hacker.org

Hey guys, tried to do some pre-processing with the maze. but after i compared the running durations with and without clearing the maze, i got almost the same result.
the following example is level 141; the most unpleasant level for me, so far.

level141 pre-processed: RDDD********************************RRDD
solving time: 8216,9s

level141 normal: RDDD********************************RRDD
solving time: 8202,6s

Btw: The pre-processing itself needs just about a second. I am using a recursive BF-algo that recognizes and avoids the "walls".



Thx, wiz

Hey, try timing each path generation that results in a dead end (ie: a point where any legal move of the available two moves is impossible) in normal mode, then see how many are they and the total time it took you to generate them, if the result is consistent with the delta of the previous totals that you've pasted, then there should be three possibilities :

1. Your algorithm simply didn't go through many paths that resulted in a dead end in normal mode at that level compared to other levels, this could be detected by doing the same test on different levels and seeing if there is a significantly larger difference in the timing ratio, on average.
2. Your algorithm is not making full use of the preprocessed map, or not operating with the preprocessed map in mind at all, this could be easily detected by printing out dead ends (there shouldn't be any)
3. You algorithm is being delayed by some very time consuming operation in both cases, and this is dwarfing the timing advantage you're getting out of preprocessing .

ps. I'm assuming that the exact same algorithm is being used in both cases.