Printable View
@ DancingOctopus:
wouldn´t it be more powerfull if you´d bake the light & shadow maps into the tiles (bitmap data)? Like copy Pixels or merge them with a differnt blending mode (soft / hard light). That way you´d keep your performance without putting anohter layer above the original tiles
yep, you're right, that would be much faster. The trade-off being you lose the shading on non-tile objects, which adds a lot of depth to the effect. What I might do is make that the medium-level option, and have objects check the average surrounding light levels to set their own overall brightness. The problem with that is it would throw a lot of extra code to pull off, whereas switching between the current methods takes a couple of conditionals..
This is all assuming I want to keep the lighting in the game. At the moment, it's just a bit of fun between working on the level editor and waiting for graphics from my artist :)
been toying finally around with AS3 the last few days and I feel satisfied and frustrated. I´m getting perfectly all the xml, array data stuff but placing sprites or graphics to the stage is atm. a pain (for some reason I can´t use the addChild() function within classes that are imported into the main class).
This is very early and my first AS3 movie, I am missing still alot of things like a text field and a nice interface but since I am conding atm. only (no flash 9 alpha used at all)- but the performance is awesome.
Adobe itself claims to improved the performance up to 10x faster- in my case for the wireframe display its 8.6x times faster as AS2 with many tricks.
here is a demo of a teapot the model has 516 faces (triangles) evem if you see atm. only the quadfaces all triangles are rendered.
http://mitglied.lycos.de/mmvr2/flash...eapot/demo.gif
http://mitglied.lycos.de/mmvr2/flash...t/ase_load.swf
have some patience as my fast computer usually needs 3-4 seconds to parse the model (the very old one 9 seconds) and you dont see any text or loading bar atm.
Wow. That is just too cool for school.
How many models can you have on screen before you get slow-down?
How about texture mapping? Is that up yet?
nice one, speed is excellent, don't really understand your addChild problem, your using it like this right:
class2 extends ADisplayObject{
}
class1 extends ADisplayObjectContainer{
var classRef2 = new class2()
addChild(classRef2)
}
thx for the coments
@mr_malee:
hey thx,- I´ll check it out- have read only 1/2 chapter about the displayObjectContainer stuff. I was just extremely pissed off as I couldn´t add anything to the diplay container within my sub- classes and that threw away my concept.
about the demo:
I think I can still tweak alot, also back with AS2 texture rendering was faster as wireframe rendering because for wireframe rendering I have to use 4 drawing API commands (moveTo, 3x lineTo) for each frame where for the texture rendering I have just 1 command.
But the pain is to re- code the crazy triangle texture function - I already noticed that I have to be carefull now with AS3 because the compiler just notices everything I forget + I can´t just experiment easy and fast in Flash with ugly code before I write the real thing.
upated the demo,- now has a textfield and FPS counter (yes learned that just in a few minutes :D)
http://mitglied.lycos.de/mmvr2/flash...t/ase_load.swf
some experiment from a few days ago:
http://mitglied.lycos.de/mmvr2/flash...2_clip_ani.gif
its a clipping method wich leaves out the polys wich dont fit within the camera bounding box in AS2.
In most cases it didn´t help that much since my models were simply to smal,- but with larger objects with more polys (like a level, or ground) it worked great. I´ll implent this one within the next days in my AS3 version.
the cool thing is actually that I support multi- objects just as you would place several objects in max without merging them to 1 mesh.Quote:
How many models can you have on screen before you get slow-down?
Regarding performance I guess that entirely depends on your computer- but what bothers me is the parsing time of the ASE file so I already planned to eventually create my own file format that is optimized for my parsing.
another demo:
http://mitglied.lycos.de/mmvr2/flash...eview_shot.gif
http://mitglied.lycos.de/mmvr2/flash...2/ase_load.swf
just added a simple face fill (4 colors) but still no z- ordering. But I am myself already very pleased with the performance now- I mean look at all the fences that´s some crazy amount of triangles there.
Wow.how about an organic mesh, like a human model?
I'd love to make a low-poly Biohazard 4.
this might be a silly question but it kinda extends what evilKris said, can flash handle nurbs and has anyone done anything with it?
whats the difference?
i've heard that MAX is good with poly's and MAYA is good with Nurbs.
anyway, loving the 3D stuff render, i seriously think that these projects are gonna change development in flash. Maybe push the flash dev team to include native 3D. Just curious, how big is the ASE file for that last upload (the fence) ?
fence ASE file 96 KB
but what takes the time atm. are the vertex coordinates wich have in this model 9 numbers after the dot "." e.g:I plan a more elegant way reading them in custom format where I store them without the dot "." and divide them once I parsed them with number I stored so that in the end I recieve my excact number. Or I reduce the decimal as I probably dont need it if I dont rescale my models.Code:x 1.001233452
But about all the 3d stuff- maybe I´ll start once a thread that will cover most- even the displaying 3d stuff and what programs, what formats,...
