kohlmeyer, Temple U) PR #3240

add support for angle styles to fix adapt and add or update "extract()" methods to some angle/bond styles and make the extract arguments more consistent (Steve Plimpton, SNL) PR #3229, PR #3236

add region style ellipsoid (Evangelos Voyiatzis, NovaMechanics Ltd) PR #2871

add "aveabs" and "sumabs" options to compute reduce (Tim Bernhard, ETH Zürich) PR #3246

add "variable" option to delete_atoms command (Axel Kohlmeyer, Temple U) PR #3243

bugfix for DOF handling in compute temp/profile (Stephen Sanderson, University of Queensland) PR #2667

more consistent behavior of dynamic groups through moving the invocation of the internal fix to a later time in the timestep (Steve Plimpton, SNL with an assist from Davie Quigley, U Warwick) #3224

make time thermo property restartable and implement a reset_timestep time option to set it manually (Axel Kohlmeyer, Temple U) PR #3248

remove support for legacy GPU hardware and CUDA toolkits in the GPU package (Axel Kohlmeyer, Temple U) PR #3213

start project to change how errors and warnings are reported and documented (Axel Kohlmeyer, Temple U and Steve Plimpton, SNL) PR #3226, PR #3227, PR #3231

Error and warning message docs from header files are removed

A new document is started collecting and discussing/explaining errors that are frequently discussed or looked up. Each paragraph is assigned an error code, corresponding error messages have a string added with an URL to the online documentation pointing to those newly added paragraphs

Individual "short" error messages like "Illegal xxxx command" are expanded to provide more details to make them (mostly) self-explanatory and thus reduce the need to look them up in the source code

Many small bug fixes, minor code improvements and refactoring, and coding style, build system, documentation, formatting updates (multiple authors) PR #3186, PR #3189, PR #3191, PR #3196, PR #3199, PR #3201, PR #3202, PR #3206, PR #3208, PR #3210, PR #3211, PR #3214, PR #3215, PR #3216, PR #3217, PR #3219, PR #3220, PR #3221, PR #3222, PR #3223, PR #3225, PR #3230, PR #3232, PR #3233, PR #3234, PR #3238, PR #3241

Backward compatibility notes:

the MESSAGE package has been removed. its functionality is provided in a more general way by the MDI package

the GPU package using CUDA now requires CUDA toolkit version 8 or later and GPU architector 3.0 (aka Kepler or later). When compiling for OpenCL driver support for OpenCL 1.2 is required, which usually supports GPUs not supported by the CUDA toolkit anymore.

the bug fix to the DOF calculation in compute temp/profile changes the results (usually by a small margin)

Source code(tar.gz)
Source code(zip)
stable_29Sep2021_update3(Mar 24, 2022)
This is the third and final update to the 29 September stable release that fixes bugs that have been reported since the second update and backported from the develop branch.

The following individual changes are included:

remove obsolete check so that larger systems can be run with ReaxFF

update the Fortran source code in tools/eam_database to produce files with fewer oscillations and better precision. add python version.

fix problems causing crashes with the MSM kspace style when continuing a run without and with changing box settings

avoid using an outdated list of special neighbors when applying the angle constraint in fix bond/create/angle

fix bug in fix rigid/npt/small and fix rigid/nvt/small leading to NaNs for 2d systems

fix memory leaks and avoid accessing uninitialized data

fix a tag caching and skip list generation issue causing crashes when using pair style sw/intel with hybrid pair styles

fix off-by-one bug in dihedral style fourier/intel

fix neighborlist and buffer creation bug in pair style lj/long/coul/long/intel

fix uninitialized data bug in pair style lj/cubic when mixed coefficients where specified explicitly

fix bug in write_coeff that could not correctly operate on full pathnames

fix a problem loading the shared library under MacOS after installing the LAMMPS python module with "make install-python"

adjust package list to avoid compilation failures when using the "most" preset

improve cythonize detection on machines using bourne shells different from bash (e.g. debian/ubuntu)

correct handling of make package-update for ML-IAP package

python 3 compatibility updates scripts in tools/python

try harder to find the stub version of libcuda.so as provided by the CUDA toolkit in case there is no CUDA driver installation

fixes to CMake build for ML-QUIP package when using CMake 3.16 and newer or recent Cray Programming Environment versions

Source code(tar.gz)
Source code(zip)
patch_24Mar2022(Mar 24, 2022)
Changes since the 17 February 2022 patch release:

Add computes for local components of the pressure tensor in cartesian and spherical coordinates (Olav Galteland, Norwegian U of Science and Technology) PR #3165

Major update of the LATBOLTZ package (Colin Denniston and collaborators, UWO) PR #3167

Code optimization for several KOKKOS styles (Nicholas Curtis and Leopold Grinberd and Gina Sitaraman, AMD, Stan Moore, SNL) PR #3147, PR #3158, PR #3169

Add fixes for damping of rotational and translational energy for extended particles (Jibril B. Coulibaly. Northwestern U) PR #3181

Add support for an inner cutoff switching function to SNAP related commands that smoothly turns of SNAP descriptors at short range (Aidan Thompson, SNL) PR #3173

Refactor code for neighbor list requests so that no direct access to the list of requests is needed and settings are as automatic and consistent as possible (Axel Kohlmeyer, Temple U) PR #3159

Refactor handling of thermodynamic output for better readability and add new thermo_style "yaml" (Axel Kohlmeyer, Temple U) PR #3179

Replace usage of non-reentrant strtok() and improve file reading and parsing using file reader and tokenizer classes (Axel Kohlmeyer, Temple U) PR #3180

Memory usage optimization and bugfix for mesocnt pair style (Philipp Kloza, Cambridge U) PR #3162

Refactor python/install.py script and change installation of the LAMMPS python module via make install-python from using distutils to setuptools. This is now a 2-step process of first creating a binary wheel (.whl) package and then installing it via pip.

