LAMMPS 专题

LAMMPS 键势与 Kokkos 操作

用实际完成的 LAMMPS 输入展示参数、提交和结果文件,附各条件的操作记录。

更新于 2026-11-18约 12 分钟阅读按任务与环境选择适用配置
阅读本篇目录

下载输入、Slurm 作业与实际结果。本例使用 LAMMPS 30 Sep 2026 的 P100 构建。附件保存 48 个已完成条件,按输入保留相互作用与操作参数。

选择条件#

解压并进入 lammps-all-bond-kokkos-operations。CASES 每行对应一个作业数组索引,inputs 保存条件输入,reference 保存实际输出。gpu 条件使用 GPU 构建,其中普通 CPU 样式按输入直接运行;kokkos 条件启用 Kokkos。

条件目录 主要样式 运行模式
v100-quartic_exp bond_style quartic/exp/kk, pair_style zero kokkos
v100-gaussian bond_style gaussian/kk, pair_style zero kokkos
v100-morse bond_style morse/kk, pair_style zero kokkos
v100-gromos bond_style gromos/kk, pair_style zero kokkos
v100-class2 bond_style class2/kk, pair_style zero kokkos
v100-fene bond_style fene/kk, pair_style zero kokkos
v100-harmonic bond_style harmonic/kk, pair_style zero kokkos
v100-nonlinear bond_style nonlinear/kk, pair_style zero kokkos
v100-fene_nm bond_style fene/nm/kk, pair_style zero kokkos
v100-harmonic_restrain bond_style harmonic/restrain/kk, pair_style zero kokkos
v100-mm3 bond_style mm3/kk, pair_style zero kokkos
v100-harmonic_shift_cut bond_style harmonic/shift/cut/kk, pair_style zero kokkos
v100-fene_expand bond_style fene/expand/kk, pair_style zero kokkos
v100-quartic bond_style quartic/kk, pair_style zero kokkos
v100-table bond_style table/kk, pair_style zero kokkos
v100-harmonic_shift bond_style harmonic/shift/kk, pair_style zero kokkos
p100-fene bond_style fene/kk, pair_style zero kokkos
p100-fene-cpu-reference bond_style fene, pair_style zero gpu
p100-fene_expand bond_style fene/expand/kk, pair_style zero kokkos
p100-fene_expand-cpu-reference bond_style fene/expand, pair_style zero gpu
p100-harmonic_shift bond_style harmonic/shift/kk, pair_style zero kokkos
p100-harmonic_shift-cpu-reference bond_style harmonic/shift, pair_style zero gpu
p100-morse bond_style morse/kk, pair_style zero kokkos
p100-morse-cpu-reference bond_style morse, pair_style zero gpu
p100-mm3 bond_style mm3/kk, pair_style zero kokkos
p100-mm3-cpu-reference bond_style mm3, pair_style zero gpu
p100-table bond_style table/kk, pair_style zero kokkos
p100-table-cpu-reference bond_style table, pair_style zero gpu
p100-quartic_exp bond_style quartic/exp/kk, pair_style zero kokkos
p100-quartic_exp-cpu-reference bond_style quartic/exp, pair_style zero gpu
p100-gaussian bond_style gaussian/kk, pair_style zero kokkos
p100-gaussian-cpu-reference bond_style gaussian, pair_style zero gpu
p100-nonlinear bond_style nonlinear/kk, pair_style zero kokkos
p100-nonlinear-cpu-reference bond_style nonlinear, pair_style zero gpu
p100-harmonic bond_style harmonic/kk, pair_style zero kokkos
p100-harmonic-cpu-reference bond_style harmonic, pair_style zero gpu
p100-harmonic_restrain bond_style harmonic/restrain/kk, pair_style zero kokkos
p100-harmonic_restrain-cpu-reference bond_style harmonic/restrain, pair_style zero gpu
p100-quartic bond_style quartic/kk, pair_style zero kokkos
p100-quartic-cpu-reference bond_style quartic, pair_style zero gpu
p100-class2 bond_style class2/kk, pair_style zero kokkos
p100-class2-cpu-reference bond_style class2, pair_style zero gpu
p100-gromos bond_style gromos/kk, pair_style zero kokkos
p100-gromos-cpu-reference bond_style gromos, pair_style zero gpu
p100-fene_nm bond_style fene/nm/kk, pair_style zero kokkos
p100-fene_nm-cpu-reference bond_style fene/nm, pair_style zero gpu
p100-harmonic_shift_cut bond_style harmonic/shift/cut/kk, pair_style zero kokkos
p100-harmonic_shift_cut-cpu-reference bond_style harmonic/shift/cut, pair_style zero gpu

