An Optimized Parallel FDTD Topology for Challenging Electromagnetic Simulations on Supercomputers

Table 3

Comparisons of virtual topology, amount of communication, and computation time on NSCC-SZ.

CPU cores	Virtual topology ()	Amount of communication	Computation time (in seconds)
48	4 × 6 × 2	4320000	1271.02
	6 × 4 × 2	4320000	1256.12
60	5 × 6 × 2	4680000	1055.02
	3 × 10 × 2	5400000	1044.47
	3 × 5 × 4	6480000	1214.09
	10 × 2 × 3	6480000	1117.17
96	6 × 8 × 2	5760000	596.99
	8 × 4 × 3	6480000	637.09
	6 × 4 × 4	7200000	604.82
	4 × 6 × 4	7200000	641.13
120	6 × 10 × 2	6480000	492.96
	10 × 6 × 2	6480000	752.48
	8 × 5 × 3	6840000	510.25
	5 × 8 × 3	6840000	566.13
240	10 × 12 × 2	8640000	286.12
	8 × 10 × 3	8640000	296.48
	10 × 8 × 3	8640000	328.07
	12 × 10 × 2	8640000	417.3
360	12 × 10 × 3	10080000	197.46
	10 × 12 × 3	10080000	278.73
	8 × 15 × 3	10440000	181.46
	12 × 15 × 2	10440000	297.34
480	10 × 12 × 4	11520000	147.48
	12 × 10 × 4	11520000	156.53
	15 × 8 × 4	11880000	157.57
	15 × 16 × 2	11880000	162.54
	12 × 8 × 5	12240000	161.85
	8 × 12 × 5	12240000	165.3