about nurbs:
think of nurbs (non rational b- splines) as 3d- vector curves (but they are called splines in this terminology) that define shapes & curves without any loss in detail as for geometric shapes. Imo. Rhino is the best NURBS modeler it´s precission and workflow are awesome.
this link should give you a good view of what a nurbs model looks like:
http://www.shonner.com/drafts/ap751_mouse.htm
so compared to polys or mesh objects it´s a more elegant and preciese way of describing a shape- esspecially industry designers love it. But it has some cons and those are that it´s almost impossible creating organic objects with it (character, nature,...) also up till now there is no realtime renderer (of what I know) that is capeable of rendering nurbs models in a unique way. Right now all renderers first convert the model into a triagnle mesh and then render that one.
@ EvilKris:
that could be very likely- a character up to 300-400 poly´s and 2 or 3 enemies on the screen at once would work I guess. With the prerendered background and masking techniques a project that would work well with Flash.
So nurbs is totally out of the question for flash, too hard to calculate? and is triangles the best way to describe an object with minimal information?
are triangles used because there is less points to consider when mapping texture's?
sorry for these questions, Im becoming intrigued
right now trianglisation is the only way deforming textures or images in any direction. Older methods like slicing techniques (raycasting) or per pixel precision (voxel) arent used today anymore because the consumer GPU´s with triangle technologies are taking overhand.
3dsmax supports Nurbs- UV maps (those are the texture coordinates ) but hardly anyone uses them as with nurbs you need additional control points (just like the pen tool in flash [ p ]) to describe the curve. In flash it´s just one point to define for that but more common are 2 points or the spline interpolation where curves are described with just points.
Anyway as for now it seems to be the easiest and most comfortable way doing it over triangles as most 3d- packes already have great tools for that. About UV- maps take a look at this link:
http://www.tamasoft.co.jp/pepakura-en/
that´s how you could discribe uv- maps as foldable paper models. So each 3d- model gets a fold map to know where the 2-d image is visible on the mesh. In the end the 2-d image is divided into many smal triangles and placed on it´s referencial 3d- space points. remember this:
http://board.flashkit.com/board/
this was my first attempt slicing triangles out of images- and it works great because once I have a default triangle of it I can skew & transform it to any triangle shape I want- thus I can now map it in 3d-space.
and yes nurbs are Imo. out of the question. You could (there have already been a few) however experiment with nurbs splines as wirefrime or sold shapes. But texturing is right now with flash out of the question.
Besides it´s more fun creating poly´s instead of nurbs- models because there are more poly tools out there. Not just max, maya, xsi ,LW, modo, silo,.. but also newer coming sculpting tools such as zBrush or the awesome mudbox are all basicly working with poly models. And even for the texturing there are great tools as well-, max already has a ver impressive UV- editor (imo. one of the best) but other tools such as deep-paint-3d or bodypaint3d can help alot too.
I think that if it continues like this and the 3d- stuff isn´t really that difficult. Its just that sandy and papervision are showing great examples, but showing how they did it and that makes people think it´s a whole new dimension. But in the end it´s just plain placing sprites transform them a little bit for each face. For me that´s just the beginning- I plan to create custom bone- tools that let me rig meshes or characters and with that interactivly animate them in flash.
I think the whole scrolling and 3d- sprite techniques are just as difficult as these 3d- things.
wow, great post, shows how little i know. I'm amazed at the current level of flash projects I'm seeing these days. Don't really have anything else to say than, the next few months are gonna be killer :D
here another clipping video from as3 (not yet worth uploading another demo):
http://mitglied.lycos.de/mmvr2/flash...view_shot2.gif
I need to find a way though to clip out the backfaces and hidden faces
finally a AS3 demo with textures :D
http://mitglied.lycos.de/mmvr2/flash...issan_view.jpg
http://mitglied.lycos.de/mmvr2/flash...3/ase_load.swf
I tried to create a extremly low poly nissan car, painted the texture myself with some reference photos merged:
http://mitglied.lycos.de/mmvr2/flash...nissan_uvw.jpg
this is only for a demo right now as I want to write a propper object manager so that I can add/load any mesh data combined with any texture data. Wich lets me for example transform meshes by object id´s like turning the wheels, letting them steer ect.
The demo is far from beeing optimal,- also the texture mapping is not quite a beauty but i´l fix that soon
excellent stuff render.
I do have one question though. I know you said it's unoptimised so maybe this could be put down to that, but I noticed when I zoom right in so only a handful of polys are visible that the CPU load stays the same (about 25-30% on my machine). I would have thought that as you draw less the load would reduce?
Can't wait to see a demo with z sorting/backface culling