SNPs | Dedicated CPU cores (P) | Time per iteration | Time for read-in + setup C | Time for solving MME | Total memory usage, GB |
---|---|---|---|---|---|

30,000 | 1 | 6 hr 44 min | 5 hr 3 min | 1 hr 41 min | 3.73 |

360,000a | 1 | 151 days | 30 days | 121 days | 537 |

360,000 | 720 | 8 hr 50 min | 1 hr 49 min | 7 hr 0 min | 831 |

Speedup | 416 | 399 | 412 | — | |

Efficiency | 57.3% | 55.4% | 57.7% | — |

Speedup is calculated as the ratio between the computing time on a single CPU core and the computing time on

*P*= 720 CPU cores. Parallel efficiency is defined as the ratio between the actual speedup and the maximum speedup, which is equal to the number of dedicated CPU cores.Results for a data set with 360,000 SNPs on a single CPU core are estimated based on results of the data set with 30,000 SNPs, cubic complexity of the solving algorithm, and square complexity of the setup of matrix^{a}**C**and memory usage.