brucefan1983 / nep_cpu Goto Github PK

Is it possible to enable OpenMP parallel computing in NEP_CPU? This may speed up the computation in some use cases. Thanks in advance!

People tested that the DFT-D3 part is equivalent to that from ASE (if using the same cutoff and damping), so I think there is no need to have DFT-D3 in NEP_CPU, as it requires using a very large neighbor list (we used a fixed neighbor list size) that can cause memory problem for some calculations. @elindgren do you agree if I remove it?

@brucefan1983 Would it be possible to revert back to using std::cout for logging/prints instead of printf? Capturing the output from nep.cpp becomes difficult if it is written with printf, as it does not write to the same output stream as std::cout.

Here is an example of the difference between the two:

Old:

std::cout << "Use the NEP2 potential with " << paramb.num_types << " atom type(s).\n";

New:

printf("Use the NEP2 potential with %d atom type.\n", paramb.num_types);

I'm not sure whether I'm misunderstanding the physics or NEP, but I get very different results when I exclude a frozen group of atoms using neigh_modify exclude group SOLID SOLID where SOLID refers to a group of atoms that is not time-integrated and have had their forces set to 0.0 0.0 0.0

Is it expected that NEP works with excluding groups of atoms?

../pair_NEP.cpp: In member function ‘virtual void LAMMPS_NS::PairNEP::init_style()’:
../pair_NEP.cpp:88:33: error: ‘int LAMMPS_NS::NeighRequest::half’ is protected within this context
88 | neighbor->requests[irequest]->half = 0;
| ^~~~
In file included from ../pair_NEP.cpp:24:
../neigh_request.h:55:7: note: declared protected here
55 | int half; // half neigh list (set by default)
| ^~~~
../pair_NEP.cpp:89:33: error: ‘int LAMMPS_NS::NeighRequest::full’ is protected within this context
89 | neighbor->requests[irequest]->full = 1;
| ^~~~
In file included from ../pair_NEP.cpp:24:
../neigh_request.h:56:7: note: declared protected here
56 | int full; // full neigh list
| ^~~~

Will do this at some poiont.

Dear Bruce
I copied the USER-NEP dir into lammps src. But when I tried "cmake -D PKG_USER-NEP=yes ." it does not add it the list of packages to be installed along with the following warning:
Manually-specified variables were not used by the project:
PKG_USER-NEP
So I added USER-NEP to the list of standard packages in lammps_root/cmake/CMakeList.txt.
Maybe it helps others.

I would like to be able to access gradients of descriptors w.r.t. atomic coordinates for some prototype work I am doing with Calorine. Would it be possible to write a function for this in NEP_CPU, @brucefan1983? From what I can tell, the gradients are currently baked directly into the force expressions which makes it hard for me to see which bits would go where.