BSSN_PUGH Scaling on BlueGene/L

Using a standard benchmark which the team has used across a number of architectures, Cactus, with the PUGH driver, showed a linear speedup for a weak scaling problem to the limit of the CPUs available, and with no detectable loss of scaling efficiency for high processor numbers.

These results strongly suggest that this application would be able to scale to the largest machines available at this time, thus allowing physicists using Cactus to conduct simulations with exceptional accuracy.

Here is the data we used with the gnuplot script to generate the scaling graph of the BSSN_PUGH on BlueGene/L (BGW).