

The gap between curves is significant in the above plot; it shows: (a) that
there's always an advantage to using 4 procs, even for 3K atoms, (b) that
systems over about 4K atoms would benefit from using 8 procs, (c)
the small gap between 2 and 1 proc/host, and (d) GbNIC only offers any
real advantage for large systems using 8 procs. These are all significant
improvements in parallel scaling performance, especially for the 2
proc/host case. The GbNIC 8 proc results for the 2.2.13 kernel (OLD)
are shown to illustrate the change.

The above shows the 8 proc results for optimizing the kernel parameter
tcp_delack_strategy on Gigabit ethernet (GbNIC); note the near
convergence of 2 and 1 proc/host curves, such that at 0 and 1 the times
are ranked by molecular system size.

As for the previous plot, but for fast ethernet (100baseT).
| Nproc | PM Ewald 2/host |
|---|---|
| N | Time |
| Speedup |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.95 | 8.91 | 7.10 | 6.98 |
| 1.94 | 1.95 | 1.94 | 1.97 | |
| 4 | 4.89 | 4.79 | 3.83 | 3.76 |
| 3.55 | 3.62 | 3.60 | 3.66 | |
| 8 | 2.97 | 2.74 | 2.28 | 2.13 |
| 5.85 | 6.34 | 6.04 | 6.47 | |
| M50.dat | ||||
| 2 | 7.92 | 7.89 | 6.38 | 6.25 |
| 1.94 | 1.95 | 1.94 | 1.98 | |
| 4 | 4.41 | 4.28 | 3.61 | 3.41 |
| 3.48 | 3.59 | 3.43 | 3.63 | |
| 8 | 2.66 | 2.46 | 2.09 | 1.93 |
| 5.78 | 6.25 | 5.92 | 6.41 | |
| M20.dat | ||||
| 2 | 3.36 | 3.35 | 2.67 | 2.68 |
| 1.94 | 1.95 | 1.94 | 1.93 | |
| 4 | 1.89 | 1.85 | 1.52 | 1.47 |
| 3.45 | 3.52 | 3.41 | 3.52 | |
| 8 | 1.37 | 1.24 | 1.02 | 0.93 |
| 4.76 | 5.26 | 5.08 | 5.56 | |
| bx30.dat | ||||
| 2 | 1.74 | 1.75 | 1.47 | 1.46 |
| 1.94 | 1.93 | 1.94 | 1.95 | |
| 4 | 1.06 | 1.03 | 0.85 | 0.84 |
| 3.18 | 3.28 | 3.34 | 3.40 | |
| 8 | 0.88 | 0.77 | 0.66 | 0.59 |
| 3.84 | 4.37 | 4.34 | 4.82 | |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.81 | 8.94 | 6.98 | 6.94 |
| 1.94 | 1.91 | 1.94 | 1.95 | |
| 4 | 5.06 | 4.96 | 3.88 | 3.80 |
| 3.38 | 3.45 | 3.49 | 3.56 | |
| 8 | 3.09 | 2.87 | 2.30 | 2.33 |
| 5.53 | 5.96 | 5.89 | 5.81 | |
| M50.dat | ||||
| 2 | 7.92 | 7.95 | 6.32 | 6.21 |
| 1.94 | 1.93 | 1.94 | 1.97 | |
| 4 | 4.43 | 4.38 | 3.50 | 3.48 |
| 3.47 | 3.51 | 3.50 | 3.52 | |
| 8 | 2.78 | 2.57 | 2.12 | 2.10 |
| 5.53 | 5.98 | 5.78 | 5.84 | |
| M20.dat | ||||
| 2 | 3.33 | 3.41 | 2.67 | 2.66 |
| 1.94 | 1.89 | 1.94 | 1.95 | |
| 4 | 1.95 | 1.89 | 1.52 | 1.47 |
| 3.31 | 3.42 | 3.41 | 3.52 | |
| 8 | 1.43 | 1.27 | 1.05 | 0.94 |
| 4.52 | 5.09 | 4.93 | 5.50 | |
| bx30.dat | ||||
| 2 | 1.74 | 1.77 | 1.45 | 1.45 |
| 1.94 | 1.91 | 1.94 | 1.94 | |
| 4 | 1.10 | 1.03 | 0.87 | 0.83 |
| 3.07 | 3.28 | 3.25 | 3.39 | |
| 8 | 0.92 | 0.79 | 0.68 | 0.59 |
| 3.66 | 4.29 | 4.16 | 4.75 | |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.83 | 8.84 | 7.01 | 6.92 |