Automatically load plugins in folders listed in the LAMMPS_PLUGIN_PATH environment variable (Axel Kohlmeyer, Temple U) PR #3170

Add tests for various compute (Axel Kohlmeyer, Temple U) PR #3163

More content and improvements for the Programmer Guide section of the Manual (multiple authors) PR #3146

update sphinx theme for the HTML version of the manual (Axel Kohlmeyer, Temple U) PR #3178

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #3138, PR #3139, PR #3141, PR #3142, PR #3145, PR #3148, PR #3153, PR #3155, PR #3156, PR #3161, PR #3164, PR #3166, PR #3168, PR #3171, PR #3174, PR #3177, PR #3185

Backward compatibility notices:

The new version of make install-python will not install a package over an existing package as the old version did. Instead, it will attempt to uninstall the old package and then install the new version. Packages installed with distutils, like the old version of make install-python, must be manually uninstalled (they don't store a suitable manifest). This should only be needed once, since from then on uninstall should be possible.

The update to the LATBOLTZ package is a significant update that changes the syntax of the fix lb/fluid command in a way that is not backward compatible. Inputs must be updated accordingly. Also fixes lb/pc and lb/pc/rigid/sphere have been removed and fixes nve and rigid should be used in their stead.

The thermodynamic output is slightly changed. This may break parser implementations for analysis of LAMMPS log files that are not indifferent to whitespace changes and width of fields. Please note the addition of thermo style "yaml" that should provide a mechanism that should make the process of extracting thermodynamic output data more reliable and easier to realize.

Source code(tar.gz)
Source code(zip)
patch_17Feb2022(Feb 17, 2022)
Changes since the 7 January 2022 patch release:

Add new pair style harmonic/cut providing a repulsive-only harmonic potential (Axel Kohlmeyer, Temple U) PR #3087

Add new fix numdiff/virial for deriving virial stress from potential energy via numerical differences (Aidan Thompson, SNL and Charlie Sievers, UC Davies) PR #3105

Add two new interlayer pair styles ilp/tmd and saip/metal (Wengen Ouyang, Wuhan University) PR #3125

Refactor source code to replace use of the virtual keyword in derived classes with the equivalent override which allows to detect some common problems at compile time (Richard Berger, Temple U) PR #3090

new option for multi-file dumps to have files that are more evenly sized to address post-processing and visualization issues for simulations of very large systems (Stan More, SNL) PR #3097, PR #3110

accelerator support (including KOKKOS) and improvements for dynamical_matrix and third_order commands (Charlie Sievers, UC Davis) PR #2233

performance improvements and added unit tests for oxdna atom style (Oliver Henrich, Strathclyde U) PR #3104

Portability updates for PyLammps (Richard Berger, Temple U) PR #3115

re-implementation of the tools/eam_database Fortran code in Python (Germain Clavier, TU Eindhoven) PR #3126

port bundled unit tests and external OpenCL ICD loader to be compiled and run on Windows using the MSVC toolchain (Axel Kohlmeyer, Temple U) PR #3102, PR #3116, PR #3127

update force style unit tests to work with Kokkos + OpenMP (Richard Berger, Temple U) PR #3035

update setup.py in python module to use setuptools instead of deprecated distutils (Richard Berger, Temple U) PR #3101

various optimizations, corrections, refactoring, and compatibility updates for the KOKKOS package (Stan Moore, SNL and multiple collaborators) PR #3082, PR #3088, PR #3092, PR #3100, PR #3111

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #3089, PR #3091, PR #3095, PR #3103, PR #3117, PR #3120, PR #3123, PR #3124, PR #3129, PR #3130, PR #3133, PR #3134, PR #3136

Backward compatibility notices:

Using the PyLammps python module now requires the presence of numpy

Source code(tar.gz)
Source code(zip)
stable_29Sep2021_update2(Jan 7, 2022)
This is the second update to the 29 September stable release that fixes bugs that have been reported since the first update and backported from the develop branch.

The following individual changes are included:

fix bug in compute temp/rotate due to uninitialized data cause by a class member shadowing a base class member

plug minor memory leaks in MACHDYN package

plug memory and file descriptor leak when reading ReaxFF potential files

fix internal fix and compute ID clash when using fix reaxff/species multiple times

fix segmentation fault when using atom style smd as part of a hybrid pair style

fix communication bug in pair style local/density

fix segmentation fault due to an uninitialized array in fix nve/gpu when the group is not "all" but there is only one atom type

fix bugs in eam/cd pair style reported in issue #3074

correct documentation and potential parameters for Lebedeva interlayer potential

update names and description of git branches in the manual to reflect the current status and policies

avoid issues when testing dump commands in parallel

avoid issues from oversubscribing processors with parallel tests using MPI

correct error in the documentation for pair style granular

fix bug in cmake parsing of the version.h file for stable branch. Fixes #3038

fix bug in bond/angle style table reported in PR #3076 (but fixed differently)

Source code(tar.gz)
Source code(zip)
bugfixes_for_stable_29Sep2021_update2.patch.gz(42.65 KB)
patch_7Jan2022(Jan 7, 2022)
Changes since the 14 December 2021 patch release:

New compute ave/sphere/atom to compute averaged properties in a sphere around atoms (Stan Moore with Mitch Wood, SNL) PR #2996

Support writing to dump files at regular time interface for variable timestep simulations (Michal Kanski, Jagiellonian U, and Steve Plimpton, SNL) PR #3052

Add centroid atomic stress support for shake, rattle and rigid/small fixes (Donatas Surblys, Tohoku U) PR #3046

Update fmtlib to version 8.1.1 (Axel Kohlmeyer and Richard Berger, Temple U) PR #3073

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #3021, PR #3060, PR #3062, PR #3064, PR #3065, PR #3068, PR #3069, PR #3071, PR #3072

Backward compatibility notices:

Forces and energies from class2 angle styles may differ due to a bugfix. Previously distances from the BondBond terms were used for BondAngle terms. Commonly those distance settings would be the same and in those cases the results will match.

