LAMMPS questions

[Linda Lo]

Hi all,

Will the LAMMPS input need to use user developed packages?
In addition, should we need to know how to execute interactive MD?
Another, should we need to know how to compile LAMMPS as a library and then coupling with other MD code?

Thanks,
Linda

[W. Michael Brown]

Hi Linda Lo,

Let me consult with Trung to give a set of packages that will be
sufficient for the test cases we might choose and reply in this thread
early next week. We probably won't require any USER packages...

There will be no requirement to use interactive MD or LAMMPS as a library.

- Mike

[Linda Lo]

That's a good news to us! We can save a lot of effort on it.

Thanks,
Linda

[Nguyen, Trung D.]

Hi Linda and all,

Mike and I discussed and decided the following packages that will be sufficient for the test cases:

1- MANYBODY
2- MOLECULE
3- KSPACE
4- ASPHERE
5- RIGID

Regarding the input script for the test cases, you are allowed to make the following changes:

1- Toggle the newton flag on or off
2- Tune the atom sorting frequency (via atom_modify sort)
3- Use the optimized version of the pair styles and fixes available in OPT, GPU, USER-OMP and USER-CUDA (see the doc page of the pair styles and fixes).
4- Change the cutoff value of the Coulombic interaction (for example, in pair styles that consisting of a /coul/long term) when the kspace solver is used.
5- Change the neighbor list build settings (via the commands neighbor and neigh_modify) as long as the number of dangerous builds reported at the end of the runs is zero; otherwise (i.e. nonzero dangerous builds), the results will be considered incorrect.
6- Use run_style verlet/split to divide the work between real space calculations and k-space solver among CPUs.
7- Use the command balance for static load balancing, or fix balance for dynamic load balancing.

Let me know if you have further questions.

Best,
-Trung

--
You received this message because you are subscribed to the Google Groups "SCC12" group.
To post to this group, send email to sc...@googlegroups.com<mailto:sc...@googlegroups.com>.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com<mailto:scc12+un...@googlegroups.com>.
Visit this group at http://groups.google.com/group/scc12?hl=en.

[Linda Lo]

Thanks! It really narrows lots of works!

Thanks,
Linda

[Stephanie Labasan]

Has anyone had any experience with the dump command for lammps, specifically, how do you visualize the output? Using XMovie or the jpeg library?

Stephanie

[Nguyen, Trung D.]

Hi Stephanie,

 

you can use dump image to generate snapshots in the .ppm or .jpg formats. The image files can be concatenated into a movie, for example, using ffmpeg.

 

You can also use dump dcd to store a series of snapshots into a DCD file, which can be read in by VMD: http://www.ks.uiuc.edu/Research/vmd/

For example, in the LAMMPS input script you specify

 

# Write atom positions into  xyz files every 100000 steps

dump 1 all xyz 100000 x_t*.xyz

# Write atom positions into a dcd file every 100 steps

dump 2 all dcd 100 trajectory.dcd

 

to  generate x_t0.xyz, x_t100000.xyz, etc. and trajectory.dcd.

 

Then with VMD installed on your machine, you can visualize them by:

 

/path_to_vmd_bin/vmd x_t0.xyz trajectory.dcd

 

The MovieMaker plugin in VMD can be used to generate movies from the DCD file: http://www.ks.uiuc.edu/Research/vmd/plugins/vmdmovie/

 

Hope it helps,

-Trung

--

You received this message because you are subscribed to the Google Groups "SCC12" group.

To post to this group, send email to sc...@googlegroups.com.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com.

[Stephanie Labasan]

HI Trung, 

I'm running the example problem indent and was able to produce a dump.indent file by uncommenting the following lines in the input file: 

dump            1 all atom 250 dump.indent

#dump           1 all image 1000 image.*.jpg type type &

#               zoom 1.6 adiam 1.5

#dump_modify    1 pad 5

I wanted to produce a jpg file, so I ran the problem again like this, but was told I cannot dump JPG file: 

#dump            1 all atom 250 dump.indent

dump           1 all image 1000 image.*.jpg type type &

               zoom 1.6 adiam 1.5

#dump_modify    1 pad 5

Is there a way to convert the dump.indent into a visual file? I'm still a little confused on how the dump command works from the documentation.

Sincerely,

Stephanie Labasan

[Ethan Madden]

In regards to these test cases, the RIGID case is in the applications example directory. Where can the others be found?

On a related note, how accurate are the results expected to be in competition?

