FAHBench is built with CMake and requires the following libraries:
Get the prerequisites
sudo apt-get install qt5-default nvidia-cuda-dev nvidia-opencl-dev
Configure an OpenMM build with CMake.
From a clean build directory, use
ccmake to configure the build with your
OpenCL, OpenMM, and Qt libraries.
make && make install!
These all have nice GUI installers.
(Optional) Download and build fftw3 for the CPU platform. I haven’t done this for the official releases. The CPU platform runs much slower without these optimized libraries.
Download OpenMM and build with CMake. The official binaries will not work with vs2015, and do not include fah-specific patches and backports. Build the release configuration. You can disable building the python bindings, C and Fortran bindings, drude, and amoeba plugins to simplify the build.
INSTALL project to install OpenMM.
Run CMake on the fahbench source directory. Finagle it until it has found all of the dependencies you just spent so long getting in order. Start by setting:
OPENMM_xxxto where you installed OpenMM.
dlls to the
Make sure your git submodules are initialized (openmm)
.tar.bz2 release of the AMDAPPSDK into this directory.
The provisioning script can’t download it because you have to accept
a license agreement through a web browser. The file has to be of the form
Enter the virtual machine with
install-fahbench.sh. Build artifacts
will be copied to the
dist/ directory in this repository.