[Patch] Specular Highlight

[Ian]

maybe its something like this ..

[Bart]

I'm not up to speed on what you have or haven't tried but could the winding parameter influence how the quads are defined? Has this been tested?

[Ian]

The real3d doesn't use the winding at all
I think what's happening is the order of the vertices for the quads is wrong. With triangles it makes no difference either way. With quads, assuming they are planar .. also makes no difference. But when they are not planar, or do other crazy stuff things can fall apart.

What I *think* might be happening is this ..
Say a quad is sent for rendering. So we have 4 vertices
v[1] v[2] v[3] v[4]
The next quad shares vertices 2 and 3 from the last buffer.
So we get
v[ ] v[ ] v[3] v[4]
Then we add the next 2 vertices so we again get
v[1] v[2] v[3] v[4]


But what we are doing is this
Quad 1
v[1] v[2] v[3] v[4]
Next quad shares 2 vertices, but we move them to the start so
v[3] v[4] v[ ] v[ ]
Then the next 2 vertices get added at the end
v[3] v[4] v[1] v[2]

The tl:dr version.
We move the shared vertices into the beginning slots. I am guessing the hardware doesn't bother unnecessarily copying them. This can lead to the diagonal split changing between each squared quad.

I am not 100% sure this is what is happening. Gonna give it a go when I have some time.

[Ian]

We need to add this

From a few minutes of debugging this evening .. I'd say that sun angle or the specular flips if the Z value of the sun becomes positive.

[HarryTuttle]

From a few minutes of debugging this evening .. I'd say that sun angle or the specular flips if the Z value of the sun becomes positive.

Do you think it could fix lightning in L.A. Machineguns?

[Ian]

Hopefully:) Just need to make some test cases. It didn't help before the coordinates were all messed up because we were doing part of the coordinate conversion in the shader, and part of it outside the shader. Which made debugging almost impossible.

My plan is to draw some big fat arrows indicating the sun direction then see where the specular highlights show up in real footage and compare. Some good specular footage here -> https://youtu.be/UL_SvP3WDv4?t=3m10s

I did find this error in the lighting of La Machine guns .. maybe deserves it's own thread.



Ambient is set to 1, so there is no shading at all here. But on the real hardware, even though the footage is poor quality you can clearly see shading. Wonder if the ambient and diffuse values for lighting are divided by the model scale, instead of normalisation ....

Say if the model matrix scaled the model 2.0 times. Instead of normalising every vertex, you could instead just divide sun ambient & diffuse by 2? Have to try that at some point

[HarryTuttle]

Beware that most of the lightning in that game is fixed shaded, sometime in combination with TranslatorMap bit set. In my build, for example, the lightning in that scene is more similar to the arcade version.

(Cannot post any screenshot right now 'cause I'm out of town again until 5/6 july).

[Ian]

I set the fixed shaded value to 0
So you can see what's lit normally (in this case the people)



Shading is totally absent in supermodel

[Bart]

I've noticed that as well. Are either specular shading or the translator mapped used on the NPCs?

[Bart]

Say if the model matrix scaled the model 2.0 times. Instead of normalising every vertex, you could instead just divide sun ambient & diffuse by 2? Have to try that at some point

That's an interesting idea...have you had a chance to try?