The experiments have been executed for different problem sizes (n = 5, 7, 30). The harder problems like SAT or Fibonacci have only been executed for n = 5,7. The results of the different problem sets can be found in the corresponding directories named queryn for n = 5,7,30.

Since every for every single run different queries have been generated, some outliers have been investigated and reexecuted with all different strategies and processor bounds. Those can be found in the *_redone directories. Note that the start states are permuted for each run, so the results have not been reproduced exactly.

The output of the experiments can be found in the files named queryn (for corresponding n). From this output, *.dat files have been generated that are space-separated value tables. Every line is a transition step. Column 1 are the applicable rules, column 2 the applied rules and column 3 the store size. Note that the output of the store in our experiments always contains an element '[]', i.e. the store size actually is increased by one inexistent element. We have not purged this element from our results. The last transition step leads to a final state, hence there are no applicable and applied rules (NA) but only a store size.

From the tables, GNUPlot plots have been generated (*.gnuplot files) that yield *.png files. The *out.tex/*out.pdf files gather all plots in one file and prints the queries and some transitions.

From the *.dat files, we also generated the table.tex files that summarize the results. We have used an R script for this purpose. For every problem instance, there are three lines in the table for the three parameters from the *.dat files (applicable, applied, store size). We computed arithmetic mean (mean1), geometric mean (mean2), minimum, maximum, median, the number of transition steps and the number of zeros (which is only interesting for the applied line, as it sums up the transition steps where no rule has been applied). In the table, those empty transitions are not purged from the number transition steps. In the paper, we have excluded those empty steps from our considerations.