输入与提交#

第一组的输入如下。单位、边界条件和参数以对应 in.lammps 为准。

units metal
atom_style bond
atom_modify map array
boundary f f f
bond_style quartic/exp/kk 
read_data model.data
bond_coeff 1 1.54 200.0 -100.0 50.0 0.0 1.0
pair_style zero 10.0
pair_coeff * *
special_bonds lj/coul 1 1 1
neighbor 0.3 bin
neigh_modify every 1 delay 0 check yes
thermo_style custom step atoms pe ebond
thermo_modify norm no format float %.16g
set atom 2 x 1.3499000000
run 0
print "1.3500000000 -0.0001000000 1.3499000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-00.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.3500000000
run 0
print "1.3500000000 0.0000000000 1.3500000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-01.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.3501000000
run 0
print "1.3500000000 0.0001000000 1.3501000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-02.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.4499000000
run 0
print "1.4500000000 -0.0001000000 1.4499000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-03.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.4500000000
run 0
print "1.4500000000 0.0000000000 1.4500000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-04.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.4501000000
run 0
print "1.4500000000 0.0001000000 1.4501000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-05.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.5499000000
run 0
print "1.5500000000 -0.0001000000 1.5499000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-06.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.5500000000
run 0
print "1.5500000000 0.0000000000 1.5500000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-07.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.5501000000
run 0
print "1.5500000000 0.0001000000 1.5501000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-08.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.6499000000
run 0
print "1.6500000000 -0.0001000000 1.6499000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-09.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.6500000000
run 0
print "1.6500000000 0.0000000000 1.6500000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-10.dump id type x y z fx fy fz modify sort id format float %.16g
set atom 2 x 1.6501000000
run 0
print "1.6500000000 0.0001000000 1.6501000000 $(pe:%.16g) $(ebond:%.16g)" append energy.dat screen no
write_dump all custom forces-11.dump id type x y z fx fy fz modify sort id format float %.16g

在附件根目录提交第一组或所选条件。

sbatch --array=0 run.slurm
# 按 CASES 的完整列表提交
sbatch run.slurm

脚本申请一个 MPI 任务、一个 CPU 核和一张 P100,根据 MODES 加载对应模块。实际调用保存在 cases.json,例如第一组的命令参数为:

lmp -k on g 1 -sf kk -pk kokkos newton on neigh half gpu/aware off -echo both -in in.lammps > native.out

查看输出#

结果写入 work-new/条件名。先查看日志中的实际热力学表与结束信息。

cd work-new/v100-quartic_exp
tail -n 35 native.out

本次第一组的原生日志片段如下。

Loop time of 0.000355656 on 1 procs for 0 steps with 2 atoms
Loop time of 0.000345854 on 1 procs for 0 steps with 2 atoms
Loop time of 0.000337825 on 1 procs for 0 steps with 2 atoms
Loop time of 0.000347672 on 1 procs for 0 steps with 2 atoms
Loop time of 0.000338413 on 1 procs for 0 steps with 2 atoms
Total wall time: 0:00:00

原始输入若包含 dump、数据表或其他导出命令,相应文件会保存在该条件目录。附件提供的参考数据文件为 energy.dat, forces-00.dump, forces-01.dump, forces-02.dump, forces-03.dump, forces-04.dump。按自己的体系修改输入时,保留对应势函数文件、单位和原子样式设置。

再次运行前,将已有的同名 work-new 条件目录改名保存。

本次输出示例#

第一组原生热力学输出的开头如下,列名由该组输入的 thermo_style 设置。

   Step        Atoms        PotEng         E_bond    
         0           2  7.979883358437999 7.979883358437999
Loop time of 0.000392439 on 1 procs for 0 steps with 2 atoms

需要根据你的体系选择资源?联系讨论计算需求。实际可用软件与权限,以当前账户和运行环境为准。

查看配图

放大后可滚动查看细节,按 Esc 关闭。

100%查看原图 ↗