Source code(tar.gz)
Source code(zip)
patch_14Dec2021(Dec 14, 2021)
Changes since the 27 October 2021 patch release:

Update bundled copy of the Kokkos library to version 3.5.0 (Stan Moore, SNL and the Kokkos developers) PR #3019, PR #3053

Allow GPU package pair styles to be used with newton_pair on (Trung Nguyen, Northwestern) PR #2867

New update to delete_atoms porosity so that atoms can now be selected by region and group (Steve Plimpton, SNL) PR #2984

New pair style sw/mod providing an empirically modified Stillinger-Weber potential suitable for compounds like MoS2 (Jin-Wu Jiang, Shanghai University, Wengen Ouyang, Wuhan University) PR #3041

New pair style nm/cut/split and bond style fene/nm for coarse grain polymer models (Joseph Dietz, Robert Hoy, USF) PR #2958

Support for binary native dump files for read_dump and rerun (Lixin Sun, Harvard) PR #3054, PR #3059

Update included fmtlib sources to version 8.0.1 (Richard Berger and Axel Kohlmeyer, Temple U) PR #2809

Refactoring of access to lists of computes, fixes and regions that do not require access to internal data structures (Axel Kohlmeyer, Temple U) PR #3001

Refactoring of data file section reading and writing for fixes for more consistency and simpler, more intuitive input (Axel Kohlmeyer, Temple U) PR #3012

Updated singularity container definition files for building LAMMPS (Richard Berger, Temple U) PR #3043

Added labels "slow" and "unstable" to some unit tests run with ctest so they can be easily excluded (Richard Berger, Temple U) PR #3033

Improved support for multi-configuration builds (required on Windows) and passing settings to external projects (Axel Kohlmeyer, Temple U) PR #2940

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #3011, PR #3013, PR #3014, PR #3015 , PR #3017, PR #3020, PR #3027, PR #3032, PR #3034, PR #3040, PR #3044, PR #3049, PR #3055, PR #3056, PR #3058

Backward compatibility notices:

delete_atoms porosity requires a group as an additional argument

The CC.Lebedeva interlayer potential file has been updated. It now uses by default parameters for "normal" conditions. The previous default parameters are tuned for high-pressure conditions and available when using "C1" instead of "C" as element.

Source code(tar.gz)
Source code(zip)
stable_29Sep2021_update1(Nov 9, 2021)
This is an update to the 29 September stable release that fixes some bugs that have been reported since then and addresses some issues related to compatibility with GitHub, Python and Python package versions used to build the manual, and running the automated unit tests.

The following individual changes are included:

Corrections to double precision support detection for the GPU package. Only report compatible GPUs. Also consider single-precision FFT requirement for pppm when running unit tests.

Fix bugs leading to deadlocks when loading INTERLAYER package potential files in parallel

Correct thermo output when expanding per-atom properties

Fix compilation issue for QM/MM front end code

Fix bugs in dump style custom when using custom per-atom data for thresholds

Update VTK package to compile and work with custom per-atom arrays

plug some minor memory leaks

fix some bugs in the KOKKOS package

update required python package versions for building the documentation to be compatible with python 3.10

fix memory allocation bug in angle and bond style gaussian leading to memory corruption on 32-bit machines

fix false positives in python unit tests when LAMMPS is compiled with an external KOKKOS library

update the citation info for the new LAMMPS paper

change references to git:// protocol to use https:// protocol instead due to heightened security requirements at GitHub

document size limitations when sorting dumps

avoid false positives in unit testing due to numerics and oversubscribing of MPI processes

Source code(tar.gz)
Source code(zip)
bugfixes_for_stable_29Sep2021_update1.patch.gz(80.91 KB)
patch_27Oct2021(Oct 28, 2021)
Important note about branch names

The LAMMPS project has changed the names of its main branches in the git repository to better represent the state of the code. The master branch was renamed to develop and the unstable branch became release. If you are following the git repository to obtain the LAMMPS source code, please update your checkouts and forks accordingly. For a limited time the master and unstable branches will be kept in sync with develop and release (through fast forwarding), but those branches will be removed eventually and then you will be required to use the new branch names or not have access to updates anymore.

Changes since the 29 September 2021 patch release:

New fix acks2/reaxff for ACKS2 charge equilibration and updates to fix qeq/reaxff that add support for fix efield to both charge equilibration methods. (Metin Aktulga, MSU and Stan Moore, SNL) PR #2931

New fix mol/swap for swapping of atom types within a single, randomly selected molecule (Steve Plimpton, SNL) PR #2992

Add support for multi-species potentials to ML-PACE package (Yury Lysogorskiy and the PACE library developers) PR #2968

Preliminary support for OpenMPTarget in KOKKOS package (Rahul Gayatri, NERSC, Chris Knight, ANL, Stan Moore, SNL) PR #2997

Implemented and applied a new namespace platform with abstractions of functionality that has either different behavior or uses different APIs on different platforms and operating systems. This reduces code redundancy and enhances platform neutrality of the core LAMMPS code. (Axel Kohlmeyer, Temple U) PR #2979

Implemented and applied a utility function to parse boolean arguments to keywords (yes/no, on/off, true/false) (Axel Kohlmeyer, Temple U), PR #2951

Semi-automatic code refactoring using clang-tidy. See issue #2946 for a discussion of individual changes. (Axel Kohlmeyer, Temple U) PR #2990, PR #2994, PR #3002

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #2973, PR #2975, PR #2976, PR #2977, PR #2983, PR #2985, PR #2989, PR #2991, PR #2993, PR #2998, PR #3003, PR #3004, PR #3007 , PR #3009

Backward compatibility notices:

Most keywords that accept "yes/no" or "on/off" parameters will now accept "yes/no", "on/off", or "true/false" as valid arguments. This is not (yet) reflected in the manual. Some cases where there are more than just two choices have been adapted accordingly, but not all. Backward compatibility is retained. All tests pass.

Source code(tar.gz)
Source code(zip)
stable_29Sep2021(Sep 30, 2021)
Below is a list of major changes since the last stable release 29 October 2020

Same as the last time, this stable release contains over 4000 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.

