The benchmark was performed on an 8-node p575-IH with 8 dual-core POWER5+
processors @ 1.9 GHz. A High Performance "Federation" switch (HPS) was used for
MPI communication. The maximal bandwidth of the HPS is about 2 GB/s and the
latency is ca. 2 microsec.

- MPI_Put() in program mod1k gave strange results for messages > 100,000 bytes.
  It is not clear whether this has to be attributed to MPI_Wtime() or something
  more fundamental is wrong.
- in mod2f 3 errors are reported for the smallest problem size. This is just
  beacause the correctness test is a bit too restrictive here. 
