This concludes the discussion of the results of our experiments using the parallel implementation. In these experiments we abused the ability for granularity control a bit, since we used it to demonstrate that the communication determines the total execution time as the problem gets larger as predicted by Amdahl's Law. We simulated making the problem larger by having it subdivided in smaller pieces. We can see that this is equivalent to making the problem larger in this application. We can just pretend that the small pieces are really very hard pieces to execute. Of course, the normal use of the granularity control would be to tune the processor usage. To finish this paper, I will give the charts with the results of the other experiments as promised. You will see that the result are less explicit than in the earlier presented charts. They would probably become clearer if we allowed the test to be more extensive and less randomized. If we compare the performances of the different approaches we also find some inconsistencies, but they should also disappear with more extensive testing since another test (which wasn't parallelized and thus not discussed in this paper) showed the results we expected.