Freebasic 3d renderer using tinyptc

[Stonemonkey]

thanks, no idea what ive done that would make it go from 2 up to 10 tho

[Optimus]

ABSOLUTELY AMAZING!!!
Never seen such a good 3d engine for freebasic before. Gotta love the commands and stuff, interesting..
27 fps here.

[Shockwave]

If I know Stonemonkey, it will get much much better.

Any chance of getting this to the stage where the running man demo could be replecated SM?
I'd love to see that running at some silly speed in 640X480 mode.

[Stonemonkey]

Sorry, not for a while shockwave. little example of something i did a while ago though with some similar effects going on  here though.

[Shockwave]

Jeepers, that's stunning.
Unbelievably quick frame rate too.

[Rbz]

@Stonemonkey:   That demo3.zip It's fantastic !

I just can't believe that this one was software rendered, welldone !

[Ghost^BHT]

Very cool indeed

[taj]

Wow, please please please tell me how you were doing the transparency in this demo...can I bribe you with Karma???

[Shockwave]

I think that he might have resorted to inline asm for a lot of the transparency stuff, Stonemonkey can verify that though.. Stonemonkey also made this nice little definition that makes quite fast alpha blending possible;

Code Select Expand


#define alpha(s,d) ((((s and &hff00ff)*(s shr 24)+(d and &hff00ff)*(256-(s shr 24)))and &hff00ff00)or(((s and &hff00)*(s shr 24)+(d and &hff00)*(256-(s shr 24)))and &hff0000))shr 8


[rdc]

Heh. I didn't see this before. Looks like the alpha stuff has already been covered. Impressive work on this Stonemonkey.

[Stonemonkey]

Yep, using inline asm but really just using the same method as in the definition Shockwave's posted there where the red and blue channels are dealt with together.

[taj]

But it looked like there was some refraction going on...did I see it wrong?

[Stonemonkey]

Ah right, sorry I forgot about that. It's done by wrapping a texture around the object and then for every frame it first renders to that texture using a camera placed inside the object. Other ways it could be done are with cube or sphere mapping using pre rendered textures although that wouldn't allow for moving objects around so well.

[Shockwave]

Been thinking about multitexturing.
Tell me if you think this would work..

Render a shadow map from the view of the light source to a texture.

Then multi texture the shadow map over the scene and render the scene from the camera's point of view.

I think that this could work quite well for shadows....

[Stonemonkey]

That's a method that's sometimes used but i've not looked into it in much depth as i think, particularly with sw rendering it would suffer quite badly from visible stepping of the shadows. I think it can be done well in hw but i'm not sure how.

[Shockwave]

There's nothing new under the sun eh?  There was me thinking I had had a brainwave!

[Stonemonkey]

heh, that's happened to me a few times too but i think it's pretty cool when it does as it shows you're thinking along the right lines.

http://en.wikipedia.org/wiki/Shadow_mapping

[taj]

Been thinking about multitexturing.
Tell me if you think this would work..

Render a shadow map from the view of the light source to a texture.

Then multi texture the shadow map over the scene and render the scene from the camera's point of view.

I think that this could work quite well for shadows....

Well it always helps to deduc things yourself from first principles. Its a technique used in OpenGL. Here is a cool article on the maths

http://www.paulsprojects.net/tutorials/smt/smt.html

[Shockwave]

Cheers Taj.

The reason the thought of doing it like this came to me wasa that I was thinking about Stonemonkeys SW engine and the way he does the shadows, he explained about the principles of having a prism like volume and everything contained in it's projection being in shadow..

I was thinking about casting shadows onto objects which are cones, spheres etc and just thought why not render the shadow and texture map it onto the object.
Shadows aren't something I can say I've ever really done, I'll most certainly be trying this technique out when I can get some stuff happening in CPP.

[taj]

Ah right, sorry I forgot about that. It's done by wrapping a texture around the object and then for every frame it first renders to that texture using a camera placed inside the object. Other ways it could be done are with cube or sphere mapping using pre rendered textures although that wouldn't allow for moving objects around so well.

Mr Monkey,

Stone, I dont quite get it. Could you be a bit more precise. I'm asking because the code to do refractions in OGL (using cubemaps) requires extensions ansd is therefore expensive in 4k. However you seem to be using 1 texture per object. The technique might be very useful, could you maybe detail it a bit more? I get rendering from inside the object, but how do you decide but how did you decide what the texture co-ordinates are on the object? What field of view do you use for rendering the view? Does it change depeding on how far away the object is or not? Any hints and tips could earn a stone banana...