Per node summary table

L , Wilson, DWF4, Staggered, NaiveStag

8 , 90, 933, 38, 23
12 , 403, 1688, 178, 113
16 , 188, 1647, 449, 295
24 , 947, 1574, 674, 553
32 , 931, 1371, 718, 643

Memory Bandwidth

Bytes, GB/s per node
786432, 40.271620
12582912, 433.611792
63700992, 905.374321
201326592, 1114.979152
491520000, 1180.241898


Communications

Packet bytes, direction, GB/s per node


GEMM

 M, N, K, BATCH, GF/s per rank fp64
16, 8, 16, 4096, 693.316363
16, 12, 16, 4096, 657.277058
16, 16, 16, 4096, 711.992616
32, 8, 32, 4096, 821.084324
32, 12, 32, 4096, 1279.852719
32, 16, 32, 4096, 2647.096674
64, 8, 64, 4096, 2630.192325
64, 12, 64, 4096, 3338.071321
64, 16, 64, 4096, 3950.899281
16, 8, 256, 4096, 1638.362501
16, 12, 256, 4096, 2377.502234
16, 16, 256, 4096, 3048.328833
32, 8, 256, 4096, 2917.384276
32, 12, 256, 4096, 4103.085151
32, 16, 256, 4096, 5102.971860
64, 8, 256, 4096, 3222.258206
64, 12, 256, 4096, 4619.456391
64, 16, 256, 4096, 5847.916650
8, 256, 16, 4096, 1728.073337
12, 256, 16, 4096, 2356.653970
16, 256, 16, 4096, 2676.876038
8, 256, 32, 4096, 2611.531990
12, 256, 32, 4096, 3451.573106
16, 256, 32, 4096, 3966.915301
8, 256, 64, 4096, 3436.248737
12, 256, 64, 4096, 4539.497945
16, 256, 64, 4096, 5307.992323



GEMM

 M, N, K, BATCH, GF/s per rank fp32
16, 8, 16, 4096, 499.017445
16, 12, 16, 4096, 731.543385
16, 16, 16, 4096, 958.800786
32, 8, 32, 4096, 1549.813550
32, 12, 32, 4096, 2147.907502
32, 16, 32, 4096, 2601.698596
64, 8, 64, 4096, 3785.446233
64, 12, 64, 4096, 5116.694843
64, 16, 64, 4096, 6109.345016
16, 8, 256, 4096, 1206.627737
16, 12, 256, 4096, 1809.699599
16, 16, 256, 4096, 2412.014053
32, 8, 256, 4096, 2406.114488
32, 12, 256, 4096, 3605.531907
32, 16, 256, 4096, 4798.444037
64, 8, 256, 4096, 4688.711196
64, 12, 256, 4096, 6990.696301
64, 16, 256, 4096, 9214.749925
8, 256, 16, 4096, 2596.307289
12, 256, 16, 4096, 3439.892562
16, 256, 16, 4096, 3907.201036
8, 256, 32, 4096, 3012.752067
12, 256, 32, 4096, 3904.217583
16, 256, 32, 4096, 4599.047092
8, 256, 64, 4096, 3721.999042
12, 256, 64, 4096, 5098.573927
16, 256, 64, 4096, 6159.080872