Error: Cannot open dump file

Submitted by cportiz on Thu, 09/23/2010 - 05:46

Hello everyone,
I'm having problems finalizing my LIGGGHTS installation. I followed the installation tutorial and completed the make step. When it came time to testing the examples, it appears to run, but at the end it returns an error saying it cannot open dump file. Any ideas? Am I having a permissions problem? I've already tested that I can write to the folder.
Thanks in advance!
Carlos

p.s. Here is some sample output:

./LIGGGHTS < examples/LIGGGHTS/cohesion/in.cohesion
LIGGGHTS 1.1.7 based on lammps-10Mar10
Created orthogonal box = (-0.05 -0.05 0) to (0.05 0.05 0.15)
1 by 1 by 1 processor grid
0 atoms in group nve_group
Particle insertion: 1001 every 2330 steps, 1000 by step 1
Setting up run ...
Memory usage per processor = 9.93308 Mbytes
Step Atoms KinEng 1 Volume
0 0 -0 0 0.0015
WARNING: Less insertions than requested
1 433 0.0010738099 0 0.0015
Loop time of 3.08804 on 1 procs for 1 steps with 433 atoms

Pair time (%) = 0.000110149 (0.00356696)
Neigh time (%) = 0.00756192 (0.244877)
Comm time (%) = 2.31266e-05 (0.000748908)
Outpt time (%) = 7.98702e-05 (0.00258643)
Other time (%) = 3.08027 (99.7482)

Nlocal: 433 ave 433 max 433 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost: 0 ave 0 max 0 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs: 2050 ave 2050 max 2050 min
Histogram: 1 0 0 0 0 0 0 0 0 0

Total # of neighbors = 2050
Ave neighs/atom = 4.73441
Neighbor list builds = 1
Dangerous builds = 0
ERROR on proc 0: Cannot open dump file
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0

Forums:

cportiz | Thu, 09/23/2010 - 06:03

I installed LAMMPS alone from the sandia website and their examples run just fine. However, LIGGGHTS is running none of its examples or sandia's examples. Also, I forgot to say earlier that I am running on the latest Ubuntu build.
I'll add more if I realize any other clues.
--C

ckloss's picture

ckloss | Thu, 09/23/2010 - 08:45

Hi Carlos,

>>LIGGGHTS is running none of its examples
Every example requires an empty directory called "post" - this was not created by git
Git was enabled for LIGGGHTS just monday. Should be fixed not - try git pull - this should revolve the issue

>>LIGGGHTS is running of sandia's examples
I don't think this is true... However, it might be that some of the examples of the current version don't run
if you try version 2010March10, all of the examples should run

Christoph

cportiz | Thu, 09/23/2010 - 15:24

LIGGGHTS really was not running sandia's examples, but it's because the makefile is not ideal for my system. I realized I had used a different makefile for LAMMPS than for LIGGGHTS, so I pulled the makefile I used for LAMMPS and used it to compile LIGGGHTS and it compiles, such that now all my LAMMPS examples finish cleanly.

However, I still get the same "Error on proc 0: cannot open dump file" error for LIGGGHTS eamples. The post folder is there and there is an empty file dump.example for each one I've tried.

I have tried both the tarball and the latest version of LIGGGHTS from git with the same results.

cportiz | Thu, 09/23/2010 - 15:31

I've gotten the examples to run by doing the following:

Modify the in.example file such that instead of "post/dump.example" it's just "dump.example". This puts the dump files in the same folder as the liggghts executable for me. I would like to get it working with the post folder - I think I must be having some kind of permissions issue. I've checked that I have read and write access to the folder, but maybe I don't have file creation ability?

ckloss's picture

ckloss | Thu, 09/23/2010 - 16:31

yes.
that was why I wrote that the examples as you download them require an empty "post" dir

mhk | Wed, 06/12/2013 - 09:51

Hello everyone,
I was wondering if the following error is serious:
WARNING: Less insertions than requested (fix_pour.cpp:433)

does this means that the simulation was not done completely?
any help is appreciated!

cstoltz | Wed, 06/12/2013 - 16:12

The error means that the code wasn't able to insert the desired number of particles at a given insertion event. This won't necessarily ruin a simulation, but is something to generally be avoided if possible. You need to either lower the insertion rate or expand the domain in which you're trying to insert particles.

Also, I see you're using fix_pour - I'd recommend using one of the fix_insert commands and upgrading to the most recent version of LIGGGHTS.

Regards,
Chris