A GPU-based package to conduct multiple-scale simulations on the dynamicaql and kinetic processes in materials.
Platform requirements:
operation platform: Linux workstation with at least one GPU card in Nvidia Kx or later serials.
compiler: PGI Fortran community Edition version.
Cuda: 8.0 or higher.
Installation requirements:
This package dose not provide a single excutable, but consists of many libraries used in MD, or MC applications. The libraries and the applications are built seperately.
To correctlly build the msmpscu libraries, the envirenment varibale should be set in file .bashrc. Run command "resetenv" as follows:
. ./resetenv
to setup the enviremental variables, or manually change the .bashrc file as follows.
By default, the installed cuda is assumed to be 8.0 with the caplbility 3.0. If the other version, for example, cuda10.1 is installed, add the variable to .bashrc file
export CUDAV=10.1
export CUDAC=7.0
If the path of msmpscu sources is SOMEWHERE in youe home path, add the following statement in your .bashrc file:
MSMPSCUSOR=$HOME/SOMEWHERE/msmpscu; export MSMPSCUSOR
PATH=$MSMPSCUSOR:$PATH; export PATH
It is recommanded to output the intermediate files (.o, .mod) and libraries (lib*.a) to a location beyond the localtion of the mdpscu sources. For example, if the your workspace is SOMEWHERE in your home path, add the following statement in your .bashrc file:
WORKSPACE=$HOME/SOMEWHERE; export WORKSPACE
PATH=$WORKSPACE/applications:$PATH; export PATH
Then, the intermediate files and libraries will be output to $HOME/SOMEWHERE/LIB, and the excutables to be build will be output to $HOME/SOMEWHERE/applications. Run the enirenment variables set, run command "bidlib" to build the libraries (see bidlib.readme). Run command "gapp" to build an specific application(see gapp.readme).