This bar chart represents the contrast in performance between the original GCEM3D code (in red), and the parallel CAPO version (in green). This example is a 130-by-66-by-34 case run on an SGI Origin 3000. While the original case did not scale beyond four processors (CPUs), the CAPO version achieved a speedup of 12.4 on 16 CPUs, improving performance of the code by a factor of 8.2. (NASA/IPG)