The LAMMPS homepage has been moved from https://lammps.sandia.gov to https://www.lammps.org which is now hosted at Temple U. This lifts bandwidth and access limitations and addresses some reliability issues.

General changes:

Continued refactoring of the core LAMMPS code and many packages to increase code reuse, simplify new additions, improve consistentcy, and benefit from C++11 features. Several convenience functions and classes were added in the process.

Reorganization of packages: the "USER" prefix has been dropped, some packages have prefixes to group them (CG-, DPD-, EXTRA-, ML-) and the contents of the USER-MISC package have been moved to appropriate existing or new packages. There are now EXTRA-PAIR, EXTRA-COMPUTE, EXTRA-FIX, EXTRA-DUMP, EXTRA-MOLECULE packages for the less commonly used styles. We recommend that external packages (i.e. those not included in the LAMMPS distribution) adopt a USER- prefix to make them recognizable as external package.

Refactored multi-cutoff neighbor list support resulting in significant performance improvements for polydisperse granular systems.

Continued improvements to the build systems, especially the CMake based version.

Refactored the handling of embedded citation strings to make them more visible.

Updates and additions to the programmer guide section of the manual including updated guidelines and information for contributing code to LAMMPS.

Add support for using clang-format in the src tree with a configuration file and suitable comments in the source files. This is recommended for new contributions and will be gradually applied to existing .cpp files as the reformatting can interfere with the readability in some cases.

Scripts to support building LAMMPS and its documentation via CMake on systems without direct internet access.

Updates and new commands or styles or packages:

Expanded support for machine learning potentials in the ML-IAP package and the added ML-HDNNP, ML-PACE, ML-RANN packages

New MDI package to allow client/server mode coupling of LAMMPS to other code using the MolSSI Driver Interface API

New BROWNIAN package for Brownian dynamics time integration of point and extended particles plus updates for fix propel/self

New DIELECTRIC package with boundary element solvers for computing induced charges at interfaces

Major refactoring of the REAXFF package to eliminate dead code and inconsistencies, modernize force field file parsing, and correctly handle computing of the virial stress under different settings and with OpenMP

Updates to the library interface and corresponding changes to the LAMMPS python module: bug fixes, additional introspection support and support for setting callbacks for fix external and other associated properties with C and Python callbacks/functions possible.

Updated GPU package with improved support for OpenCL which enables using the GPU package with GPUs embedded in Intel processors (Iris, UHD, and HD graphics using the "NEO" OpenCL runtime).

Diverse updates, performance improvements, and bugfixes for the following packages: DRUDE, PLUMED, COLVARS, KOKKOS, KIM, ML-IAP, SNAP, REACTION

Many bugfixes, updates, and individual contributed commands and styles.

Backward compatibility notes:

The LAMMPS python module is now a folder with multiple Python script files instead of a single lammps.py file. This should not affect how the module can be used "in-place" by setting environment variables. The python module contains a version check that should help detecting when a LAMMPS shared library that does not match the python module.

Multi-cutoff neighbor list and communication is using a new and more flexible code path with a few additional constraints over the old code path, which is still available as a backup that must be explicitly requested.

KIM package commands have been reorganized into a single wrapper and multiple sub-commands: kim_<something> has become kim <something>

The CMake preset "minimal" has been renamed to "basic". There are no more "make yes/no-standard" and "make yes/no-user" due to the package reorganization; instead "make yes/no-basic" and a "make yes/no-most" targets have been added to reflect the corresponding CMake presets

Source code(tar.gz)
Source code(zip)
bugfixes_for_stable_29Sep2021.patch.gz(99.24 KB)
patch_29Sep2021(Sep 30, 2021)
This patch includes some final changes and bugfixes to create the stable release.

There are no substantial new features in this patch, only a variety of small updates and bug fixes for reported issues and based on extended testing with instrumented code and on platforms or with settings we do not run automatic tests for. Notable changes:

bugfix for EAM/alloy pair style with KOKKOS on GPUs (Stan Moore, SNL) PR #2956

bugfix for hybrid pair styles not enforcing a single pair_coeff statements for potentials that require it (Axel Kohlmeyer, Temple U) PR #2963

logging of queried KIM collections locations with kim init (Ryan Elliott, UMN) PR #2949

added section on parallel algorithms to manual adapted from new LAMMPS paper, PR #2917

small bugfix for COLVARS package (Jerome Henin and Giacomo Fiorin) PR #2959

small bugfix for handling of nevery keyword in fix bond/react (Jake Gissinger, NASA) PR #2969

bugfix to neighbor list: prevent it from copying "unique" stencil/bin data (Stan Moore and Steve Plimpton, SNL) PR #2965

updates to traditional make build system to enable C++11 compliance by default, list packages in alphabetical order and make package maintenance commands less verbose (Axel Kohlmeyer, Temple U) PR #2961

collected small changes (Axel Kohlmeyer, Temple U) PR #2952, PR #2962, PR #2966

improved detection of double precision hardware with OpenCL and a workaround for neighbor list issues on sparse systems in the GPU package

improved C++ standard compliance for improved compatibility with compilers in C++20 mode

improved C library standard compliance by avoiding calling memset() and fwrite() with a null pointer

fix bug in traditional make build system support for ML-PACE package

add missing documentation for building with threaded FFTW3 support using non-standard library locations

plug memory leaks and compilation issue with 64-bit atom ids in MSCG package

avoid division by zero in fix dt/reset

fix portability issues and detect incompatibilities when compiling external libraries using CMake

prohibit KOKKOS from lowering the requested C++ standard

Source code(tar.gz)
Source code(zip)
patch_20Sep2021(Sep 21, 2021)
This is the release candidate for the next stable release. We will only accept urgent bugfixes and updates to documentation until then. For people packaging stable LAMMPS releases into binary packages, please note that you have about one week left to submit any changes to LAMMPS where you have discovered issues related to building and packaging that we have overlooked. Once the stable release is out, you will have to maintain such changes as external patches until the next stable release or package a following patch release with those changes included.

