Exactly, when drawing a solid wall you dont have to transform each pixel into texture space ... you are just drawing horizontal lines!

Either way it looks great ... but i was wondering how the texture is mapped. It looks like an 8x8 texture but it appears to have taken a triangle slice of the texture? Or is the whole 8x8 texture not mapped to a single triangle (since it should look distorted when you map a square to a triangle).

Either way it looks great ... but i was wondering how the texture is mapped. It looks like an 8x8 texture but it appears to have taken a triangle slice of the texture? Or is the whole 8x8 texture not mapped to a single triangle (since it should look distorted when you map a square to a triangle).

It map a texture according with the U,V coordinate in texture space given at the 3 points who made the triangle

Matref, your backface culling is very interesting. Since I rewamp the polygons sytem, I was forced to remove z-sorting, and I start to study this method in fact  . I will be happy if you can explain me this.

Anyway, some new screeenshots (and I work now hard on the lib to release it as fast as I can !)

That looks AMAZING!!!

Keep up the good work, as always!

thanks 

Matref, I finally don't need your backface culling, because I manage to create something user-friendly (well I hope so  ) and which can manage backface culling, as well as a LOT of other things 
But, I really want to see how you do it 

Anyway, I am close to the update, but there is some very bad bugs I want to correct before the release, so maybe tomorrow. I don't think I will have the time to correct everything this evening, though 

The textured face looks really good, and it runs at a decent speed too!

A litlle screenshot to show the backface culing thing. Note that a lot of lines can be remove, I currently work on this. 
Vertex isn't process if face is remove.

It's currently run at +-31 fps, but I hope gain more speed.

I simply perform the normal vector of each faces using a cross product. With that, I have the plan equation of the face, and I check the camera pos according to this equation. The result is positive or negative, and I hide face according to data definition. The only thing that we have to worry is  when all the points of the face is not on the same plan (sqare only), but one workaround is to increase geometry complexity (or simply be careful)

Yeah he is doing it a more traditional way (and thus more accurate). The method we used was the 'quick and easy' way, with the tradeoff being accuracy (though it is minimized if you sort faces, though with proper culling you may not need to sort if you are using convex objects).

This tuto is weird. I use this method, but I only do the cross product. Taking the tuto code, I remove the

         double len=sqrt(A*A+B*B+C*C);
         A=A/len; B=B/len; C=C/len;

line, since by definition, the cross product give one of the possible normal vector of the face. No need to normalize the vector. This vector give automatically the plane equation of the face, ax+by+cz-d=0. I then replace the x,y,z value by the x,y,z camera position , and check the sign of the result.