CUDAOPTIM with AMBER 12 test input/output
Rosemary Mantell (rgm38@cam.ac.uk)
September 2015
---------------------------
CUDAOPTIM currently works with CUDA 5.5 and above and the ifort/icc and gfortran/gcc compilers. 
Compilation is disabled for the gfortran/gcc compilers as Intel compilers give the best performance. 
You can build CUDAOPTIM using cmake and ccmake. This input is for a hybrid eigenvector-following 
transition state search, using gradient-only variational eigenvector refinement and tangent space 
minimisation, with LBFGS energy minimisation for the pathway. The system is the truncated 
haemagglutinin monomer. Expected output can be found in the 'expected_output' subdirectory.


1. Make a build directory

mkdir ~/svn/OPTIM/build && cd build

2. Specify compilers and run cmake

FC=ifort CC=icc cmake ../source

3. Run ccmake and set flags

ccmake .

In the options menu, set WITH_CUDA on using the arrow keys and the enter key. Then hit 'c'
to configure, 'e' to exit (ignore the error about the SDK root directory), 'c' and 'e' again
to generate the build files.

4. Build CUDAOPTIM

make -j

CUDA specific OPTIM keywords in odata
------------------------------------

CUDA A

This specifies that CUDA with the AMBER 12 interface should be used. All other OPTIM keywords
can be researched in the documentation here:

http://www-wales.ch.cam.ac.uk/OPTIM.doc/