Changes since the 31 August 2021 patch release:

Updates and additions to the programmer guide section of the manual. Specifically, updated guidelines and information for contributing code to LAMMPS (Axel Kohlmeyer, Temple U) PR #2927

Add new command line flag -skiprun testing input files for syntax error as this will skip main loops in run, minimize and similar commands PR #2919

Avoid crashes when loading modules like NumPy multiple times into deleted/cleared LAMMPS instances. For that Py_Finalize() needs to be treated like MPI_Finalize() and thus called from main.cpp or a wrapper function in the library interface (Axel Kohlmeyer and Richard Berger, Temple U) PR #2923

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #2919, PR #2924, PR #2925, PR #2929, PR #2935, PR #2939, PR #2941

Source code(tar.gz)
Source code(zip)
patch_31Aug2021(Aug 31, 2021)
Changes since the 30 July 2021 patch release:

Add KOKKOS accelerated styles for shear flow simulations with the nvt/sllod integrator (Emily Kahl, The University of Queensland) PR #2882

New fix ttm/grid with improved file handling and distributed grid. The existing ttm fixes are refactored and modernized. This also generalizes and unbundles the grid communication class from the KSPACE package and thus updates all PPPM and MSM kspace style variants accordingly (Steve Plimpton, SNL) #2902

New pair style coul/exclude to remove excluded Coulomb interactions to be used with kspace style scafacos for molecular systems (Axel Kohlmeyer, Temple U) PR #2886

Add functionality to fix property/atom to support per-atom arrays in addition to per-atom vectors (Steve Plimpton SNL) PR #2342, PR #2889 PR #2903

Update the COLVARS package to version 2021-08-06 (Giacomo Fiorin, NIH, and the colvars developers) PR #2863 and PR #2866

Update to fix bond/swap to support a second kind of bond swapping model (Steve Plimpton, SNL) PR #2888

File reader refactor, bug fixes, and unit tests for INTERLAYER package pair styles (Axel Kohlmeyer, Temple U) PR #2879, PR #2897

Optimizations for SNAP with KOKKOS for large systems (enabled via config option) (Evan Weinberg, Nvidia) PR #2905

Various updates and bugfixes for the KOKKOS package (Stan Moore, SNL) PR #2890, #2900

Update factory functions to create or replace fixes and computes to return a pointer to the created class instance (Axel Kohlmeyer, Temple U) PR #2869

Updates to the introductory pages of the LAMMPS documentation (Axel Kohlmeyer, Temple U) PR #2893

Many small bug fixes, minor code improvements, and coding style, build system, documentation, formatting updates (multiple authors) PR #2804, PR #2862, PR #2861, PR #2865, PR #2868, PR #2876, PR #2880, PR #2881, PR #2885, PR #2887, PR #2889, PR #2891 , PR #2895, PR #2896, PR #2904, PR #2908, PR #2909, PR #2910, PR #2911, PR #2912, PR #2913, PR #2914

Backward compatibility notices:

MC package fixes atom/swap, gcmc, widom, and charge/regulation were under some circumstances calling fixes that normally would be called at the "end_of_step()" stage during a time step instead in the "pre_exchange()" stage. This could lead to unexpected output or use of data from the previous time step (e.g. with fix print, fix ave/time etc.). These calls have been removed and the expected behavior has been restored.

the output file format for the fix ttm electron temperature grid files has changed

Source code(tar.gz)
Source code(zip)
patch_30Jul2021(Jul 30, 2021)

Changes since the 28 July 2021 patch release:

This patch release contains a fix for a serious bug reported in issue #2858 that was present in LAMMPS since the 10 February 2021 patch release. We urge everybody that are using this version or any other versions released since then to upgrade to this patch release.

In addition, this patch release contains mostly minor code refactoring changes and a few fixes for issues were either a result of the recent package reorganization or were exposed by it. PR #2856, PR #2857, and PR #2860
Source code(tar.gz)
Source code(zip)
patch_28Jul2021(Jul 28, 2021)
Changes since the 2 July 2021 patch release:

Completion of the package reorganization started in the 2 July 2021 release (Axel Kohlmeyer, Temple U) PR #2829

Significant refactor of the REAXFF package and some updates to the QEQ package and corresponding code in the OPENMP and KOKKOS packages to have it better aligned with LAMMPS coding conventions, remove obsolete and dead code and implement missing features like support for compute stress/atom in the OpenMP version of the reaxff pair style. (Axel Kohlmeyer, Temple U, Aidan Thompson, SNL) PR #2708

New fix pair/tracker and pair style tracker to track properties of pair interactions (Joel Clemmer, SNL) PR #2788

New lammps_gather_bonds() function in the C library interface and corresponding Python module wrappers (Axel Kohlmeyer, Temple U, and Robert Meißner, TU Hamburg) PR #2850

Updates to the CG-DNA package including a new atom style to address issues with tracking contact pairs (Oliver Henrich, Strathclyde U) PR #2827 and PR #2835

Updates to the INTEL package to better support the new LLVM based Intel compilers and transition to use OpenMP pragmas for SIMD (Mike Brown, Intel) PR #2842 and PR #2849

Add support for atom_modify map hash in the KOKKOS package (Stan Moore, SNL) PR #2815

Updates to the C library interface and Python module to support all fix external functionality and make those behave more consistently, also document the calls (Axel Kohlmeyer, Temple U) PR #2837

A new compute for the TALLY package to measure virial heat flow from one group of atoms to another (Donatas Surblys, Tohoku U) PR #2841

Address issues and inconsistent behavior when using pair_modify nofdotr and compute stress/atom for pair styles in the MANYBODY and OPENMP packages. Add functionality to the unit tests to check for consistency in energy and global pressure (Axel Kohlmeyer, Temple U) PR #2736

Various small bug fixes, coding style, build system, documentation and formatting updates (multiple authors) PR #2823, PR #2826, PR #2824, PR #2828, PR #2830, PR #2831, PR #2832, PR #2833, PR #2836, PR #2838, PR #2839, PR #2844, PR #2848, PR #2853, PR #2854, #2855

Backward compatibility notices:

This patch release completes the package reorganization from the 2 July 2021 release. The USER-MISC package has been removed and its contents distributed to existing packages or one of the new packages INTERLAYER, ORIENT, EXTRA-COMPUTE, EXTRA-DUMP, EXTRA-FIX, EXTRA-MOLECULE, EXTRA-PAIR. Also some styles from MISC, MOLECULE and the src folder have been moved to more suitable packages. In case you are missing a style in your executable, LAMMPS should report which package contains the missing style and needs to be added.

