Conclusions

The k-point parallelized codes can certainly provide more efficient use of computation cycles as it can be demonstrated in tests cases ()-() and application . There are also other cases, where either the original code does not scale at all to larger number of cores -see test cases (),() , or the original code does not even work on large number of cores (see test case ()) or exterior limitations such as the 12 hour queue on HECToR in application () do not allow large simulations. In all these cases we demonstrated the calculations could be achieved with the k-point parallelized code.