Hello everyone! I have managed to achieve this quite well....

My problem is, Im not sure how it would scale in terms of speed.

Please, please do me the huge favour of reading this through. My deadline is christmas, and its drawing dangerously near:O



I currently have all tiles and objects(including npcs) in zordered linked list.
I draw the full list, including the models in the right order.

Unfortunately, I cant really cache the 3d models when its their turn to be rendered. It uses way too much memory. So, I must render them each time they are needed.

I have implemented a "dirty rectangles" system, so only the objects(tiles/npcs/etc) that intersect a dirty rectangle will be redrawn.

Here are the steps in order...IGNORING DIRTY RECTS:)

-Iterate through map objects, painting each at the right location...
-If its a tile, then paint a tile...
-If its an avatar(npc/player), then render model to 200x200 framebuffer,
with a MAGENTA background. Create new image with MAGENTA set to TRANSPARENT. Paint onto main backbuffer.


Now, that works. It really does. The problem is it seems like a lot of trouble. Is there any ways I can improve this?

Is faster/possible to make the map into some kind of 3d object itself, dynamically?
Is it possible to set the framebuffer jpct renders onto transparent, to cut out an extra step?

Also, how on earth could I have several different avatars?
I would think a separate world for each model would work perfectly,
but is it too memory intensive?

Thank you in advance for this!


BTW: My FPS is ~80 when im not painting anything new, but drops to
~30 when im in fullscreenmode, and moving my character.
In window mode, im lucky if i get 18, and its only ONE character:(.

Good job:D

Well done man:) I knew that would be possible.


I didnt read everything, so i dont know exactly what you did, but this is what id do...


2 Cameras. One for front, one for rear.

render view from REAR to bufferedimage.


render front view, and blit buffered image into place.

Is is possible to do the following......


load model of naked body...

load model of tshirt, both transformed to 0,0.

render....


and get something that doesnt z fight?
I havent tried it yet, but I am curious.


EDIT:

Is it possible to "envmap" in  jpct? i mean, set objects faces so they are
color of background, and so, invisible effectively. however, they keep their  position, in that objects behind are not seen either.

What the heck are instance objects?

No problem:) Im glad I can be of some help.

A few months back, I did read up on this. I found a document on the quake
engine, which was quite helpful. Ill try to track it down for you:).

I use the grid system, as I said above. I use a simple textual message protocol for things like this.
Without going into too much depth, I use TCP for important data, and UDP for positional data, and everything else.

When a player moves on the map, his client sends something like this to the server...
move <locationid>
The server then checks that there isnt any speed hacks, and that the new location is next to the old one. If its OK, then it broadcasts a similar message(plus player's id) to all other clients on the same map.

I also have "face <direction>", etc.


For you to properly send data in a 3d game, I cant give the definative answer. I havent looked into this more than I needed to, because it wasnt relevant. I would start trying something like this though....

I would not focus on exact positional data. I would, however send it periodically, to ensure the players are in the right place.

I would focus on sending rotational data(as in, what direction are they facing), and speed. So, velocity essentially.

If you were to implement jumping, you would have to make sure the client knew just how high they could jump.

In answer to your question about synchronised animation frames, I really wouldnt try it hytparadisee:).
There isnt really a need to do it, and people will not really notice the difference.
The IDEAL game to look at with this, is Unreal Tournament!
You just tell the client whats happening and it is up to it to cycle through walking animations, or swinging a sword.

I hope this is even slightly helpful:) I am here a lot now,
so I will try to answer all I can:)

If you happen to be working on a game like UT, then you will need to know how to swap guns and items of clothing on your 3d models.
I am trying to get a foolproof method for this too:)

Edit: Yes, it would be smart to send a stop command when the player stops. It will be tricky, as all things are. But im certain you will get the hang of it. Im just glad I dont have to do any of this right now;)

thanks for the info:)

Ill try it out:)

Argh, I wish I was fluent in german, aber mein deutsch ist nicht so gut weil ich bin doof.

Es ist nacht, und ich schlaffe musse(must? no special symbol:P).

My grammar is all wrong for the most part.