Intel Ethernet 10 Gigabit LAMMPS Performance Study

iWARP vs. Infiniband: LAMMPS
LAMMPS (Large-scale Atomic/Molecular
Massively Parallel Simulator) is a classical
molecular dynamics code that models an
ensemble of particles in a liquid, solid, or
gaseous state. It can model atomic, poly-
meric, biological, metallic, granular, and
coarse-grained systems using a variety of
force elds and boundary conditions. The
application can model systems with only a
few particles up to several billion.
In the most general sense, LAMMPS inte-
grates Newton’s equations of motion for
collections of atoms, molecules, or macro-
scopic particles that interact via short- or
long-range forces with a variety of initial
and/or boundary conditions. For computa-
tional efciency, LAMMPS uses neighbor
lists to keep track of nearby particles. The
lists are optimized for systems with par-
ticles that are repulsive at short distances,
so that the local density of particles never
becomes too large. On parallel machines,
LAMMPS uses spatial-decomposition tech-
niques to partition the simulation domain
into small 3D sub-domains, one of which
is assigned to each processor. Processors
communicate and store “ghost” atom
information for atoms that border their
sub-domain. LAMMPS is most efcient (in
a parallel sense) for systems whose par-
ticles ll a 3D rectangular box with roughly
uniform density.
LAMMPS is designed to be easy to modify
or extend with new capabilities, such as
new force elds, atom types, boundary
conditions, or diagnostics. LAMMPS is a
freely-available open-source code, distrib-
uted under the terms of the GNU Public
License
1
. The current version is written in
C++. Earlier versions were written in F77
and F90. LAMMPS was originally developed
under a US Department of Energy (DOE)
Cooperative Research and Development
Agreement between two DOE labs and
three companies. It is distributed by Sandia
National Labs
2
.
LAMMPS runs efciently on single-proces-
sor desktop or laptop machines, but it is
designed for parallel computers. It will run
on any parallel machine that compiles C++
and supports the MPI
3
message-passing
library. This includes distributed- or shared-
memory parallel machines and Beowulf-
style clusters.
For more information, see the LAMMPS FAQ
page
4
.
Test Scenario
The lithium-ion batteries used in cell
phones and laptop computers are based
on a liquid electrolyte in which a lithium
salt is dissolved, and lithium is the cation
that is transferred across the electrolyte
during charge and discharge. Replacing the
liquid electrolyte with a polymer based
“solid” electrolyte, termed “solid polymer
electrolyte” offers advantages in weight,
size, exibility, safety, and end-of-life
disposal. However, the conductivity of
these electrolytes falls short of required
standards. The study of cation transport
in solid polymer electrolytes is very impor-
tant for overcoming this challenge.
While experimental techniques provide
information on the diffusion coefcient,
polymer segmental relaxation, and the
content of mobile ions, it is difcult to
determine a transport mechanism from
these measurements. This testing uses
molecular dynamics simulation to study
ion transport and backbone mobility of
a polyethylene oxide-based single-ion
conductor for potential lithium ion bat-
tery application. In single-ion conductors,
or ionomers, the anion is incorporated
in the polymer chain. The conductivity
then arises exclusively from the cation,
which can eliminate unwanted buildup of
anions on the electrodes. The simulation
contains 27 molecules with a total number
of atoms close to 6,000. Although this is a
modest size, observation of cation dynam-
ics into the diffusive regime requires simu-
lation runs up to 500 ns, depending on the
cation identity, the anion identity, and the
temperature.
Figure 1. LAMMPS iWARP versus InfiniBand* performance-testing
results (lower y-axis figures are better).
0.00
50.00
100.00
150.00
200.00
250.00
300.00
350.00
400.00
450.00
500.00
16 Cores 32 Cores 64 Cores 128 Cores
Time (mins)
Cores
Infinband*
iWARP
2
Intel® Ethernet 10 Gigabit iWARP LAMMPS Performance Study