\(\renewcommand{\AA}{\text{Å}}\)

fix atom_weight/apip command

Syntax

fix ID group-ID atom_weight/apip nevery fast_potential precise_potential lambda_input lambda_zone group_lambda_input [no_rescale]

  • ID, group-ID are documented in fix command

  • atom_weight/apip = style name of this fix command

  • nevery = perform load calculation every this many steps

  • fast_potential = eam or ace for time measurements of the corresponding pair_style or float for constant time

  • precise_potential = ace for a time measurement of the pair_style pace/apip or float for constant time

  • lambda_input = lambda/input for a time measurement of pair_style lambda/input/apip or float for constant time

  • lambda_zone = lambda/zone for a time measurement of pair_style lambda/zone/apip or float for constant time

  • group_lambda_input = group-ID of the group for which lambda_input is computed

  • no_rescale = do not rescale the work per processor to the measured total force-computation time

Examples

fix 2 all atom_weight/apip 50 eam ace lambda/input lambda/zone all
fix 2 all atom_weight/apip 50 1e-05 0.0004 4e-06 4e-06 all
fix 2 all atom_weight/apip 50 ace ace 4e-06 4e-06 all no_rescale

Description

This command approximates the load every atom causes when an adaptive-precision interatomic potential (APIP) according to (Immel) is used. This approximated load can be saved as atomic variable and used as input for the dynamic load balancing via the fix balance command.

An adaptive-precision potential like eam/apip and pace/apip is calculated only for a subset of atoms. The switching parameter that determines per atom, which potential energy is used, can be also calculated by pair_style lambda/input/apip. A spatial switching zone, that ensures a smooth transition between two different interatomic potentials, can be calculated by pair_style lambda/zone/apip. Thus, there are up to four force-subroutines, that are computed only for a subset of atoms and combined via the pair_style hybrid/overlay. For all four force-subroutines, the average work per atom is be measured per processor by the corresponding pair_style. This fix extracts these measurements of the pair styles every nevery steps. The average compute times are used to calculates a per-atom vector with the approximated atomic weight, whereas the average compute time of the four subroutines contributes only to the load of atoms, for which the corresponding subroutine was calculated. If not disabled via no_rescale, the so calculated load is rescaled per processor so that the total atomic compute time matches the also measured total compute time of the whole pair_style. This atomic weight is intended to be used as input for fix balance:

variable nevery equal 10
fix weight_atom all atom_weight/apip ${nevery} eam ace lambda/input lambda/zone all
variable myweight atom f_weight_atom
fix balance all balance ${nevery} 1.1 rcb weight var myweight

Furthermore, this fix provides the over the processors averaged compute time of the four pair_styles, which are used to approximate the atomic weight, as vector.

Restart, fix_modify, output, run start/stop, minimize info

No information about this fix is written to binary restart files. None of the fix_modify options are relevant to this fix.

This fix produces a per-atom vector that contains the atomic weight of each atom. The per-atom vector can only be accessed on timesteps that are multiples of nevery.

Furthermore, this fix computes a global vector of length 4 with statistical information about the four different (possibly) measured compute times per force subroutine. The four values in the vector are as follows:

  1. average compute time for one atom using the fast pair_style

  2. average compute time for one atom using the precise pair_style

  3. average compute time of lambda/input/apip for one atom

  4. average compute time of lambda/zone/apip for one atom

The compute times are computed as average of all processors that measured at least one computation of the corresponding style. The vector values calculated by this fix are “intensive” and updated whenever the per-atom vector is computed, i.e., in timesteps that are multiples of nevery.

The vector and the per-atom vector can be accessed by various output commands.

No parameter of this fix can be used with the start/stop keywords of the run command. This fix is not invoked during energy minimization.

Restrictions

This fix is part of the APIP package. It is only enabled if LAMMPS was built with that package. See the Build package page for more info.

Default

no_rescale is not used by default.

(Immel) Immel, Drautz and Sutmann, J Chem Phys, 162, 114119 (2025)