Further developments in a parallel configuration interaction code for applications on complex systems

We have continued the development of our new parallel atomic structure codes, improving the methods of constructing matrices. Our configuration interaction code now features a dynamic workload distribution method to form the Hamiltonian matrix, proving to be much more scalable than the previous static version. This allows for much more efficient use of many processors. The Davidson procedure has also been improved to allow a larger initial approximation to be considered in negligible time. We report the application of our codes to the calculation of the oscillator-strength ratio of two Fe XVII transitions, which has recently been shown to finally agree with experiments (arXiv:2201.09070).