Conclusion

The key aims of this project were: to complete the parallelisation SCATCI and to port it to HECToR. This has been achieved and a scalable parallel version of UKRMol-in has been developed for use on HECToR and other large HPC resources. The Hamiltonian build and diagonalisation steps were implemented by calls to SLEPc and PETSc. Further improvements to the initial load balancing and performance of the parallel UKRMol-in code were made such that scalability to several nodes on HECToR is now achievable and Zdenek Masin, a PhD student at the Open University has performed runs using 640 processes on HECToR. Much larger problem sizes may now be investigated with UKRMol-in. This work has been made available to HECToR users of UKRMol-in and has been submitted to CCPForge.