36 lines
1.2 KiB
Plaintext
36 lines
1.2 KiB
Plaintext
These are input scripts designed for use with the GPU package.
|
|
|
|
To run them, you must first build LAMMPS with the GPU package
|
|
installed, following the steps explained in Section 2.3 of
|
|
doc/Section_start.html and lib/gpu/README. An overview of building
|
|
and running LAMMPS with the GPU package is given in Section 5.6 of
|
|
doc/Section_accelerate.html. Note that you can choose the precision
|
|
at which computations are performed on the GPU in the build process.
|
|
|
|
Note that lines such as this in each of the input scripts:
|
|
|
|
package gpu force/neigh 0 1 1
|
|
|
|
are set for running on a compute node with 2 GPUs. If you
|
|
have a single GPU, you should comment out the line, since
|
|
the default is 1 GPU per compute node.
|
|
|
|
The scripts can be run in the usual manner:
|
|
|
|
lmp_g++ < in.gpu.melt.2.5
|
|
lmp_g++ < in.gpu.melt.5.0
|
|
lmp_g++ < in.gpu.phosphate
|
|
lmp_g++ < in.gpu.rhodo
|
|
|
|
mpirun -np 4 lmp_g++ < in.gpu.melt.2.5
|
|
mpirun -np 4 lmp_g++ < in.gpu.melt.5.0
|
|
mpirun -np 4 lmp_g++ < in.gpu.phosphate
|
|
mpirun -np 4 lmp_g++ < in.gpu.rhodo
|
|
|
|
The first set of commmands will run a single MPI task using a single
|
|
GPU (even if you have 2 GPUs).
|
|
|
|
The second set of commands will run 4 MPI tasks, with 2 MPI tasks per
|
|
GPU (if you have 2 GPUs), or 4 MPI tasks per GPU (if you have a single
|
|
GPU).
|