PDF version of this document
As part of speed-up through code modification, an adaptive precision approach and the like deserve to be looked into a little further. As the figures in the earlier mentioning of this issue show (e.g. Figure cap:Automatic-precision-and), the rate of convergence is changed as the process goes on and so is the speed of the algorithm. There is more to be investigated to ensure the approach invariantly results in gains. It is also worthwhile to see if the choice of tolerance can be made more preferable, based on some empirical evidence. For instance, experiments with varying values for tolerance might be helpful.