Skip to main content

Table 1 PyEvolve benchmarking. Time taken was estimated as time for optimisation. Number of runs per condition ranged from 1 to 5. 1Model – See text for details of the codon and dinuc substitution models; 2Levels – indicates whether Simulated Annealing (SA), Likelihood Function (LF) or BOTH parallelisation levels were used; 3Parallel degree refers to the number of virtual cpu's at the 4SA or LF levels (for the LF level, this is defined per SA virtual cpu); 5the number of likelihood function evaluations made during the optimisation for 610 or 20 sequences, expressed in thousands. See text for details of the data and hardware used.

From: PyEvolve: a toolkit for statistical modelling of molecular evolution

Model 1 Levels 2 Total cpus Parallel degree 3 lfe (1000's) 5 Total Time (minutes) Time (seconds) per 1000 lfe
    SA 4 LF 4 10 6 20 6 10 20 10 20
codon Serial 1 1 1 56 121 124 269 133.06 133.65
  LF 2 1 2 56 121 81 182 86.49 90.16
   4 1 4 56 122 55 130 58.69 63.78
   8 1 8 57 122 41 100 43.99 49.02
   16 1 16 57 120 35 82 36.52 41.02
  SA 2 2 1 57 122 85 178 89.40 87.22
   4 4 1 57 121 58 121 60.19 60.15
   8 8 1 57 122 44 99 46.10 48.83
   16 16 1 58 122 38 88 39.48 43.16
  BOTH 4 2 2 57 125 56 125 59.39 60.03
   8 2 4 58 122 40 89 41.57 43.74
   8 4 2 57 121 39 85 40.92 42.37
   16 2 8 56 121 30 69 32.28 34.47
   16 4 4 58 121 28 63 29.31 31.27
   16 8 2 57 121 31 71 32.70 35.30
dinuc Serial 1 1 1 54 119 17 37 19.22 18.47
  LF 2 1 2 54 119 11 24 12.59 12.29
   4 1 4 54 119 7 16 7.80 7.82
   8 1 8 54 117 5 11 5.30 5.55
   16 1 16 55 119 4 9 4.04 4.41
  SA 2 2 1 53 118 11 23 12.19 11.51
   4 4 1 54 118 7 15 8.32 7.77
   8 8 1 54 118 5 12 5.91 5.89
   16 16 1 53 118 4 10 4.73 4.86
  BOTH 4 2 2 54 118 7 15 8.07 7.69
   8 2 4 54 118 5 10 5.14 5.04
   8 4 2 54 117 5 10 5.61 5.28
   16 2 8 54 118 3 7 3.76 3.74
   16 4 4 54 119 3 7 3.76 3.57
   16 8 2 54 119 4 8 4.13 4.10