Fig. 4.

Running times for the Hill dataset with 5151 individual alignments. (a) Total running times when instances are sorted by individual running times. For any fraction x%, we calculate the total running time of the x%, instances for which the alignment was computed faster than for any of the remaining instances. For example at 50% one can find the running time that was needed to compute the 50% fastest instances. For each algorithm, instances were sorted separately. Note the logarithmic y-axis. (b) Individual running times for the 200 slowest instances of the classical DP algorithm. Instances are sorted by their running time for the classical DP algorithm. One can see that running times of the classical DP are outperformed by that of the sparse DP