I made an attempt to implement DeWitter's game loop (http://dewitters.koonsolo.com/gameloop.html) in C#, but during testing it seemed that my implementation is inaccurate. I am aware that it will never be perfectly accurate but the measured results don't look quite right.
I am not sure if the problem is with the loop itself or simply with the way it is measured.
public abstract class Game
{
private Stopwatch Clock;
private TimeSpan nextTick;
private int _targetTicksPerSecond = 30;
private TimeSpan _targetFrameLength;
private int _maxFrameSkip = 5;
private void GameLoop()
{
int framesSkipped = 0;
float interpolation = 0f;
nextTick = Clock.Elapsed; // Clock initialized by caller method
_targetFrameLength = TimeSpan.FromMilliseconds(1000.0 / (double)_targetTicksPerSecond);
while( _window.IsOpen() )
{
framesSkipped = 0;
while (Clock.Elapsed > nextTick && framesSkipped < _maxFrameSkip)
{
Window.DispatchEvents();
Update();
nextTick += _targetFrameLength;
framesSkipped++;
}
interpolation = (float)((Clock.Elapsed + _targetFrameLength - nextTick).TotalMilliseconds / _targetFrameLength.TotalMilliseconds);
Draw(Renderer, interpolation);
}
}
private Stopwatch testWatch;
private TimeSpan prev;
protected virtual void Update()
{
Debug.WriteLine((testWatch.Elapsed - prev).ToString());
prev = testWatch.Elapsed;
}
// empty method in derived test class
protected abstract void Draw(RenderTexture renderer, float elapsed);
}
A typical output looks like this:
00:00:00.0281242 | 00:00:00.0331724
00:00:00.0336379 | 00:00:00.0306271
00:00:00.0304994 | 00:00:00.0309980
00:00:00.0343010 | 00:00:00.0329622
00:00:00.0331906 | 00:00:00.0320139
00:00:00.0275725 | 00:00:00.0276520
00:00:00.0323255 | 00:00:00.0309111
00:00:00.0283663 | 00:00:00.0338622
00:00:00.0431601 | 00:00:00.0312895
00:00:00.0214689 | 00:00:00.0302318
00:00:00.0331552 | 00:00:00.0343154
00:00:00.0304954 | 00:00:00.0309841
I tried the loop with different measuring approaches using StopWatch.ElapsedMilliseconds, StopWatch.Elapsed.TotalMilliseconds etc, but over the long run, measured results are pretty much the same. All the methods called are empty except for SFML's DispatchEvents() (though removing it made no difference).
I'd really like any pointers on what I might have done wrong with the loop and/or measuring and how it might be improved.
