searchusermenu
  • 发布文章
  • 消息中心
点赞
收藏
评论
分享
原创

分子动力学模拟——超算集群运行LAMMPS与数据可视化

2023-03-02 13:30:47
422
0

1.lammps的数据可视化

通常我们使用ovito/vmd来进行数据可视化,这里着重讲ovito,一方面是因为它比vmd好用(虽然不能建模),另一方面也是因为vmd的安装实在是太简单了

1.1 Windows安装ovito

还是有不少演示机器会希望直接使用Windows来看数据,所以ovito的exe下载量还是蛮高的

直接进入官网:https://www.ovito.org/,下载Windows版本
下载对应的exe
下载好后以管理员身份运行,执行默认安装即可,完成后打开软件如下:
我们进入lammps的安装目录下的examples文件夹,我们这里安装到了:C:\Users\WY\AppData\Local\LAMMPS 64-bit 22Dec2022-MPI\Examples\melt,同样是用melt做测试,但是一般默认的例子会把输出数据的命令注释掉,我们将in文件的dump这行注释出来即可;
运行这个in文件之后,会发现输出了两个文件,一个是log,一个是dump.melt,也就是轨迹文件,将这个文件用ovito打开即可看到轨迹

1.2 CentOS安装ovito

还是一样进入官网下载linux的basic版本,不用解压,直接xftp之类的工具传输过去即可

# 解压
tar -zxvf ovito-basic-3.7.12-x86_64.tar.xz
# 完成后bin目录下可以直接运行
./ovito
# 为了方便调用,我们将其设置为环境变量,由于是root安装的我们这里是:/usr/local/bin
vi .bashrc
# 加入以下行
export PATH=/usr/local/bin:$PATH
# 执行后,直接ovvito即可使用
source .bashrc

1.3 vmd的安装

步骤都跟ovito一样,这几就不赘述了,但是要注意vmd打开是会出现三个窗口的,这个是正常现象

 

2.使用slurm运行Lammps

上一篇文章介绍了很多hpc和slurm的关系,但是在示例部分我们实际上是直接使用我们的终端(Windows/linux)来运行的,换而言之如果碰到比较复杂的模型我们应该怎么办呢?总不能给hpc装一个Windows系统吧,于是我们就需要使用hpc提交作业的脚本来提交in文件

具体的集群搭建可以参考这篇文章:https://www.ctyun.cn/developer/article/363542369067077

搭建好了集群之后,我们可以使用sbatch的方法来提交作业,sbatch一些基本的命令如下:

-o job.%j.out       # 脚本执行的输出将被保存在当job.%j.out文件下,%j表示作业号;
-p C032M0128G       # 作业提交的指定分区为C032M0128G;
--qos=low           # 指定作业的QOS为low;
-J myFirstJob       # 作业在调度系统中的作业名为myFirstJob;
--nodes=1           # 申请节点数为1,如果作业不能跨节点(MPI)运行, 申请的节点数应不超过1;
--ntasks-per-node=1 # 每个节点上运行一个任务,默认一情况下也可理解为每个节点使用一个核心,如果程序不支持多线程(如openmp),这个数不应该超过1;

这里我们提交的测试脚本为:

#!/bin/bash
 
#SBATCH --job-name=lammps-test
#SBATCH --output=lammps-test.out
#SBATCH --partition=p1

mpirun -np 1 lmp -in in.melt
mpirun -np 1 lmp -in in.comb.Cu

这样就实现了同时提交两个in文件(提交命令为:sbatch lammps-test.sh),需要注意的有以下几点:

1.需要用普通用户去创建运行文件的目录,因为lammps运行的时候需要输出日志,没创建文件的权限会报错

2.不推荐使用root用户运行,即使加了--allow-run-as-root,相关的mpi依然有可能报错

3.有可能需要用lmp_mpi而不是lmp,看lammps安装的具体情况,我们的示例中为lmp

4.需要使用 1 作为slot,slurm会自动分配多个节点来计算

5.如果缺少势文件,可以从window系统里面传输到运行的目录下,或者在lammps目录下新增一个+改变in问阿金调用势文件的路径

 

结果如下(注释掉in文件的dump就可以输出轨迹文件,进而在ovito里面观看):

LAMMPS (23 Jun 2022 - Update 2)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (src/comm.cpp:98)
using 1 OpenMP thread(s) per MPI task
Lattice spacing in x,y,z = 1.6795962 1.6795962 1.6795962
Created orthogonal box = (0 0 0) to (16.795962 16.795962 16.795962)
1 by 1 by 1 MPI processor grid
Created 4000 atoms
using lattice units in orthogonal box = (0 0 0) to (16.795962 16.795962 16.795962)
create_atoms CPU = 0.000 seconds
Generated 0 of 0 mixed pair_coeff terms from geometric mixing rule
Neighbor list info ...
update every 20 steps, delay 0 steps, check no
max neighbors/atom: 2000, page size: 100000
master list distance cutoff = 2.8
ghost atom cutoff = 2.8
binsize = 1.4, bins = 12 12 12
1 neighbor lists, perpetual/occasional/extra = 1 0 0
(1) pair lj/cut, perpetual
attributes: half, newton on
pair build: half/bin/atomonly/newton
stencil: half/bin/3d
bin: standard
Setting up Verlet run ...
Unit style    : lj
Current step  : 0
Time step     : 0.005
Per MPI rank memory allocation (min/avg/max) = 3.222 | 3.222 | 3.222 Mbytes
Step          Temp          E_pair         E_mol          TotEng         Press     
0   3             -6.7733681      0             -2.2744931     -3.7033504    
50   1.6842865     -4.8082494      0             -2.2824513      5.5666131    
100   1.6712577     -4.7875609      0             -2.281301       5.6613913    
150   1.6444751     -4.7471034      0             -2.2810074      5.8614211    
200   1.6471542     -4.7509053      0             -2.2807916      5.8805431    
250   1.6645597     -4.7774327      0             -2.2812174      5.7526089    
Loop time of 0.362727 on 1 procs for 250 steps with 4000 atoms

Performance: 297744.252 tau/day, 689.223 timesteps/s
98.9% CPU use with 1 MPI tasks x 1 OpenMP threads

MPI task timing breakdown:
Section |  min time  |  avg time  |  max time  |%varavg| %total
---------------------------------------------------------------
Pair    | 0.30078    | 0.30078    | 0.30078    |   0.0 | 82.92
Neigh   | 0.050034   | 0.050034   | 0.050034   |   0.0 | 13.79
Comm    | 0.0051255  | 0.0051255  | 0.0051255  |   0.0 |  1.41
Output  | 0.00021937 | 0.00021937 | 0.00021937 |   0.0 |  0.06
Modify  | 0.005321   | 0.005321   | 0.005321   |   0.0 |  1.47
Other   |            | 0.001243   |            |       |  0.34

Nlocal:           4000 ave        4000 max        4000 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost:           5506 ave        5506 max        5506 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs:         151788 ave      151788 max      151788 min
Histogram: 1 0 0 0 0 0 0 0 0 0

Total # of neighbors = 151788
Ave neighs/atom = 37.947
Neighbor list builds = 12
Dangerous builds not checked
Total wall time: 0:00:00
LAMMPS (23 Jun 2022 - Update 2)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (src/comm.cpp:98)
using 1 OpenMP thread(s) per MPI task
Lattice spacing in x,y,z = 3.615 3.615 3.615
Created orthogonal box = (0 0 0) to (14.46 14.46 14.46)
1 by 1 by 1 MPI processor grid
Created 256 atoms
using lattice units in orthogonal box = (0 0 0) to (14.46 14.46 14.46)
create_atoms CPU = 0.000 seconds
Reading comb potential file ffield.comb with DATE: 2011-02-22
Pair COMB:
generating Coulomb integral lookup table ...
element[1] = Cu, z = 0.454784
will not apply over-coordination correction ...
Neighbor list info ...
update every 1 steps, delay 1 steps, check yes
max neighbors/atom: 2000, page size: 100000
master list distance cutoff = 12.5
ghost atom cutoff = 12.5
binsize = 6.25, bins = 3 3 3
1 neighbor lists, perpetual/occasional/extra = 1 0 0
(1) pair comb, perpetual
attributes: full, newton on
pair build: full/bin/atomonly
stencil: full/bin/3d
bin: standard
Setting up Verlet run ...
Unit style    : metal
Current step  : 0
Time step     : 0.0002
Per MPI rank memory allocation (min/avg/max) = 7.47 | 7.47 | 7.47 Mbytes
Step          Temp          TotEng         PotEng         E_vdwl         E_coul         Press          Volume           Lx             Ly             Lz             Xz      
0   10.1          -3.5063151     -3.5076155     -3.5076155      0              27.496055      3023.4645      14.46          14.46          14.46          0            
1   10.099643     -3.5063151     -3.5076155     -3.5076155      0              27.512983      3023.4645      14.46          14.46          14.46          0            
2   10.098572     -3.5063151     -3.5076153     -3.5076153      0              27.563765      3023.4645      14.46          14.46          14.46          0            
3   10.096788     -3.5063151     -3.5076151     -3.5076151      0              27.64839       3023.4645      14.46          14.46          14.46          0            
4   10.094291     -3.5063151     -3.5076148     -3.5076148      0              27.766843      3023.4645      14.46          14.46          14.46          0            
5   10.09108      -3.5063151     -3.5076144     -3.5076144      0              27.919101      3023.4645      14.46          14.46          14.46          0            
6   10.087158     -3.5063151     -3.5076139     -3.5076139      0              28.105138      3023.4645      14.46          14.46          14.46          0            
7   10.082524     -3.5063151     -3.5076133     -3.5076133      0              28.324919      3023.4645      14.46          14.46          14.46          0            
8   10.077179     -3.5063151     -3.5076126     -3.5076126      0              28.578403      3023.4645      14.46          14.46          14.46          0            
9   10.071123     -3.5063151     -3.5076118     -3.5076118      0              28.865545      3023.4645      14.46          14.46          14.46          0            
10   10.06436      -3.5063151     -3.5076109     -3.5076109      0              29.186292      3023.4645      14.46          14.46          14.46          0            
Loop time of 0.100091 on 1 procs for 10 steps with 256 atoms

Performance: 1.726 ns/day, 13.902 hours/ns, 99.909 timesteps/s
98.1% CPU use with 1 MPI tasks x 1 OpenMP threads

MPI task timing breakdown:
Section |  min time  |  avg time  |  max time  |%varavg| %total
---------------------------------------------------------------
Pair    | 0.099538   | 0.099538   | 0.099538   |   0.0 | 99.45
Neigh   | 0          | 0          | 0          |   0.0 |  0.00
Comm    | 0.00031282 | 0.00031282 | 0.00031282 |   0.0 |  0.31
Output  | 0.00015368 | 0.00015368 | 0.00015368 |   0.0 |  0.15
Modify  | 4.3161e-05 | 4.3161e-05 | 4.3161e-05 |   0.0 |  0.04
Other   |            | 4.356e-05  |            |       |  0.04

Nlocal:            256 ave         256 max         256 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost:           4375 ave        4375 max        4375 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs:              0 ave           0 max           0 min
Histogram: 1 0 0 0 0 0 0 0 0 0
FullNghs:       172544 ave      172544 max      172544 min
Histogram: 1 0 0 0 0 0 0 0 0 0

Total # of neighbors = 172544
Ave neighs/atom = 674
Neighbor list builds = 0
Dangerous builds = 0
Total wall time: 0:00:00

如果导入ovito,结果如图;

0条评论
0 / 1000
才开始学技术的小白
23文章数
2粉丝数
才开始学技术的小白
23 文章 | 2 粉丝
原创

分子动力学模拟——超算集群运行LAMMPS与数据可视化

2023-03-02 13:30:47
422
0

1.lammps的数据可视化

通常我们使用ovito/vmd来进行数据可视化,这里着重讲ovito,一方面是因为它比vmd好用(虽然不能建模),另一方面也是因为vmd的安装实在是太简单了

1.1 Windows安装ovito

还是有不少演示机器会希望直接使用Windows来看数据,所以ovito的exe下载量还是蛮高的

直接进入官网:https://www.ovito.org/,下载Windows版本
下载对应的exe
下载好后以管理员身份运行,执行默认安装即可,完成后打开软件如下:
我们进入lammps的安装目录下的examples文件夹,我们这里安装到了:C:\Users\WY\AppData\Local\LAMMPS 64-bit 22Dec2022-MPI\Examples\melt,同样是用melt做测试,但是一般默认的例子会把输出数据的命令注释掉,我们将in文件的dump这行注释出来即可;
运行这个in文件之后,会发现输出了两个文件,一个是log,一个是dump.melt,也就是轨迹文件,将这个文件用ovito打开即可看到轨迹

1.2 CentOS安装ovito

还是一样进入官网下载linux的basic版本,不用解压,直接xftp之类的工具传输过去即可

# 解压
tar -zxvf ovito-basic-3.7.12-x86_64.tar.xz
# 完成后bin目录下可以直接运行
./ovito
# 为了方便调用,我们将其设置为环境变量,由于是root安装的我们这里是:/usr/local/bin
vi .bashrc
# 加入以下行
export PATH=/usr/local/bin:$PATH
# 执行后,直接ovvito即可使用
source .bashrc

1.3 vmd的安装

步骤都跟ovito一样,这几就不赘述了,但是要注意vmd打开是会出现三个窗口的,这个是正常现象

 

2.使用slurm运行Lammps

上一篇文章介绍了很多hpc和slurm的关系,但是在示例部分我们实际上是直接使用我们的终端(Windows/linux)来运行的,换而言之如果碰到比较复杂的模型我们应该怎么办呢?总不能给hpc装一个Windows系统吧,于是我们就需要使用hpc提交作业的脚本来提交in文件

具体的集群搭建可以参考这篇文章:https://www.ctyun.cn/developer/article/363542369067077

搭建好了集群之后,我们可以使用sbatch的方法来提交作业,sbatch一些基本的命令如下:

-o job.%j.out       # 脚本执行的输出将被保存在当job.%j.out文件下,%j表示作业号;
-p C032M0128G       # 作业提交的指定分区为C032M0128G;
--qos=low           # 指定作业的QOS为low;
-J myFirstJob       # 作业在调度系统中的作业名为myFirstJob;
--nodes=1           # 申请节点数为1,如果作业不能跨节点(MPI)运行, 申请的节点数应不超过1;
--ntasks-per-node=1 # 每个节点上运行一个任务,默认一情况下也可理解为每个节点使用一个核心,如果程序不支持多线程(如openmp),这个数不应该超过1;

这里我们提交的测试脚本为:

#!/bin/bash
 
#SBATCH --job-name=lammps-test
#SBATCH --output=lammps-test.out
#SBATCH --partition=p1

mpirun -np 1 lmp -in in.melt
mpirun -np 1 lmp -in in.comb.Cu

这样就实现了同时提交两个in文件(提交命令为:sbatch lammps-test.sh),需要注意的有以下几点:

1.需要用普通用户去创建运行文件的目录,因为lammps运行的时候需要输出日志,没创建文件的权限会报错

2.不推荐使用root用户运行,即使加了--allow-run-as-root,相关的mpi依然有可能报错

3.有可能需要用lmp_mpi而不是lmp,看lammps安装的具体情况,我们的示例中为lmp

4.需要使用 1 作为slot,slurm会自动分配多个节点来计算

5.如果缺少势文件,可以从window系统里面传输到运行的目录下,或者在lammps目录下新增一个+改变in问阿金调用势文件的路径

 

结果如下(注释掉in文件的dump就可以输出轨迹文件,进而在ovito里面观看):

LAMMPS (23 Jun 2022 - Update 2)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (src/comm.cpp:98)
using 1 OpenMP thread(s) per MPI task
Lattice spacing in x,y,z = 1.6795962 1.6795962 1.6795962
Created orthogonal box = (0 0 0) to (16.795962 16.795962 16.795962)
1 by 1 by 1 MPI processor grid
Created 4000 atoms
using lattice units in orthogonal box = (0 0 0) to (16.795962 16.795962 16.795962)
create_atoms CPU = 0.000 seconds
Generated 0 of 0 mixed pair_coeff terms from geometric mixing rule
Neighbor list info ...
update every 20 steps, delay 0 steps, check no
max neighbors/atom: 2000, page size: 100000
master list distance cutoff = 2.8
ghost atom cutoff = 2.8
binsize = 1.4, bins = 12 12 12
1 neighbor lists, perpetual/occasional/extra = 1 0 0
(1) pair lj/cut, perpetual
attributes: half, newton on
pair build: half/bin/atomonly/newton
stencil: half/bin/3d
bin: standard
Setting up Verlet run ...
Unit style    : lj
Current step  : 0
Time step     : 0.005
Per MPI rank memory allocation (min/avg/max) = 3.222 | 3.222 | 3.222 Mbytes
Step          Temp          E_pair         E_mol          TotEng         Press     
0   3             -6.7733681      0             -2.2744931     -3.7033504    
50   1.6842865     -4.8082494      0             -2.2824513      5.5666131    
100   1.6712577     -4.7875609      0             -2.281301       5.6613913    
150   1.6444751     -4.7471034      0             -2.2810074      5.8614211    
200   1.6471542     -4.7509053      0             -2.2807916      5.8805431    
250   1.6645597     -4.7774327      0             -2.2812174      5.7526089    
Loop time of 0.362727 on 1 procs for 250 steps with 4000 atoms

Performance: 297744.252 tau/day, 689.223 timesteps/s
98.9% CPU use with 1 MPI tasks x 1 OpenMP threads

MPI task timing breakdown:
Section |  min time  |  avg time  |  max time  |%varavg| %total
---------------------------------------------------------------
Pair    | 0.30078    | 0.30078    | 0.30078    |   0.0 | 82.92
Neigh   | 0.050034   | 0.050034   | 0.050034   |   0.0 | 13.79
Comm    | 0.0051255  | 0.0051255  | 0.0051255  |   0.0 |  1.41
Output  | 0.00021937 | 0.00021937 | 0.00021937 |   0.0 |  0.06
Modify  | 0.005321   | 0.005321   | 0.005321   |   0.0 |  1.47
Other   |            | 0.001243   |            |       |  0.34

Nlocal:           4000 ave        4000 max        4000 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost:           5506 ave        5506 max        5506 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs:         151788 ave      151788 max      151788 min
Histogram: 1 0 0 0 0 0 0 0 0 0

