Your advice : PPL vs Dynamic Lightmapping ?

First, check the 2 news demos on our webiste :
http://ibelgique.ifrance.com/Slug-Production/

Next, do you think that it’s possible to use Dynamic Lightmapping instead of Per Pixel Lightning or it’s more powerfull to use LightMapping for static Light and Per Pixel Lightning for Dynamic Light.

Here’re 2 screenshot :

Per Pixel Lightning :

Dynamic LightMapping :

per pixel lighting (shouldnt it be texel?) lighting is far better + once u add it dynamic lightmapping eg quake2/3 will look pretty bad afterwards. the problem of course is not all cards (eg tnt2) support per pixel lighting

Now you can download the demos; I’ve updated the site.

Light maps work and look good when combined with per pixel lighting (it’s per pixel; vectors vary per pixel and bump texture vector triplets are filtered and change direction per pixel).

For correctness the light map could be, generated from occulters as a shadow texture (not a shadow map that implied depth info). The question arises should you bake to tangent space or project. For moving objects interracting with light correctly in the scene you want to move away from usual light map tangent space representation and move to a projected texture, this really becomes a shadow texture representation IMHO. Artists could apply light modulating textures for an artistic effect or a classic tangent space light map from a global illumination indirect calculation could apply a base level ambient term in your first pass but that may look flat without doing something clever with your bump map to modulate it, I have some ideas on that if you’re interested.

The scene is so trivial, it is pretty difficult to say that ‘there is no difference’

With a single quad, I would think you could duplicate almost any effect with a single texture map.

It really depends on what you are doing. You really need to light a scene that is close to what you want to have in your finished product before you can decide.

Now, instead of a quad, we’ve implemented the Dynamic LightMapping into our Slug 3D Engine, but we don’t have a lost of performance and it’s not so bad
And you ?

Check : http://ibelgique.ifrance.com/Slug-Production/

Now, the Radeon user can try the demos with the same effect.

It’s seems that for downloading, we (me too) need a password.

I’ve sended a mail to the technical service and I hope that it will be fixed soon.

But you can try …

Me? Considering the time frame for me to be able to complete anything (my full time job is not related to graphics programming), it would be foolish for me to consider anything except a full programmable pipeline like what will follow the NV30 and 9700. In other words, full OpenGL 2.0, with little to no consideration for backwards compatability.

But, that is -my- situation. Right now my considerations are purely people looking at my resume 8 to 16 months from now after I have had enough time to complete a graphics portfolio :slight_smile:

You’re right, so for you, if you don’t have the god 3D card, stay behind.

I am just trying to keep from saying “Why are you even asking, of course you want to use per-pixel-lighting”, because then you will say, “some people I am programming for have old cards!”

The fact that my audience is game developers who are often ahead of the bleeding edge (they sometimes get prerelease hardware) sort of biases my opinion :slight_smile:

More on topic, the reason per pixel lighting is not much slower must be because basic light mapping does not stress the hardware you are using at all. So, more complicated effects, like per pixel lighting, could be added for free.