Programming environments and tools for massively parallel computers and large scale applications

The material presented in this thesis covers a wide range of different and diverse topics. Although it might look like the various subjects are rather unrelated there is a thread throughout. It is given by the necessity to evaluate and to achieve performance with IFS a large production weather forecasting code and to look at this problem from various angles. When we set out it was quite unclear what the real issues were. Was it the communication causing unavoidable overheads limiting overall program performance or was it load imbalance that would prevent us from achieving our performance goals? It was still widely believed that the most serious threat to achieving performance would come from communication in a parallel program. Load imbalance was seen as another area that potentially could cause severe problems. After some investigation it turned out that the port of IFS is rather well behaved in these regards; the static work distribution that had been build in by design was sufficient to limit serious problems caused by load imbalances.In addition to the more performance oriented goal of the study the potential use of HPF for applications like IFS has been investigated in much detail.