On Tuesday, October 2, 2012 11:29:43 AM UTC-4, Nguyen, Trung D. wrote:

Hi Linda and all,

Mike and I discussed and decided the following packages that will be sufficient for the test cases:

1- MANYBODY
2- MOLECULE
3- KSPACE
4- ASPHERE
5- RIGID

Regarding the input script for the test cases, you are allowed to make the following changes:

1- Toggle the newton flag on or off
2- Tune the atom sorting frequency (via atom_modify sort)
3- Use the optimized version of the pair styles and fixes available in OPT, GPU, USER-OMP and USER-CUDA  (see the doc page of the pair styles and fixes).
4- Change the cutoff value of the Coulombic interaction (for example, in pair styles that consisting of a /coul/long term) when the kspace solver is used.
5- Change the neighbor list build settings (via the commands neighbor and neigh_modify) as long as the number of dangerous builds reported at the end of the runs is zero; otherwise (i.e. nonzero dangerous builds), the results will be considered incorrect.
6- Use run_style verlet/split to divide the work between real space calculations and k-space solver among CPUs.
7- Use the command balance for static load balancing, or fix balance for dynamic load balancing.

Let me know if you have further questions.

Best,
-Trung

To unsubscribe from this group, send email to scc12+un...@googlegroups.com<mailto:scc12+unsub...@googlegroups.com>.

[Nguyen, Trung D.]

Hi Stephanie,

you can visualize the snapshots stored in the file dump.indent using VMD:

vmd -lammpstrj dump.indent

or from the VMD menu: File/New Molecule; choose the file dump.indent and specify the file type as LAMMPS Trajectory. After the file is loaded, you can go through the snapshots stored in that file using the slide bar in the VMD Main window. To generate a picture for each snapshot (i.e. frame), you can go to File/Render, choose the Render tool (e.g. Tachyon), select File name and click on Start Rendering. To generate a movie from all the snapshots, go to Extensions/Visualization/Movie Maker; select Renderer as Tachyon, and Movie Settings as Trajectory..

If you want to produce jpg files with dump image, you need to install libjpeg (http://en.wikipedia.org/wiki/Libjpeg), and then rebuild LAMMPS with -DLAMMPS_JPEG added to LMP_INC and the paths JPG_INC, JPG_PATH, JPG_LIB correctly specified in the Makefile.${machine} you are using. See more details at Step 7 in the following url: http://lammps.sandia.gov/doc/Section_start.html

Hope that helps,

-Trung

From: Stephanie Labasan <labasan....@gmail.com<mailto:labasan....@gmail.com>>
Date: Monday, October 15, 2012 5:28 PM
To: "sc...@googlegroups.com<mailto:sc...@googlegroups.com>" <sc...@googlegroups.com<mailto:sc...@googlegroups.com>>

From: sc...@googlegroups.com<javascript:> [mailto:sc...@googlegroups.com<javascript:>] On Behalf Of Stephanie Labasan
Sent: Friday, October 12, 2012 5:43 PM
To: sc...@googlegroups.com<javascript:>
Subject: SCC12 Re: LAMMPS questions

Has anyone had any experience with the dump command for lammps, specifically, how do you visualize the output? Using XMovie or the jpeg library?

Stephanie

On Thursday, September 27, 2012 8:53:30 PM UTC-7, Linda Lo wrote:
Hi all,

Will the LAMMPS input need to use user developed packages?
In addition, should we need to know how to execute interactive MD?
Another, should we need to know how to compile LAMMPS as a library and then coupling with other MD code?

Thanks,
Linda
--
You received this message because you are subscribed to the Google Groups "SCC12" group.

To post to this group, send email to sc...@googlegroups.com<javascript:>.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com<javascript:>.

[Nguyen, Trung D.]

Hi Ethan,

the example input scripts for the given packages can be found in the following folder:

1- MANYBODY: examples/eim
2- MOLECULE: examples/micelle and examples/peptide
3- KSPACE: examples/peptide
4- ASPHERE: examples/ellipse
5- RIGID: examples/rigid

You can also look into examples/gpu to see how to use GPU acceleration.

Regarding result accuracy, we will look at certain thermodynamic quantities (like temperature, pressure, etc.) reported in the log files and the final structures stored in the dump files generated by the runs. We don't compare numerical results bit to bit, but if you are using GPU acceleration (GPU or USER-CUDA), please make sure that the package(s) is built with mixed precision.

-Trung

From: Ethan Madden <crazeh...@gmail.com<mailto:crazeh...@gmail.com>>
Date: Monday, October 15, 2012 6:20 PM
To: "sc...@googlegroups.com<mailto:sc...@googlegroups.com>" <sc...@googlegroups.com<mailto:sc...@googlegroups.com>>

Cc: Trung Nguyen <nguy...@ornl.gov<mailto:nguy...@ornl.gov>>

To post to this group, send email to sc...@googlegroups.com<javascript:><mailto:sc...@googlegroups.com<javascript:>>.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com<javascript:><mailto:scc12+un...@googlegroups.com<javascript:>>.

[Ethan Madden]

Awesome! Just what I was wondering. Thank you!

To unsubscribe from this group, send email to scc12+un...@googlegroups.com<javascript:><mailto:scc12+unsub...@googlegroups.com<javascript:>>.

[rzyrzy...@gmail.com]

Hello, sorry to disturb you.
You refer that "Change the cutoff value of the Coulombic interaction (for example, in pair styles that consisting of a /coul/long term) when the kspace solver is used. "
I just wonder that whether it will affect the accuracy of the result?