| 1 |
< |
OOPSE |
| 1 |
> |
What is OpenMD? |
| 2 |
|
|
| 3 |
< |
OOPSE is an open-source Object-Oriented Parallel Simulation Engine. |
| 4 |
< |
It is primarily used to perform molecular dynamics simulations on |
| 5 |
< |
"strange" atom types that are not normally handled by other simulation |
| 6 |
< |
packages. This includes atoms with orientational degrees of freedom |
| 7 |
< |
(point dipoles, sticky atoms), as well as transition metals under the |
| 8 |
< |
Embedded Atom Method (EAM) or Sutton-Chen (SC) potentials. |
| 3 |
> |
OpenMD is an open source molecular dynamics engine which is capable of |
| 4 |
> |
efficiently simulating liquids, proteins, nanoparticles, interfaces, |
| 5 |
> |
and other complex systems using atom types with orientational degrees |
| 6 |
> |
of freedom (e.g. "sticky" atoms, point dipoles, and coarse-grained |
| 7 |
> |
assemblies). Proteins, zeolites, lipids, transition metals (bulk, flat |
| 8 |
> |
interfaces, and nanoparticles) have all been simulated using force |
| 9 |
> |
fields included with the code. OpenMD works on parallel computers |
| 10 |
> |
using the Message Passing Interface (MPI), and comes with a number of |
| 11 |
> |
analysis and utility programs that are easy to use and modify. An |
| 12 |
> |
OpenMD simulation is specified using a very simple meta-data language |
| 13 |
> |
that is easy to learn. |
| 14 |
|
|
| 15 |
< |
Simulations are started in OOPSE using two files: |
| 15 |
> |
Simulations are started in OpenMD using a single Molecular Dynamics (.md) |
| 16 |
> |
file. These files must start with the <OpenMD> tag and must |
| 17 |
> |
have two sections: |
| 18 |
|
|
| 19 |
< |
1) a C-based meta-data (.md) file, and |
| 19 |
> |
1) a C-based <MetaData> section, and |
| 20 |
|
|
| 21 |
< |
2) a modified XYZ format for initial coordinate and velocity information. |
| 21 |
> |
2) a <Snapshot> block for initial coordinate and velocity information. |
| 22 |
|
|
| 23 |
< |
Detailed descriptions of the structures of these two files are |
| 23 |
> |
Detailed descriptions of the structures of these files are |
| 24 |
|
available in the "doc" directory. Sample simulations are |
| 25 |
|
available in the "samples" directory. |
| 26 |
|
|
| 27 |
< |
What you need to compile and use OOPSE: |
| 27 |
> |
What you need to compile and use OpenMD: |
| 28 |
|
|
| 29 |
< |
1) Good C, C++ and Fortran95 compilers. We've built and tested OOPSE |
| 29 |
> |
1) Good C++ and C compilers. We've built and tested OpenMD |
| 30 |
|
on the following architecture & compiler combinations: |
| 31 |
|
|
| 32 |
< |
Architecture CC CXX F90 Notes |
| 33 |
< |
------------------------- ---- ----- ----- ---------------------- |
| 34 |
< |
ix86-pc-linux-gnu icc icpc ifort (Intel versions 7-9) |
| 35 |
< |
powerpc-apple-darwin8.4.0 gcc g++ xlf (GNU v.4 / IBM XL v. 8.1) |
| 36 |
< |
x86_64-unknown-linux-gnu pgcc pgCC pgf95 (Portland Group v. 6.0) |
| 37 |
< |
sparc-sun-solaris2.8 cc CC f95 (Sun ONE Studio 10) |
| 31 |
< |
|
| 32 |
< |
We've successfully compiled OOPSE with the Pathscale c, c++, and |
| 33 |
< |
Fortran95 compilers on the x86_64-unknown-linux-gnu architecture, |
| 34 |
< |
but a bug in the exception handling on these compilers causes |
| 35 |
< |
OOPSE to abort (rather than providing a useful error message) when |
| 36 |
< |
an error is found in the meta-data file. |
| 37 |
< |
|
| 38 |
< |
The GNU compilers, gcc and g++, work very well, but you will need to |
| 39 |
< |
use a Fortran *95* compler for the fortran side of OOPSE, and |
| 40 |
< |
pgf90 does not implement the required portions of the f95 |
| 41 |
< |
language. Fortran77 and Fortran90 (i.e. g77 and pgf90) are *not* |
| 42 |
< |
sufficient to compile the fortran portions of OOPSE. Gfortran |
| 43 |
< |
does not yet build our fortran code, but g95 might. If you use |
| 44 |
< |
one of these compilers, please let us know of any build or runtime |
| 45 |
< |
errors. |
| 32 |
> |
Architecture CC CXX Notes |
| 33 |
> |
------------------------- ---- ----- ---------------------- |
| 34 |
> |
ix86-pc-linux-gnu icc icpc (Intel v. 10) |
| 35 |
> |
i386-apple-darwin11.1.0 gcc g++ (GNU v.4.3) |
| 36 |
> |
i386-apple-darwin11.1.0 icc icpc (Intel v. 12) |
| 37 |
> |
x86_64-unknown-linux-gnu gcc g++ (Pathscale 3.2) |
| 38 |
|
|
| 39 |
|
2) GNU make (also known as gmake). Regular make won't work. |
| 40 |
|
Really. We've tried. Don't bother with regular make. |
| 41 |
|
Seriously. You need GNU make. Did we mention that you |
| 42 |
|
need GNU make? |
| 43 |
|
|
| 44 |
< |
3) Perl. Compilation dependencies in Fortran95 are somewhat |
| 53 |
< |
complicated, so the build process uses a perl script called |
| 54 |
< |
filepp to do this job. You need perl for filepp, so you |
| 55 |
< |
need perl to build OOPSE. |
| 44 |
> |
3) Perl. |
| 45 |
|
|
| 46 |
< |
4) MPI is optional for the single processor version of OOPSE, |
| 47 |
< |
but is required if you want OOPSE to run in parallel. |
| 46 |
> |
4) MPI-2 is optional for the single processor version of OpenMD, but |
| 47 |
> |
is required if you want OpenMD to run in parallel. We like |
| 48 |
> |
OpenMPI. Other implementations of MPI-2 might work, but we haven't |
| 49 |
> |
tried. You can get Open MPI here: http://www.open-mpi.org/ |
| 50 |
|
|
| 60 |
– |
We like MPICH-1.2.*. Other implementations might work, but we |
| 61 |
– |
haven't tried. You can get MPICH here: |
| 62 |
– |
http://www-unix.mcs.anl.gov/mpi/mpich/ |
| 63 |
– |
|
| 51 |
|
INSTRUCTIONS |
| 52 |
|
|
| 53 |
|
1) Get, build, and test the required pieces above. |
| 54 |
< |
2) ./configure (or ./configure --with-mpi=/usr/local/mpich) |
| 54 |
> |
2) ./configure |
| 55 |
|
3) make |
| 56 |
|
4) make install |
| 57 |
|
|
