Table 1.
Average run times over 10 runs of the SpiNNTools algorithms, in order of execution, when executing Conway's Game of Life on various grid sizes, broken down into phases of execution.
| CONWAY'S GAME OF LIFE SIZE | |||||
| Number vertices / cores | 100 | 400 | 900 | 1,600 | 2,500 |
| Number of edges | 800 | 3,200 | 7,200 | 12,800 | 20,000 |
| Number of chips used | 7 | 28 | 57 | 103 | 160 |
| Number of boards used | 1 | 1 | 3 | 3 | 6 |
| MACHINE DISCOVERY (μs) | |||||
| Machine allocator | 8,711,589 | 10,799,373 | 10,175,818 | 12,076,720 | 6,731,204 |
| Machine generator | 8,040,498 | 8,470,635 | 6,963,204 | 5,674,133 | 10,230,590 |
| Virtual chip allocator | 86 | 85 | 85 | 93 | 91 |
| MAPPING (μs) | |||||
| Network specification report | 3,705 | 14,953 | 31,667 | 56,875 | 90,130 |
| Placement | 12,078 | 37,719 | 84,610 | 145,805 | 230,735 |
| Routing | 29,002 | 128,009 | 370,468 | 747,004 | 1,682,054 |
| IP tag allocator | 12,715 | 42,762 | 96,990 | 171,202 | 275,011 |
| IP tag report | 945 | 553 | 624 | 590 | 581 |
| Edge N keys required | 4,164 | 11,085 | 23,426 | 40,706 | 62,953 |
| Routing key allocator | 100,638 | 410,597 | 906,178 | 1,608,309 | 2,515,832 |
| Routing key report | 3,007 | 7,767 | 16,441 | 28,254 | 42,442 |
| Routing table generator | 5,533 | 30,833 | 81,163 | 285,620 | 517,796 |
| Executable type locator | 1,602 | 3,451 | 6,390 | 10,444 | 15,819 |
| Buffer manager creator | 4,280 | 7,453 | 11,580 | 18,214 | 26,184 |
| DATA SPECIFICATION (μs) | |||||
| Data specification writer | 91,863 | 361,230 | 810,060 | 1,447,989 | 2,247,400 |
| Graph provenance gatherer | 3,073 | 8,069 | 16,864 | 30,550 | 45,736 |
| LOADING (μs) | |||||
| Router initialization | 16,681 | 67,168 | 141,138 | 256,470 | 410,147 |
| Executable name gatherer | 5,787 | 17,242 | 36,305 | 65,802 | 100,398 |
| Routing table compression | 2,017,828 | 2,110,796 | 2,287,198 | 2,644,092 | 3,270,905 |
| IP tag loader | 3,617 | 3,352 | 8,953 | 9,757 | 18,219 |
| Load data | 335,054 | 1,411,766 | 3,221,989 | 5,937,052 | 9,586,231 |
| Load executable images | 962,930 | 969,326 | 965,471 | 979,613 | 1,014,863 |
| RUNNING (μs) | |||||
| Runtime updater | 57,257 | 199,705 | 441,409 | 798,570 | 1,281,730 |
| Database writer | 243 | 563 | 1,007 | 1,672 | 2,442 |
| Database notification | 1,311 | 1,161 | 1,287 | 1,400 | 1,213 |
| Application runner | 303,362 | 303,475 | 303,622 | 303,897 | 304,273 |
| Core provenance gatherer | 1,964 | 3,012 | 5,045 | 8,908 | 11,060 |
| Router provenance gatherer | 76,984 | 75,238 | 243,184 | 240,695 | 530,201 |
| Profile data gatherer | 1,681 | 2,955 | 4,763 | 8,535 | 10,945 |
Note that the data recorded starts when the user calls the run function, and stops at the end of the Running phase; in particular, setup, graph creation and data extraction times are not included here as they are not currently recorded by the tools. The number of chips and boards used by the simulations is also reported to show the scaling. These numbers were obtained on a server with 128GB RAM, an Intel Xeon E5520 2.27GHz CPU and an SSD drive, running in the same machine room as the SpiNNaker machine, connected to it via a 1GB/s interface.