OpenMP benchmark on IBM p690 cluster (1.7 GHz POWER4+ processors)
Remarks:
- mod1a  : On 16 processors the performance was so low that the repeat
           count for small loops had to be lowered. Also the performance
           for 16 processors fluctuates wildly.

- mod1b  : The xlf compiler appears to have problems with allocating 
           arrays of length 10,000,000 resulting in the following message:

           1525-108 Error encountered while attempting to allocate a data
           object.  The prog ram will stop.

           So, there are no results for this program.

- mod1f  : On 16 processors a segmentation fault occurs.

- mod2a  : The results for the dotproduct implementation for N > 100
           are incorrect when running on more than one processor.

- mod2d  : An exeutable is produced without any warning but the execution
           stalls: it keeps running indefinitely without producing a
           result.

- mod2i  : For nprocs > 1 the program gives a segmentation fault or
           stalls indefinitely without warning.
