MPI Timing Comparisons

Figure 1 below demonstrates, at least for the referenced code, that both the PGI and GNU compilers generate executables with comparable performance.

Figure 2 demonstrates the performances of the same code on SCV's IBM pSeries, IBM Bluegene, Pentium III Linux Cluster, and the Katana AMD Opteron CLuster. Please note that the purpose of these figures is to provide a reference point. It is by no means an indication of a general trend as many factors (algorithm, compiler, parallel paradigm, etc.) can affect the performance of a code on a specific machine.  

Figure 1. Performance of an MPI code Compiled with PGI and GNU compilers.  
Figure 2. MPI timing comparison of the four SCV computer systems.

 
Figure 3. MPI timings of a Boundary Element code for 3 SCV computer systems.