Warning
This page is located in archive.

Python tournament

Each player was called 100 times for each maze. Everything was run on the same machine with minimal outside load. Experiment was repeated with differing number of calls.

Total time in seconds is the time it took for 4000 searches including loading and saving data. Both the program and the data were loaded to RAM drive to speed up the whole process and minimize disk speed influence.

Results

Java tournament

Each player was tested on the maps, which were meant for validation.

The tournament was ranked according to these three criteria:

  1. The total amount of found paths
  2. The total amount of found optimal paths
  3. The total time taken

Tournament settings

The tournament executor has been optimized through the following steps:

  1. Tournament will be run in single thread on a quad core CPU. (To minimize chance of interruption by other applications)
  2. It will be run and averaged 100 times. (Again to minimize the effect of interruptions.)
  3. JVM will be run in -server mode - JVM internal optimization will be run before startup, not during run.
  4. Java Heap fixed to 2GBs of memory. (-Xmx2G and -Xms2G)
  5. Garbage collection before each run (System.gc(); System.runFinalization(); Thread.sleep(5000)) - memory will be emptied before each run and there is enough time for GC
  6. HotSpot compilation before each run (Xcomp) - so there will not be JIT compilation during repeats.

Results

courses/ae4b99rph/labs/maze/tournament_results/start.txt · Last modified: 2013/10/04 13:02 (external edit)