LAMMPS Patch Release 10 March 2021

LAMMPS Patch Release Overview

What is LAMMPS used for?

LAMMPS is a classical molecular dynamics (MD) code that models ensembles of particles in a liquid, solid, or gaseous state. It can model atomic, polymeric, biological, solid-state (metals, ceramics, oxides), granular, coarse-grained, or macroscopic systems using a variety of interatomic potentials (force fields) and boundary conditions. It can model 2D or 3D systems with only a few particles up to millions or billions.

LAMMPS can run on single processor laptops or desktops, but is designed for parallel computers using message-passing techniques and a spatial-decomposition of the simulation domain. This includes shared-memory boxes and distributed-memory clusters and supercomputers. Many of its models have versions that provide accelerated performance on CPUs, GPUs, and Intel Xeon Phis. The code is designed to be easy to modify or extend with new functionality.

Interested in getting faster results?
Learn more about LAMMPS Certified GPU Workstations and Servers starting at $4,300

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/reres 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.

This release has 2 assets:

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

Visit the release page to download them.

About LAMMPS

LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) 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).