Some semantics of the fix external support functions to set global energy and virial have been changed to be more consistent.

Source code(tar.gz)
Source code(zip)
patch_2Jul2021(Jul 2, 2021)
Changes since the 27 May 2021 patch release:

New package DIELECTRIC with boundary element solvers for computing induced charges at interfaces (Trung Nguyen and Monica Olvera del la Cruz, Northwestern) PR #2793

New SO3 descriptor for the MLIAP package (Byungkyun Kang and Qiang Zhu, UNLV) PR #2796

New fabric tensor compute (Joel Clemmer, SNL and Ishan Srivastava, LBL) PR #2795

Variable pH support and bugfixes for Ewald self-energy in fix charge/regulation (Tine Curk, Northwestern) PR #2802

Correctly handle KOKKOS package initialization with the library interface and the fact that the Kokkos library may only be initialized once (Stan Moore, SNL) PR #2470

GPU package fixes for compilation with the ROCm/HIP backend (Richard Berger, Temple U) PR #2789

Updates to Python package and examples (Richard Berger, Temple U) PR #2787

Various small bug fixes, coding style, build system, documentation and formatting updates (multiple authors) PR #2786, PR #2790, PR #2798, #2803, #2810, #2813, #2817

Backward compatibility notes:

The LAMMPS home page has been moved to https://www.lammps.org

To reflect changes in the development process and the overall code quality we no longer differentiate between "standard" and "user packages. Thus many packages have been renamed. There are no more "user" packages included in LAMMPS except for USER-MISC. Also some packages have been renamed to better represent their content and also to group similar ones (e.g. machine learning or coarse grain or DPD packages). Some more reorganization of individual style files will follow in the next patch release.

There is a new package (DPD-BASIC) and several dpd pair styles have been moved there from USER-MISC or the src folder.

The "minimal.cmake" preset has been renamed to "basic.cmake" to match the corresponding "yes/no-basic" target in the conventional build and to avoid conflicts.

Source code(tar.gz)
Source code(zip)
patch_27May2021(May 27, 2021)
Changes since the 14 May 2021 patch release:

New package USER-HDNNP which provides a new pair style hdnnp, which is an interface to the N2P2 library for Behler-Parinello neural network potentials (Andreas Singraber, U Vienna) PR #2626

New package USER-RANN which provides a new pair style rann and matching potential files for using "rapid atomistic neural network (RANN) potentials (Christopher Barrett and Doyl Dickel, Mississippi State) PR #2570

Update of the bundled Kokkos library to version 3.4.1 (Stan Moore, SNL and the Kokkos developers) PR #2783

Update references to the LAMMPS homepage to reference www.lammps.org and docs.lammps.org instead of lammps.sandia.gov and add tool to check for it and convert such links, also corresponding updates to the LAMMPS manual itself. (Axel Kohlmeyer and Richard Berger, Temple U) PR #2778, PR #2768

Various small bug fixes, coding style and formatting updates (multiple authors) PR #2640, PR #2770, PR #2771, PR #2772, PR #2773, PR #2775, PR #2776, PR #2777, PR #2781, PR #2782, PR #2784

Backward compatibility notes:

the LAMMPS home page has been moved from https://lammps.sandia.gov to https://www.lammps.org (please report cases where links need to be updated).

Source code(tar.gz)
Source code(zip)
patch_14May2021(May 15, 2021)
Changes since the patch release 8 April 2021:

New package USER-PACE which provides a new pair style pace computing interactions using the Atomic Cluster Expansion (ACE) (Yury Lysogorskiy and collaborators at ICAMS, Ruhr-Uni Bochum) PR #2699

New Package USER-BROWNIAN for Brownian dynamics time integration of point and extended particles plus updates for fix propel/self (Sam Cameron, U Bristol) PR #2511

New Package USER-MDI to enable client/server mode coupling of LAMMPS with other codes using the MolSSI Driver Interface (MDI) (Taylor Barnes, MolSSI) PR #2611

New fix style "charge/regulation" which implements Monte Carlo sampling of charge regulation (Tine Curk, Northwestern U and Jiaxing Yuan, Shanghai Jiao Tong U) PR #2585

New generalized DPD pair styles: dpd/ext and dpd/ext/tstat (Martin Svoboda, Karel Sindelka, and Martin Lisal, ICPF) PR #2710

Faster and more flexible support for multi style neighbor lists (Joel Clemmer, SNL) PR #2536

New pair style hybrid/scaled that allows combining contributions from different pair styles with constant or variable scaling factors (Axel Kohlmeyer, Temple U) PR #2702

Updated version of pair style bop with bugfixes and performance improvements (Xiaowang Zhou, SNL) PR #2750, PR #2753

New base class Command that all command styles are now derived from instead of the Pointers class (Axel Kohlmeyer, Temple U) PR #2707

GPU package version of pair style lj/smooth (Gurgen Melikyan, HSE University) PR #2706

Update of the bundled Kokkos library to version 3.4.0 (Stan Moore, SNL and the Kokkos developers) PR #2733

Support for atom style spin in the KOKKOS package (Julien Tranchida, Stan Moore, and Svetoslav Nikolov, SNL) PR #2693

