I would not pass the time since last frame. The problem is with the "time tick resolution". Like if you are getting 500fps and there are only 1000 ticks, you are passing either 1 or 2 ticks as "time since last interval". You can see how there is a 100% difference between 1 and 2! 50% between 2 and 3. That little difference shows up as herky-jerky movement.

Pass the current time. Not all objects will use time the same way. If an object needs interval time, it can calculate that itself. Or you can keep TIME and INTERVAL_TIME as global variables so objects can use either one and only calculate them once. whatever you want.

here is a timer class

class CTimer{  protected:    double m_Frequency;    __int64 m_StartClock;    float m_FrameTime;  float m_FrameStart;  float m_FrameEnd;  public:    float GetFrameTime() { return m_FrameTime; }    double GetTime();    void Init();  void Update();  };

#include <windows.h>#include "timer.h"  double CTimer::GetTime(){  __int64 EndClock;    QueryPerformanceCounter((LARGE_INTEGER*)&EndClock);    return (double)(EndClock-m_StartClock)*m_Frequency;}  void CTimer::Init(){  __int64 rate;    // Get the performance frequency  QueryPerformanceFrequency((LARGE_INTEGER*)&rate);    // Invert it so we can multiply instead of divide  m_Frequency = 1.0/(double)rate;    // Get the start time  QueryPerformanceCounter((LARGE_INTEGER*)&m_StartClock);    m_FrameTime  = 0.0f;  m_FrameStart = (float)GetTime();  m_FrameEnd   = 0.0f;}  void CTimer::Update(){  // Cap the frame rate to the timer frequency  // The timer frequency is normally always faster than the frame rate  do {    m_FrameEnd = (float)GetTime(); // Get the end frame time  } while( m_FrameEnd == m_FrameStart);    //   m_FrameTime  = m_FrameEnd - m_FrameStart; // Get the elapsed time  m_FrameStart = m_FrameEnd; // Set the end time to be the next start time}

here is how you can use it

CTimer timer;
timer.Init();

// game loop
{
// update the timer
timer.Update();

// update sprite animation at 1 frame per second
sprite.updateAnimation(1.0, timer.GetFrameTime());
}

	QueryPerformanceFrequency(&frame_delay);	frame_delay.QuadPart/=85;//set at 85 fps	QueryPerformanceCounter(&ticks);	QueryPerformanceCounter(&last_frame);//------------------------GAME LOOP-------------------------------	do	{		do		{			while(PeekMessage(&Msg, NULL, 0, 0, PM_REMOVE))				{				TranslateMessage(&Msg);				DispatchMessage(&Msg);				switch(Msg.message)				{				case WM_LBUTTONDOWN:				break;				case WM_QUIT:				game_over=true;				}			}		QueryPerformanceCounter(&ticks);		}while(ticks.QuadPart<last_frame.QuadPart+frame_delay.QuadPart);		QueryPerformanceCounter(&last_frame);			Inputs(Game);		UpdateWorld(Game);		HitDetection(Game);		Refresh(Game);	}while(!game_over);

and i have constant deltas for 85 fps as one would expect. to support constant framerate that are determined at start up i could easily gives those deltas a coefficients. now if im going to be locking the framerate at the refresh rate in this manner, is it worth worying about rare instances of poor performance? (this is a 2d game anyway)

[edited by - unliterate on May 27, 2004 9:49:29 PM]

well, you only want to call SDL_GetTicks() ONCE. Otherwise some of your objects will move ahead of / more than / less than other objects. You want them all to move the same amount each frame so give everybody the same time. You also want to call it only once because it cuts down on function calls as well, especially if you have thousands of objects that need the time. By making time a global, you can also cut down on passing it around every loop as too.

lei, (i know you werent adressing me but) would that apply if the framerate is locked? that doesnt seem to take account the only problem applicable to the way im doing things, and that problem would be certain game function calls taking way longer than expected (because of something going on in the game, like lots of explosions or something). also, by ''time'' do you mean the diference in between the current time and the last time check?

i don''t want to sound like my way is the best way, but the way *I* do it is just to let the gameloop run as fast as it can and update everything as fast as i can. It doesn''t matter if the game is bogged down at 10 FPS or is running like the Devil at 500 FPS, objects still all move and such at the same rate. Instead of saying an object should move X pixels per frame, you make it pixels per second.

I pass the current absolute time to my objects. Different objects handle that time differently. Some completely ignore it. Some keep track of the last update and calculate the difference. Some only care about the absolute time (if i want something to happen at exacty 10 seconds into the game or something), some need the difference (moving objects), some need the time since they were born (particles), some need the time relative to their whole lifespan (conditional animations), etc.

Now to (not) answer your question, i''m not sure how to get the framerate to be nice while doing other CPU intensive activities like AI besides using threads. I haven''t started AI functions for my project. I''m having to much fun designing special FX :-) But when the screen gets really bogged down, the framerate slows to a crawl, but objects still move at the same pace.

Haven't read everybody's responses, but this is how I do it, and it works fine for me.

1. set a time on which the game objects move at "normal" speed. this is like the standard time.
2. get the time spent to render a single frame. compare this time with the standard time.
3. adjust all movements of objects according to the comparison.

example:

// the game objects move at normal speed when it's 60fpsfloat standard_time = 60.0f/1000.0f;// this will adjust all movements of the objectsfloat fps_factor = 1.0f;void render(){   float start_time = SDL_GetTicks();   // do stuff   objectA.x += objectA.speed * fps_factor;   objectB.x += objectB.speed * fps_factor;   float end_time = SDL_GetTicks();   float difference = start_time - end_time;   fps_factor = difference * standard_time;}  

[edited by - alnite on May 27, 2004 12:34:51 AM]