refactor loop and if structure

[huppd]

Refactoring loop and if-statement structure to improve OpenACC performance.

[huppd]

For our CPU measurements it only had a small impact:
Reference:
MAX AVG
1.937 1.390
1.947 1.391
1.953 1.390

Refactor:
MAX AVG
1.965 1.404
1.968 1.405
1.970 1.406

On the NEC it seems to have even less of an impact:
Reference:
MAX AVG
112.410, 110.948
112.694, 111.067
113.348, 111.763

Refactor:
MAX AVG
112.703, 111.195
112.848, 111.124
113.095, 111.495

For GPU the kernel took before 17.6ms and could be split into two which take 3.59ms and 235μs.

Unfortunately, these are all different comparison, so one cannot compare across platform.
Is that something we could consider to bring back to the main branch?