Fig. 1

Benchmarking results for SFREEMAP. The curve labeled SIMMAP-1 was obtained using a single simulation, SIMMAP-10 using 10 simulations and SIMMAP-20 using 20 simulations. a) Increase in execution times with numbers of trees (256 terminals) during the mapping of a single 4-state character. Average performances were estimated from slopes (±SE), computed using R’s linear model (lm function), as seconds spent on each tree (SIMMAP-1 8.64 ± 0.11 s/tree, SIMMAP-10 14.80 ± 0.11 s/tree, SIMMAP-20 21.77 ± 0.12 s/tree, SFREEMAP 1.78 ± 0.02 s/tree). Intercepts were forced to 0 and determination coefficients (R2) exceeded 0.99 in all experiments. b) Increase in execution times with number of taxa for the reconstruction of a 4-state character on a single tree. Average performances were estimated from slopes (±SE) as seconds added per taxon (SIMMAP-1 0.038 ± 0.001 s/taxon, SIMMAP-10 0.063 ± 0.001 s/taxon, SIMMAP-20 0.088 ± 0.001 s/taxon, SFREEMAP 0.0066 ± 0.0002 s/taxon). Intercepts were forced to 0 and determination coefficients (R2) were over 0.97 in all experiments. c) Exponential increase in execution time with number of states. Experiments were run using one tree with 256 terminals. Lines were fit using the function nls, also implemented in R, assuming the formula y = bx^a. We verified goodness-of-fit by adjusting a linear regression to log-log transformed data. R2 exceeded 0.89 in all experiments