TGB Crowd Test 3 & 4

So I needed a break from my other projects and decided to finally knock out my next installment of my crowd generator. There were two specific things I wanted to do: The first was to remove one layer of my runner object by combining the skin layer and the outline of the skin into one imagemap (duh). Considering each runner in the test consisted of 2-5 layers/objects (depending on randiness of course), it made sense to combine the 2 "always-on" layers/objects into one. That was the only change in Crowd Test 3, so no need to really publish a blog about it, it was only about 10 minutes worth of work in a photo editor and some block qoutes to kill the extra object. But that little change lead to an obvious performance increase (see graphs below). However, it didn't test the true performance bottleneck: Object creation/destruction. For that, I went and founded my re-engineering based on Stephen Zepp's recommendation from my Crowd Test 2 blog: To load the objects into a SimSet, then instead of deleting each object at the end of the run, I simply move them back to start to run again.

See the change for yourself:
You'll notice the object handling bottleneck show up at 0:32 and 0:41 as I increase the saturation in two steps. 0:32 is an increase from 6546 objects to 8183 objects, and 0:41 is a further increase to 10228 objects. You'll notice that once the object creation is completed, the FPS stables and runs consitant at the given saturation level.

Arbitrary Screenshot!


And now for the data!
Crowd001 - The first and 2nd versions of the Crowd Test use the same object handling code, so both versions are represented here.

Crowd003 - Crowd Test 3 was just a reduction in layers in the runner, reducing each runner by 1 object.

Crowd004 - Crowd Test 4 contains the new object handling code that moves runners back to start rather than destruction and recreation.

A table showing the revisions:


And now some chart fun:






Psuedocode showing change in object creation:

function newRunner()
{
  %runner = new t2dAnimatedSprite() 
  {
    ...object code...
    Worldlimit is set to NULL // changed from KILL
    %runner.runSpeed=... // stores objects randomly generated speed for later use:
    $runnerSimSet.add(%this); // first of 2 SimSets, this one is for tracking 
                              // actual runner count

  }
  ... // the other objects are created 
      // just like Crowd Test 2, with 
      // the base and skin object combined 
      // to 1 and each object had some vars 
      // removed as well.

}

New code for handling world limit:

function runnerClass::onWorldLimit(%this)
{
  $runnerFinishedSimSet.add(%this);
  
  // $saturationArray is an array of graded runner saturations
  // $runnerSaturation is an arbirary value from 1(conjested)-10(light)

    // if more runners are out than governed by the current saturation
    // setting, then delete them  
  if($runnerSimSet.getCount()>$saturationArray[$runnerSaturation]) 
    %this.safeDelete();
  else
  {
    %this.setPositionX("-70") //reset to start
    %this.setLinearVelocityX(%this.runSpeed); // make sure he runs at orig speed
  }
}

Code to call the generator

function levelLoop()
{
   if($runnerSimSet.getCount()<$saturationArray[$runnerSaturation])
     newRunner(); // if there aren't enough object to satisfy
                  // the saturation requirement, then make them
   $runnerCounter = getRandom($runnerSaturation,$runnerSaturation*5);
   schedule($runnerCounter,0, "levelLoop");
}