grid infile keywords in Processors command

Hello,

I have a question about grid infile keywords in Processors command. I put the following line on top of my input file.

processors grid infile processors_info.txt

The content of the processors_info.txt file is the following.

1 1 2
0 1 1 1
1 1 1 2

According to the explanation from , the first line should be px py pz, and the next lines shoulds be ID I J K. I am trying to use two cores, so I assigned the processor “0” on 1 1 1 and the processor “1” on 1 1 2. Postdoctoral Associate Center for Advanced Vehicular Systems Mississippi State University

olfcglgniehenpkf.png

Hello,

I have a question about grid infile keywords in Processors command. I put
the following line on top of my input file.

processors grid infile processors_info.txt

The content of the processors_info.txt file is the following.

1 1 2
0 1 1 1
1 1 1 2

According to the explanation from http://lammps.sandia.gov/doc/
processors.html, the first line should be px py pz, and the next lines
shoulds be ID I J K. I am trying to use two cores, so I assigned the
processor "0" on 1 1 1 and the processor "1" on 1 1 2.

However, the output I get is the following.

ERROR: Expected integer parameter in input script or data file
(../comm.cpp:319)
Last command: processors grid infile processors_info.txt

Could you please tell me if this is my mistake?

​the mistake is not in the file, but in the processors command itself. it
is incorrect in multiple places.

"processors" must *always* be followed by three numbers (or *). that is the
reason for the error message.
but there is more. "infile" is not a valid keyword, but you are missing the
"custom" keyword:

so the total command would be:

processors 1 1 2 grid custom processors_info.txt

Now, I would like to explain a reason why I would like to use this
command. I and my colleagues have made some changes (added bond order term)
to MEAM algorithm, and we modified the C++ version of the code. In order to
make sure that the results (e.g. energies) from a serial execution match
those of a MPI execution. However, I found the atom lists per core keep
changing at each run. As an example, please see the picture of the simple 5
acetylene system below.

Sometimes, I get the information about top three molecules and bottom two
next, and reversed other times. I would like to get the results in the same
order each time for debugging purposes. I thought the "grid infile" keyword
could assign processors in a certain order.

​atoms are still in arbitrary order on each node, so i don't see what you
are gaining here. instead, you can uniquely identify each atom by its atom
ID (if you assign them accordingly) and then you can sort your output by
atom ID and have always a consistent output.

axel.​

olfcglgniehenpkf.png

Dr. Alex,

Thank you for your detailed explanation and correcting my mistake. It seemed we had random segmentation fault when we used multiple cores, but it actually was the neighbor indexing. For now, it seems working fine.

Thank you,

Sungkwang

olfcglgniehenpkf.png