| 1.94 | 1.94 | 1.94 | 1.97 | |
| 4 | 5.51 | 4.88 | 4.38 | 3.74 |
| 3.11 | 3.51 | 3.10 | 3.64 | |
| 8 | 5.48 | 3.71 | 3.97 | 2.76 |
| 3.13 | 4.62 | 3.43 | 4.93 | |
| M50.dat | ||||
| 2 | 7.83 | 7.92 | 6.29 | 6.22 |
| 1.94 | 1.92 | 1.94 | 1.96 | |
| 4 | 4.97 | 4.34 | 3.97 | 3.52 |
| 3.06 | 3.50 | 3.07 | 3.47 | |
| 8 | 5.19 | 3.40 | 3.81 | 2.55 |
| 2.93 | 4.47 | 3.20 | 4.79 | |
| M20.dat | ||||
| 2 | 3.35 | 3.35 | 2.68 | 2.65 |
| 1.94 | 1.94 | 1.94 | 1.96 | |
| 4 | 2.54 | 2.03 | 2.10 | 1.63 |
| 2.56 | 3.20 | 2.48 | 3.19 | |
| 8 | 4.00 | 2.24 | 2.87 | 1.60 |
| 1.62 | 2.90 | 1.81 | 3.25 | |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.87 | 8.90 | 6.89 | 6.83 |
| 1.94 | 1.93 | 1.94 | 1.96 | |
| 4 | 5.65 | 5.12 | 4.40 | 3.95 |
| 3.05 | 3.36 | 3.04 | 3.38 | |
| 8 | 5.35 | 3.69 | 3.94 | 2.77 |
| 3.22 | 4.66 | 3.39 | 4.83 | |
| M50.dat | ||||
| 2 | 7.74 | 7.81 | 6.31 | 6.18 |
| 1.94 | 1.92 | 1.94 | 1.98 | |
| 4 | 5.18 | 4.42 | 4.00 | 3.53 |
| 2.90 | 3.40 | 3.06 | 3.47 | |
| 8 | 5.29 | 3.33 | 3.77 | 2.64 |
| 2.84 | 4.51 | 3.25 | 4.64 | |
| M20.dat | ||||
| 2 | 3.31 | 3.35 | 2.65 | 2.64 |
| 1.94 | 1.92 | 1.94 | 1.95 | |
| 4 | 2.58 | 2.01 | 2.08 | 1.60 |
| 2.49 | 3.19 | 2.47 | 3.21 | |
| 8 | 3.85 | 2.26 | 2.70 | 1.60 |
| 1.67 | 2.84 | 1.90 | 3.21 | |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.97 | 8.90 | 6.98 | 6.83 |
| 1.94 | 1.96 | 1.94 | 1.98 | |
| 4 | 6.01 | 5.28 | 4.75 | 4.18 |
| 2.90 | 3.30 | 2.85 | 3.24 | |
| 8 | 6.72 | 4.64 | 4.91 | 3.49 |
| 2.59 | 3.75 | 2.76 | 3.88 | |
| M50.dat | ||||
| 2 | 8.00 | 7.81 | 6.50 | 6.29 |
| 1.94 | 1.99 | 1.94 | 2.00 | |
| 4 | 5.42 | 4.70 | 4.42 | 3.83 |
| 2.86 | 3.30 | 2.85 | 3.29 | |
| 8 | 6.34 | 4.34 | 4.68 | 3.31 |
| 2.45 | 3.58 | 2.69 | 3.81 | |
| M20.dat | ||||
| 2 | 3.35 | 3.35 | 2.67 | 2.67 |
| 1.94 | 1.94 | 1.94 | 1.94 | |
| 4 | 3.00 | 2.30 | 2.49 | 1.90 |
| 2.17 | 2.83 | 2.08 | 2.73 | |
| 8 | 5.11 | 3.16 | 3.69 | 2.30 |
| 1.27 | 2.06 | 1.40 | 2.25 | |
| Nproc | PM Ewald 2/host | PM Ewald 1/host | Spherical 2/host | Spherical 1/host |
|---|---|---|---|---|
| M100.dat | ||||
| 2 | 8.88 | 9.07 | 7.07 | 6.89 |
| 1.94 | 1.90 | 1.94 | 1.99 | |
| 4 | 6.26 | 5.78 | 4.89 | 4.43 |
| 2.75 | 2.98 | 2.80 | 3.10 | |
| 8 | 7.10 | 5.35 | 5.15 | 4.22 |
| 2.43 | 3.22 | 2.66 | 3.25 | |
| M50.dat | ||||
| 2 | 7.97 | 8.17 | 6.31 | 6.29 |
| 1.94 | 1.89 | 1.94 | 1.95 | |
| 4 | 5.79 | 5.23 | 4.57 | 4.03 |
| 2.67 | 2.96 | 2.68 | 3.04 | |
| 8 | 6.77 | 4.93 | 4.92 | 3.89 |
| 2.28 | 3.14 | 2.49 | 3.15 | |
| M20.dat | ||||
| 2 | 3.33 | 3.46 | 2.69 | 2.69 |
| 1.94 | 1.87 | 1.94 | 1.94 | |
| 4 | 3.15 | 2.52 | 2.52 | 1.98 |
| 2.05 | 2.56 | 2.07 | 2.64 | |
| 8 | 5.18 | 3.42 | 3.75 | 2.57 |
| 1.25 | 1.89 | 1.39 | 2.03 | |