Dear LAMMPS users,
I found different behavior of LAMMPS running on windows (ICMS 26Jan17 vs latest version 5Feb18) in calculating number of cluster using compute cluster/atom command. I suspect there is some error since the new build of windows installer introduced in 11Aug17 version.
This was observed when using compute cluster/atom as post process (with rerun command), to be specific it is triggered by the usage of dynamic group. In the ICMS version the number of clusters is calculated correctly, while in the latest patch version it shows 0 cluster.
For more details, please refer to the log files and histo file outputs, uploaded to google drive.
https://drive.google.com/open?id=1odQbr0t1Kq0Gi9CdhIftuc0n-sMmQ4dg
Data files and input script are also provided to reproduce the error.
Thank you for your attention.
thanks for the files. i am not 100% in agreement with your assessment.
the issue seems to be originating in compute chunk/atom when using the
compress yes option.
the same problem also seems to show up on the first step of a regular run.
the following patch gives the correct behavior for your inputs. would you
need me to build a new windows installer with this patch included to be
able to test it?
axel.
p.s.: here is the change
diff --git a/src/compute_chunk_atom.cpp b/src/compute_chunk_atom.cpp
index dcfd2ff..aa1210c 100644
--- a/src/compute_chunk_atom.cpp
+++ b/src/compute_chunk_atom.cpp
@@ -712,13 +712,14 @@ void ComputeChunkAtom::compute_ichunk()
return;
}
- invoked_ichunk = update->ntimestep;
Dear Axel,
Thank you for detail explanations. It would be great if you could build a new windows installer with the patch, since I have not yet tried to build windows installer using fedora.
Thank you.
Best Regards,
Kristanto
Dear Axel,
Thank you for detail explanations. It would be great if you could build a
new windows installer with the patch, since I have not yet tried to build
windows installer using fedora.
OK, i've built and posted unofficial windows installer builds with the
proposed (but not yet released) changes included (plus some other pending
changes).
this version will identify itself as version 14Feb2018, which is an
indication, that this is not an official LAMMPS release (as this doesn't
exists).
i'm also not updating the "latest" links. this will keep pointing to the
official last patch released as 5Feb2018.
please check it out and let us know, if there are still problems or
inconsistencies.
thanks,
axel.
Dear Axel,
Thank you for creating a new installer build. I have tested the new build, it gives the correct cluster chunk calculation. So far I didn’t find any inconsistencies.