of the theoretical
communication bandwidth of 110 MB/s for the NPACI SP-2.
The above exercise also highlights the importance of the optimal
distribution of data among the processors such that the communication
between them is minimized.
It is clear that case 2 is the best case to run for this problem
as it requires no interprocessor communication and is hence the
fastest and case 3 is the worst as it requires the maximum communication
time. The cases 1, 4 and 5 seem to take almost the same time
here but may be different on different machines depending how the compiler
handles the corresponding HPF commands (i.e. the other CSHIFT command
and the triplet notation).