New clang-format configuration file in src tree. All files in that tree are prepared so that clang-format can be applied without destroying functionality. (Axel Kohlmeyer, Temple U) PR #2762

Refactoring of core LAMMPS code and packages to increase code reuse and readability, simplify text parsing, improve thread safety (Richard Berger and Axel Kohlmeyer, Temple U) PR #2704, PR #2719, PR #2727, PR #2738, PR #2749

Bugfixes and updates for the GPU package addressing issues caused by the changes/improvements in PR #2603 (Mike Brown, Intel Corp., Trung Nguyen, Northwestern U, and Richard Berger, Temple U) PR #2711, PR #2720, PR #2723, PR #2744, PR #2745, PR #2756

More unit tests for MPI load balancing and file operations (Richard Berger and Axel Kohlmeyer, Temple U) PR #2642, PR #2725

Scripts to support building LAMMPS and its documentation on systems without direct internet access (Richard Berger, Temple U) PR #2742

Various small bug fixes, coding style and formatting updates (multiple authors) PR #2703, PR #2705, PR #2712, PR #2713, PR #2714, PR #2715, PR #2717, PR #2721, PR #2722, PR #2724, PR #2726, PR #2737, PR #2739, PR #2741, PR #2743, PR #2748, PR #2756, PR #2758, PR #2760, PR #2761, PR #2765, PR #2769

Backward compatibility notes:

using comm_modify multi now requires to also use neighbor multi. In case the new neighbor multi functionality causes problems, it is possible to revert to the previous implementation with neighbor multi/old. In that case comm_modify multi/old and comm_modify cutoff/multi/old need to be used for enabling and customizing multi-cutoff ghost atom communication.

binary restart files containing data for pair style coul/cut will be incompatible due to bugfixes (but they were broken anyway)

external packages that add command styles will have to be adapted to the changes to the class topology in PR #2707

Source code(tar.gz)
Source code(zip)
patch_8Apr2021(Apr 7, 2021)
Changes since the patch release 10 March 2021:

New PLUGIN package providing a "plugin" command that allows loading and unloading of LAMMPS styles from separately compiled shared object files at runtime (Axel Kohlmeyer, Temple U) PR #2650

Refactoring of core LAMMPS code and packages to increase code reuse and readability, simplify text parsing, simplify internal fix/compute generation, improve thread safety (Richard Berger and Axel Kohlmeyer, Temple U) PR #2655, PR #2657, PR #2674, PR #2678, PR #2683, PR #2685, PR #2694, PR #2700

Additional unit test tools and test inputs to increase coverage of automated testing (Richard Berger and Axel Kohlmeyer, Temple U) PR #2308, PR #2653, PR #2658, PR #2679, PR #2681, PR #2682, PR #2695

Updates and additions to PyLammps examples (Richard Berger, Temple U and Evangelos Voyiatzis, Royal DSM) PR #2662, PR #2664

Updates and bugfixes to the USER-REACTION package (Jake Gissinger, NASA) PR #2675

Updates and bugfixes to the KOKKOS package (Stan Moore, SNL) PR #2677

Updates and bugfixes to the GPU package (Mike Brown, Intel Corp. and Nicholas Curtis, AMD, and Richard Berger, Temple U) PR #2691, PR #2692

Improved error/exception detection and handling in the LAMMPS python module (Richard Berger, Temple U) PR #2668

Various small bug fixes, coding style and formatting updates (multiple authors) PR #2654, PR #2659, PR #2661, PR #2663, PR #2666, PR #2670, PR #2671, PR #2685, PR #2686, PR #2689, PR #2690, PR #2696, PR #2697, PR #2698

Backward compatibility notes:

there are more strict checks about the order of rigid fixes and fixes that change box dimensions. LAMMPS will now correctly error out where it previously would silently ignore some fixes.

checks for using run_style verlet versus run_style respa have been made consistent across the code and fully aware of suffixes

the meaning of the request ID argument for accessing neighbor lists through the library and python interface has changed. it will function the same for the typical case of having just a single neighbor lists, but see the manual for other cases.

Source code(tar.gz)
Source code(zip)
patch_10Mar2021(Mar 10, 2021)
Changes since the patch release 10 February 2021:

Major update of the GPU package with significant performance and compatibility improvements (Mike Brown and Nitin Dhamankar, Intel Corp.) PR #2603, PR #2610, use statically linked OpenCL driver loader with CMake (Axel Kohlmeyer, Temple U) PR #2617, HIP compilation fixes (Richard Berger, Temple U) PR #2622

New pair style lj/relres which allows fine-grained/coarse-grained simulations for lj/cut type interactions (Mark Chaimovich, Russian School of Mathematics, and Aviel Chaimovich, Drexel University) PR #2590

Refactoring of the KIM package commands. Instead of multiple kim_<something> commands there is now only the kim command which accepts <something> as subcommand. Additional cleanups and improvements to the kim commands (Yaser Afshar, Ryan S. Elliott, Ellad Tadmor UMN) PR #2620, PR #2625, PR #2649

Direct support for neural network (nn) models added to MLIAP package (Pedro Antonio Santos Florez, University of Nevada - Las Vegas) PR #2641

Multiple small improvements and bugfixes to the USER-REACTION package (Jake Gissinger, NASA) PR #2600, PR #2616, PR #2627

Reader classes for LAMMPS log files and fix ave/chunk output added to the LAMMPS Python module (Richard Berger, Temple U) PR #2631

Unit tester tools for dihedrals and impropers with inputs for most styles (Abhishek Shandilya, RPI and Axel Kohlmeyer, Temple U) PR #2290

Cleanup and small performance updates for SNAP support in KOKKOS on GPU (Evan Weinberg, Nvidia) PR #2599

Various small bug fixes, coding style and formatting updates (multiple authors) PR #2601, PR #2602, PR #2604, PR #2609, PR #2612, PR #2613, PR #2614, PR #2615, PR #2618, PR #2624, PR #2629, PR #2632, PR #2633, PR #2634, PR #2635, PR #2637, PR #2638, PR #2644, PR #2651, PR #2652

Backward compatibility notes:

The STUBS library is compiled with C++ so there are no symbol name collisions when a serial LAMMPS library and mpi4py are loaded into a Python interpreter at the same time

All KIM package commands of the kind kim_<command> are renamed to kim <command>

There are several changes to the package gpu command:

The Ngpu parameter to the package now accepts 0 (for automatic determination of number of GPUs)

The gpuID keyword now only takes 1 argument.

