There are several requirements for a CPU-GPU version of MOPAC. These are:
The computer must have a NVIDIA GPU Card installed. This card needs to have a Compute Capability equal or superior to 2.0. To check if your GPU is supported, go to https://developer.nvidia.com/cuda-gpus
Toolkit must be installed.
the CUDA toolkit is not installed, go to
and download and install it.
MOPAC2016 was written using CUDA-5.5. If a higher version of CUDA, e.g., CUDA-6.5 or CUDA-7, is already installed, then, for Linux systems, download http://openmopac.net/Linux_mopac_cuda_5.5_libs.zip, and for Windows systems, http://openmopac.net/Windows_mopac_cuda_5.5_libs.zip. Follow the instructions in README.txt. This will allow MOPAC to run using local CUDA-5.5 libraries.
:The Cuda Toolkit environment variables need to be defined in your $HOME/.bashrc file. To do this, add the following lines to .bashrc:
(Usually /usr/local/cuda-5.5 is the default CUDA_TOOLKIT_INSTALLATION_PATH.)
number of threads used in your MOPAC2016 can be defined using
export MKL_NUM_THREADS=XX ,
, whereXX is the number of threads you want to use. XX must be equal or smaller than the number of physical cores you have available in your system (MKL does not benefit from hyper-threading).
The operating system must
use glibc version 2.14. CPU-GPU MOPAC will not run if the OS uses
glibc 2.12. This means that it will not run on CentOS.
use glibc version 2.14. CPU-GPU MOPAC will not run if the OS uses glibc 2.12. This means that it will not run on CentOS.
When these requirements have been met, install the CPU-GPU version of MOPAC. During testing, be aware that GPU will not be used on systems of less than 100 atoms. Such systems will be run without using the GPU.