different output for the example provided by compute xrd?

LAMMPS LAMMPS Mailing List Mirror
1

Dear Sir or Madam

I recently try to use the ‘compute xrd’ command. So I first try to reproduce the results from example. But for xrd I got is:

Histogrammed data for fix 1

TimeStep Number-of-bins Total-counts Missing-counts Min-value Max-value

Bin Coord Count Count/Total

0 200 90152 90152 2.9957e-30 1.52751e+08
1 40.1 384 0.00425947
2 40.3 96 0.00106487
3 40.5 288 0.0031946

However, the correct provided by the example is:

Histogrammed data for fix 1

TimeStep Number-of-bins Total-counts Missing-counts Min-value Max-value

Bin Coord Count Count/Total

0 200 2.11404e+09 0 40.0105 79.9779
1 40.1 3.06101e-21 1.44795e-30
2 40.3 1.30992e-21 6.19629e-31

There is big difference between the Count.

Do anyone what cause this problem?

Thanks very much for your help. Wish you all the best!

Hao

2

Dear Sir or Madam

I recently try to use the ‘compute xrd’ command. So I first try to reproduce the results from example. But for xrd I got is:

which input exactly from which version of LAMMPS did you run and which file is your reference output?

axel.

3

Dear Axel

The LAMMPS version I am using is “31 MAR 2017”. The input file I use is the example file provided by LAMMPS and it is in /examples/USER/diffraction/BulkNi.in.

The reference output is the output provided by LAMMPS and it is /examples/USER/diffraction/Output/bulkNi.hist.xrd.correct

I make LAMMPS as follows: make yes-user-diffraction; make mpi

The command I use to generate the output file is: mpirun -np 4 ./lmp_mpi -in BulkNi.in

Thanks very much for your help.

Wish you all the best!

Hao

4

Dear Axel

The LAMMPS version I am using is “31 MAR 2017”. The input file I use is the example file provided by LAMMPS and it is in /examples/USER/diffraction/BulkNi.in.

The reference output is the output provided by LAMMPS and it is /examples/USER/diffraction/Output/bulkNi.hist.xrd.correct

I make LAMMPS as follows: make yes-user-diffraction; make mpi

The command I use to generate the output file is: mpirun -np 4 ./lmp_mpi -in BulkNi.in

i don’t know about this old version of LAMMPS, but with the current version, there is a small, but significant difference between the file examples/USER/diffraction/BulkNi.in and examples/USER/diffraction/Output/BulkNi.in. if i run the latter, i get results that are comparable within the range of floating point accuracy, i.e. all the “near zero” values are of the same magnitude and the one peak is represented exactly to the output precision.

so, please try the input BulkNi.in file in the Output folder and compare those results. if they fail to reproduce, you may want to try upgrading LAMMPS.

axel.

5

Thanks very much, Axel. Now it works. I tried small system and it is relatively fast. However, when I used my real model which contain around 1,600,000 atoms, the calculations become extremely slow. Do anyone know how to make it faster? Thanks very much.

6

Thanks very much, Axel. Now it works. I tried small system and it is relatively fast. However, when I used my real model which contain around 1,600,000 atoms, the calculations become extremely slow. Do anyone know how to make it faster? Thanks very much.

i would not have expected it to be different. these kind of calculations tend to be difficult to scale up to large systems.
please note, that the computes support multi-threading via OpenMP (independent of the USER-OMP package) so you may see some performance improvement using hybrid MPI and OpenMP parallelization.

beyond that, i think your best chances to get advice is to contact the author of the package directly.
the only other way i know would to speed things up would be to use a smaller model, but i don’t think that suggestion is of much help to you. :wink:

good luck,
axel.