Table 2 Execution time scaling of a Wiener filter reconstruction on a grid of size \(8192 \times 8192\)
From: d2o: a distributed data object for parallel high-performance computing in Python
\(\#\mathrm {nodes}\) | 1 | 1 | 1 | 1 | 2 | 4 | 8 | 16 | 32 |
|---|---|---|---|---|---|---|---|---|---|
\(\#\mathrm {processes:~} n\) | 1 | 2 | 3 | 4 | 8 | 16 | 32 | 64 | 128 |
t [S] | 1618 | 622.0 | 404.2 | 364.2 | 181.7 | 94.50 | 46.79 | 18.74 | 8.56 |
\(s_n = 2t_2/t_n\) | 0.769 | 2.00 | 3.08 | 3.42 | 6.85 | 13.2 | 26.6 | 66.4 | 145 |
\(q_n = {1/(1+\log (\frac{n}{s_n}))}\) | 0.900 | 1.00 | 1.01 | 0.94 | 0.94 | 0.92 | 0.93 | 1.02 | 1.06 |