[gameprogrammer] Re: 2d side scroller

I would leave threads out of it for now.
Get a timer going, render as fast as you can to keep framerate up, and update the scene (collision detection / response, movement, physics) every ~30ms
input is event driven so that can be processed straight away to change to state of your game world (ie fire key spawns a new bullet object with a certain position/vector/sprite), then return to rendering/updating.
while (1)
has30ms passed since i was last here?
update world(elapsed Time) - updates the world, player, physics, enemy, move platforms, check for trigged events etc
reset timer
renderEverything()
end while

You can further optimise this game loop by allocating less time for AI updates or similar.. eg only update AI strategy twice a second..

As for optimising there are two things (if you arent already doing these)

Optimise the gfx pipeline... dont draw everything and count on opengl to do the culling for you. Instead add a bounding square/circle around everything, and if it is outside of your screen view, then dont render it. this might mean breaking your backgrounds down into half screen sized chunks.. each one is an object you need to check if it is inside/intersecting your view before sending it to opengl
its a simple test before rendering,
for each object
if objects bounding sphere/square is inside or intersecting my screen
object.draw()
else
nothing
end for
- Use display lists for the static geometry so it is stored on the gfx card.

Optimise the collision detection - using the same bounding square/circle
example for bullet
if the bullet isnt inside the bounding circle of the player or any objects, ignore it this update
if the bullet is inside the bounding circle, then do per pixel collision test against the bullet.
example for enemy
if the enemy is outside of my view, dont update his positon, dont process his AI, etc (unless he has seen the player and is chasing or hunting the level for you)

this requires that you have a bounding square/circle for your platforms floor/boxes/roof etc and you will need to check for collisions with them also (these can be automatically generated from the models as they are loaded)
If you have diagonal passages, this might be more difficult, as you might have to do point in polygon test.. not sure how that works.

I dont know what language you are using, but if its OO you should be able to make this general behaviour that can apply to players/enemies/projectiles in the same way.

-Josh

Nick Howes wrote:

Yes it's best to do rendering and logic in two separate loops. The looping itself isn't intensive so you won't get any slowdown from that (until you have massive numbers of entities). The extra advantage of this is you don't necessarily have to render each time you update the world. They could even be separate threads running at their own speed (provided everything is thread-safe). You could control the frame rate without affecting the actual running speed of the game world.

Alan Wolfe wrote:

Thats a good point. I was trying to be more efficient by taking advantage of loops as they happened, ie instead of looping through all enemies and doing movement/physics/logic and then looping through them all again later to do the drawing, i tried to make it all part of the same loop.

Guess that's not the best idea eh? :P

Thanks for pointing that out, that probably explains some wierdness I was having with the moving platforms.


On 11/7/05, *Ken Johnson* <johnsk16@xxxxxxxxx <mailto:johnsk16@xxxxxxxxx>> wrote:

    One thing that I noticed is that I would render the screen as the
    last operation. You render enemies/player before you test if an
    enemy may have been killed by a players bullet and you also render
    the player before you test if he needs to be moved from a moving
    platform.


    On 11/7/05, *Alan Wolfe* < alan.wolfe@xxxxxxxxx
    <mailto:alan.wolfe@xxxxxxxxx> > wrote:

        Hey everybody!

        I'm working on a 2d side scroller in SDL/OpenGL and this is
        the first actual one i have ever made.  Because of that, I
        have found some things I did which were not the most efficient
        way to do things and have changed them, but I have found other
        things which i know must not be the best way, but am unsure of
        what the best way would be.

        Here is my order of operations in my game loop:

        1) Check which keys are down and move the player accordingly
        or spawn bullets etc. (movement checks that there are no
        blocks in the way, as well as checks if they touch spikes
        etc).  This section also figures out which frame of animation
        to draw for the main character whether he looks like he is
        jumping running shooting etc.

2) Handle gravity and other physics for player, moving as necesary

        3) Handle the scrolling of the camera to keep player on screen
        (the game uses mouse look so the screen really scrolls based
        on mouse coordinates.

        4) Draw paralax layers that are behind the map

        5) Draw the map

        6) Draw the main character

        7) Draw and handle enemies - draws enemies as well as handles
        enemy physics and logic (AI)

8) Draw projectiles - both enemy and player projectiles. Draws them and does physics and logic, also checks to see if
friendly bullets have hit enemies (if so take away life from
the enemy it hit) and checks to see if enemy bullets have hit
the player (if so take away life from the player).

        9) Draw game objects such as powerups etc.  Does physics and
        logic as well for each.

        10) Draw moving platforms (along with path logic as well -
        such as if it moves in a circle, it will move it the next
        step). This part also checks to see if each moving platform is
        going to hit the player when it moves.  If so, it pushes the
        player out of the way (checking to make sure it doesnt push
        the player through a wall or anything like that).

        11) Draw paralax layers that are in front of the map

        12) Draw game GUI such as lives left etc

        13) see if the player has touched any bad guys.  If so, take
        away life and mark the player as being in the "injured"
        animation state.

        14) see if the player has touched any power ups.  If so, give
        them energy, extra life, or whatever.

        Does anyone see anything sticking out which might be optomized
        or done in a better way?



        A secondary question is how is the best way to program moving
        platforms?  I am in the middle of implementing them and right
        now my tactic is this...

        -----
        Loop through each moving platform...

        If the player is on the current platform, move them the same
        amount as the platform is moving (this makes sure if they are
        standing on a platform that it carries the player with it) -
        (taking into account other walls etc so you don't push them
        into space etc)

        Move the platform.

        If the player wasnt on a platform but is now, move the player
        the same amount as the platform is moving (taking into account
        other walls etc so you don't push them into space etc)
        -----

        I know this isnt the best way because this makes the player
        "float" slightly above the platform at times and rarely, the
        player will get stuck inside the platform.

        Is there a better way to do this?

        Thanks!
        Alan





---------------------
To unsubscribe go to http://gameprogrammer.com/mailinglist.html


Other related posts: