Re: [Arianne-devel] gaining server speed

[Miguel A. B. L. <mig...@ho...>]

Hi Jo,

I would like to release 0.51 tomorrow.
If you feel confident about that it works correctly, just commit it. :)

Anyway IMHO it doesn't hit the main problem that is Creature.logic
Most of time is wasted there checking if there is a player around and 
calling pathfinding.
I were working on A* pathfinding to fix the issue of no detecting players, 
but I have failed ( again ). I will try again tomorrow. ( I know I can do it 
:P )

On Creatures.logic at the end is resolved the attack and move actions of the 
creature, now I see that this is a terrible mistake and that the resolution 
of the action should be split in a similar way it is done with players.

And about server paralelization, keep in mind that it wasn't done to be 
paralelized, so concurrent access problems may appear due to design that 
need to be fixed on design.

Regards,
Miguel

>From: Jo Seiler <jo....@gm...>
>Reply-To: ari...@li...
>To: ari...@li...
>Subject: [Arianne-devel] gaining server speed
>Date: Tue, 20 Jun 2006 21:53:21 +0200
>
>Hello,
>
>I'm not sure how to understand the part about feature freeze in last  
>meeting notes.
>So here's a question on if I should submit it or not.
>
>I implemented a simple task switcher in the RulesProcessor's endTurn:
>
>	synchronized public void endTurn() {
>         Log4J.startMethod(logger, "endTurn");
>         long start = System.nanoTime();
>         int aktTurn = getTurn();
>         try {
>             for (RespawnPoint point : respawnPoints) {
>                 point.logic();
>             }
>             switch (aktTurn % 5) {
>                 case 4:
>                     for (RespawnPoint point : respawnPoints)
>                         point.checkRespawn(aktTurn);
>                     break;
>                 case 3:
>                     for (PlantGrower plantGrower : plantGrowers)
>                         plantGrower.regrow(aktTurn);
>                     break;
>                 case 2:
>                     for (Corpse corpse : corpses)
>                         corpse.logic(aktTurn);
>                     break;
>                 case 1:
>                     for (Blood blood : bloods)
>                         blood.logic(aktTurn);
>                     break;
>                 default:
>                     for (Door door : doors) {
>                         door.logic(aktTurn);
>                     }
>                     break;
>             }
>             scripts.logic();
>         } catch (Exception e) {
>             logger.error("error in endTurn", e);
>         } finally {
>             logger.debug("End turn: " + (System.nanoTime() -  start) / 
>1000000.0 + " (" + (aktTurn % 5) + ")");
>             Log4J.finishMethod(logger, "endTurn");
>         }
>     }
>
>and changed the classes that are called from here.
>
>I changes Respawn point to split between NPC.logic() and respawn  logic and 
>changed all functions called in the switch statement so  they work in the 
>expected way even if they are not called every turn.
>
>The result of the work based on debug output in the same timeframe  since 
>server restart is:
>
>before the change:
>
>27163 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>48.763
>27475 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>59.88
>27747 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>30.956
>28042 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>25.506
>28367 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>49.897
>28643 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>25.122
>28944 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>25.514
>29245 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>25.56
>29577 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>56.59
>29861 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>37.79
>30162 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>36.788
>30479 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>52.618
>
>after the change:
>
>27230 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>23.367 (2)
>27534 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>26.395 (3)
>27853 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>44.554 (4)
>28133 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>23.71 (0)
>28434 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>23.408 (1)
>28734 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>23.63 (2)
>29051 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>39.153 (3)
>29340 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>27.841 (4)
>29636 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>22.706 (0)
>29951 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>37.32 (1)
>30237 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>22.881 (2)
>30538 DEBUG  games.stendhal.server.StendhalRPRuleProcessor  - End  turn: 
>23.527 (3)
>
>it's not a major breakthrough but should give us a little air to  breathe 
>and the possibility to optimize it even more.
>
>I checked all the functionality that is involved with the functions  and 
>all is working normal it seems.
>
>So, do we have a code freeze or shall I submit it?
>
>Regards,
>Jo
>
>
>
>
>




>_______________________________________________
>Arianne-devel mailing list
>Ari...@li...
>https://lists.sourceforge.net/lists/listinfo/arianne-devel