intel/intel-optimized-lammps repository overview

⁠Introduction

Stack image for LAMMPS⁠ (Large-scale Atomic/Molecular Massively Parallel Simulator) based on Intel HPC Toolkit⁠ (ver. ubuntu22.04⁠) and LAMMPS package⁠ (ver. 28Mar2023⁠).

⁠Configuration parameters

In order to modify the test execution, add the parameters from the list below, as environment variables. Otherwise, default values will be executed.

• LAMMPS_MODEL: the model to be executed. Default: in.intel.airebo. See list of valid models.
• OMP_NUM_THREADS: the number of Open MP threads to use. Default: auto, the detection results in using the number of cores per socket of the machine.
• GATED: whether to run a shortened version of the benchmark, the default is to run a long version and any non-empty string will cause the short version to run.
• USE_PPN: Whether to use PPN as an argument to mpiexec, the default is to not use it and any non-empty string will cause it to be used.
• USE_LRT: Whether to use LRT as an argument to the LAMMPS binary, the default is to not use it and any non-empty string will cause it to be used.
• NUM_PROCESSES: the number of MPI processes to use. Default: auto, the detection results in using the product of the cores per socket and the number of sockets.
• MULTIPLIER: the multiplier value to use to scale the length of the test. This value overrides the GATED option.

⁠Valid models

• in.intel.airebo
• in.intel.dpd
• in.intel.eam
• in.intel.lc
• in.intel.lj
• in.intel.rhodo
• in.intel.sw
• in.intel.tersoff
• in.intel.water

⁠System Requirements

Intel® AVX-512 support.