The device keyword is replaced with 'platform' and 'device_type'

The automatic OpenCL platform detection will choose the platform with the best GPU (in terms of compute cores)

OpenCL, device configuration parameters for NVIDIA, AMD, INTEL are automatically determined at runtime More details about this are in the manual

Since the changes to the GPU package were substantial, some remaining bugs and compatibility issues are to be expected

Source code(tar.gz)
Source code(zip)
patch_10Feb2021(Feb 11, 2021)
Changes since the patch release 24 December 2020:

Add support for machine learning potentials using python frameworks to the MLIAP package (Nicholas Lubbers, LANL, and Aidan Thompson, SNL) PR #2508

Add 'shift' keyword for multiple Tersoff pair style variants to allow adjusting lattice constants in the case of hetero materials (Wengen Ouyang, Tel Aviv University, Davide Mandelli, Istituto Italiano di Tecnologia, Aidan Thompson, SNL) PR #2514

New 'eam/he' pair style for modeling He bubbles in metals (Xiaowang Zhou, SNL) PR #2556

New Wang-Frenkel pair style ('wf/cut'), a smooth, short range Lennard-Jones like potential (Xipeng Wang and Simón Ramírez-Hinestrosa, Chinese Academy of Sciences, U Cambridge) PR #2435

New potential files in version 2 format for the AGNI pair style and support for reading/processing those potential files (James Chapman, Lawrence Livermore National Lab) PR #2479

Update of bundled Kokkos library to version 3.3.1 (Stan Moore, SNL, and the Kokkos developers) PR #2540

Optimizations for running SNAP potentials with KOKKOS on GPUs (Evan Weinberg, Nvidia) PR #2549

Add capability to fix bond/react to have reactions create atoms (Jake Gissinger, NASA) PR #2584

Standardize the way how contributions from fix styles to energy and virial are handled. Remove THERMO_ENERGY fix mask. (Steve Plimpton, SNL) PR #2560

Replace redundant 3-vector inline functions from pair styles and math_vector.h with calls to MathExtra namespace functions. Obsolete math_vector.h and math_complex.h header files that were using common c++ class names without a protecting namespace (Axel Kohlmeyer, Temple U) PR #2558

More granular control over communication for pair style and fixes with KOKKOS (Stan Moore, SNL) PR #2532

Refactor processing of embedded citation strings to make them more visible (Axel Kohlmeyer, Temple U and Steve Plimpton, SNL) PR #2545

Add introspection support for compile time settings of accelerator packages (Axel Kohlmeyer, Temple U), PR #2554

Add ArgInfo class for simpler processing of references to computes, fixes, variables and more. Refactor the code using those (Axel Kohlmeyer, Temple U) PR #2571

Add `utils::strdup()`` convenience function and apply it to many files to collapse repetitive multi-line code to a single line (Axel Kohlmeyer, Temple U) PR #2583

Detect non-ASCII characters in input and try to correct known cases where equivalent UTF-8 characters are used (Axel Kohlmeyer, Temple U) PR #2564

Add version check to the LAMMPS python module so that the version of the python module must match the version in the shared library. (Axel Kohlmeyer, Temple U) PR #2589

Change sign conventions for forward/reverse Fourier transforms in LAMMPS to be aligned with common conventions (Steve Plimpton and Stan Moore, SNL) PR #2506

Move definition of "INVOKED_*" bitmap constants to an enumerator in the compute.h header file for consistency (Axel Kohlmeyer, Temple U) PR #2572

Add to CI checks runs via GitHub Actions (instead of Jenkins on ci.lammps.org): Static code analysis for C++ and Python via CodeQL, and compilation and unit tests on MacOS. (Axel Kohlmeyer and Richard Berger, Temple U) PR #2401

Various small bug fixes, coding style and formatting updates (multiple authors) PR #2547, PR #2553, PR #2555, PR #2557, PR #2559, PR #2561, PR #2562, PR #2565, PR #2566, PR #2567, PR #2569, PR #2576, PR #2579, PR #2580, PR #2582, PR #2586, PR #2591, PR #2592, #2594, #2596

Backward compatibility notice:

the processing and meaning of fix_modify energy and fix_modify virial has been changed. It is now only available for fixes that contribute to the potential energy and the virial. The energy contributions from thermostats and similar fixes can be accessed with the thermo keyword ecouple and the conserved quantity (i.e. etotal + ecouple) is available as econserve. There are corresponding changes to the implementation, so external fixes will have to be adapted or they will no longer compile. All affected fixes in the LAMMPS distribution have been updated.

the LAMMPS python module now contains a version check: When installing the python module, the current LAMMPS version is stored in a __version__ module variable and upon initialization of a LAMMPS instance it is checked against the output from the lammps.version() method which returns the version of the loaded shared library. If they don't match an exception is raised. This helps to detect inconsistent installations of the shared library or module or incorrectly setting LD_LIBRARY_PATH or PYTHONPATH.

Arguments to LAMMPS commands that contain references to compute, fixes, or variables (c_ID, f_ID, v_ID with and without indices) are now more strictly checked for correctness. Inputs that are incorrect but have previously worked by chance will now cause failures.

the KOKKOS package option "cuda/aware" has been renamed to "gpu/aware" since Kokkos now also supports AMD GPUs via HIP.

this release has several "invasive" changes to core parts of LAMMPS, so there is an increased risk that something has been broken in a part of the code that is not (yet) covered by our automated tests (which currently cover about 1/3rd of the total code base).

Source code(tar.gz)
Source code(zip)