CHARMM supports a broad range of crystal lattice types and operations, but those with rotations (P21, etc.) cannot be simulated with the fast DOMDEC engine, or whenever the COLFFT feature is included in the compiled code. One can still build parallel code with PME, however, but it it will be noticeably slower. More advanced features of the build processes are required in order to create an executable for this. Be sure to setup the compiler and MPI library build environment first, before running either build command.

With the procedure, the following works:

# cleanup previous builds
rm -rf build/em64t_M lib/em64t_M
# compile for Intel

For the cmake configure process, in the new, empty build directory:

../charmm/configure --with-intel --without-openmm --without-domdec --without-mkl --without-fftw -r COLFFT -a ASYNC_PME

The '--without-openmm' is optional, but I use it to speed up the configure process, so it won't waste time looking for CUDA and OpenCL tools. It possible that the '-r COLFFT -a ASYNC_PME' may be enough and the remaining '--without' options are not needed, but I have not tested that.

I have built and tested c43b1 with the above procedures, and have previously used such code to run PME simulations with full crystal symmetry, including rotations.

Rick Venable
computational chemist