Total # of neighbors = 151788
Ave neighs/atom = 37.947
Neighbor list builds = 12
Dangerous builds not checked
Total wall time: 0:00:00
LAMMPS (23 Jun 2022 - Update 2)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (src/comm.cpp:98)
using 1 OpenMP thread(s) per MPI task
Lattice spacing in x,y,z = 3.615 3.615 3.615
Created orthogonal box = (0 0 0) to (14.46 14.46 14.46)
1 by 1 by 1 MPI processor grid
Created 256 atoms
using lattice units in orthogonal box = (0 0 0) to (14.46 14.46 14.46)
create_atoms CPU = 0.000 seconds
Reading comb potential file ffield.comb with DATE: 2011-02-22
Pair COMB:
generating Coulomb integral lookup table ...
element[1] = Cu, z = 0.454784
will not apply over-coordination correction ...
Neighbor list info ...
update every 1 steps, delay 1 steps, check yes
max neighbors/atom: 2000, page size: 100000
master list distance cutoff = 12.5
ghost atom cutoff = 12.5
binsize = 6.25, bins = 3 3 3
1 neighbor lists, perpetual/occasional/extra = 1 0 0
(1) pair comb, perpetual
attributes: full, newton on
pair build: full/bin/atomonly
stencil: full/bin/3d
bin: standard
Setting up Verlet run ...
Unit style    : metal
Current step  : 0
Time step     : 0.0002
Per MPI rank memory allocation (min/avg/max) = 7.47 | 7.47 | 7.47 Mbytes
Step          Temp          TotEng         PotEng         E_vdwl         E_coul         Press          Volume           Lx             Ly             Lz             Xz      
0   10.1          -3.5063151     -3.5076155     -3.5076155      0              27.496055      3023.4645      14.46          14.46          14.46          0            
1   10.099643     -3.5063151     -3.5076155     -3.5076155      0              27.512983      3023.4645      14.46          14.46          14.46          0            
2   10.098572     -3.5063151     -3.5076153     -3.5076153      0              27.563765      3023.4645      14.46          14.46          14.46          0            
3   10.096788     -3.5063151     -3.5076151     -3.5076151      0              27.64839       3023.4645      14.46          14.46          14.46          0            
4   10.094291     -3.5063151     -3.5076148     -3.5076148      0              27.766843      3023.4645      14.46          14.46          14.46          0            
5   10.09108      -3.5063151     -3.5076144     -3.5076144      0              27.919101      3023.4645      14.46          14.46          14.46          0            
6   10.087158     -3.5063151     -3.5076139     -3.5076139      0              28.105138      3023.4645      14.46          14.46          14.46          0            
7   10.082524     -3.5063151     -3.5076133     -3.5076133      0              28.324919      3023.4645      14.46          14.46          14.46          0            
8   10.077179     -3.5063151     -3.5076126     -3.5076126      0              28.578403      3023.4645      14.46          14.46          14.46          0            
9   10.071123     -3.5063151     -3.5076118     -3.5076118      0              28.865545      3023.4645      14.46          14.46          14.46          0            
10   10.06436      -3.5063151     -3.5076109     -3.5076109      0              29.186292      3023.4645      14.46          14.46          14.46          0            
Loop time of 0.100091 on 1 procs for 10 steps with 256 atoms

Performance: 1.726 ns/day, 13.902 hours/ns, 99.909 timesteps/s
98.1% CPU use with 1 MPI tasks x 1 OpenMP threads

MPI task timing breakdown:
Section |  min time  |  avg time  |  max time  |%varavg| %total
---------------------------------------------------------------
Pair    | 0.099538   | 0.099538   | 0.099538   |   0.0 | 99.45
Neigh   | 0          | 0          | 0          |   0.0 |  0.00
Comm    | 0.00031282 | 0.00031282 | 0.00031282 |   0.0 |  0.31
Output  | 0.00015368 | 0.00015368 | 0.00015368 |   0.0 |  0.15
Modify  | 4.3161e-05 | 4.3161e-05 | 4.3161e-05 |   0.0 |  0.04
Other   |            | 4.356e-05  |            |       |  0.04

Nlocal:            256 ave         256 max         256 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost:           4375 ave        4375 max        4375 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs:              0 ave           0 max           0 min
Histogram: 1 0 0 0 0 0 0 0 0 0
FullNghs:       172544 ave      172544 max      172544 min
Histogram: 1 0 0 0 0 0 0 0 0 0

Total # of neighbors = 172544
Ave neighs/atom = 674
Neighbor list builds = 0
Dangerous builds = 0
Total wall time: 0:00:00

如果导入ovito,结果如图;

文章来自个人专栏
HPC基础技术
4 文章 | 1 订阅
0条评论
0 / 1000
请输入你的评论
0
0