Table 1. The total time (in seconds) needed for computation using FamSeq at one million positions.
| Method | Loops | PU | Pedigree Size | |||||
| 7 | 8 | 9 | 10 | 11 | 12 | |||
| E-S | N | CPU | 13 | 12 | 15 | 16 | 22 | 34 |
| MCMCa | N | CPU | 100,920 | 129,030 | 160,170 | 177,740 | 240,650 | 296,600 |
| Y | CPU | 117,460 | 233,490 | 289,720 | 362,630 | 432,760 | 496,750 | |
| BN | N | CPU | 242 | 605 | 2,003 | 6,483 | 23,404 | 73,485 |
| N | GPUb | 2,472 (150) | 2,907 (169) | 3,312 (239) | 3,856 (397) | 4,519 (946) | 6,452 (2,717) | |
| Y | CPU | 250 | 902 | 2,013 | 6,731 | 22,078 | 70,417 | |
| Y | GPUb | 2,548 (150) | 2,986 (170) | 3,123 (239) | 3,602 (399) | 4,396 (954) | 6,605 (2,726) | |
PU: processing unit; E-S: Elston-Stewart algorithm; MCMC: Markov chain Monte Carlo algorithm; BN: Bayesian network algorithm; N: No, inbreeding loops are not considered; Y: Yes, inbreeding loops are considered.
a
We called only 100,000 variants due to excessive running time for the MCMC algorithm. The time shown here is 10× the time required to call 100,000 variants.
b
The time in parentheses is the GPU computing time.