Benchmark Results

The W216 test case was use to benchmark these changes. The table below gives the runtimes on a range of processor counts on HECToR and the resulting speedup:

Table 6: Comparison of W216 runtime before and after rank reordering for load balance

Cores	128	256	512	1024	2048
Before(s)	5998	3499	2448	1569	2565
After(s)	4800	2859	2096	1425	2166
Speedup(%)	25	23	16	10	18