Hi Mike,
I tried to understand the content of the VAT textures by myself, especially the specificities because I did not find any documentation on this subject.
I tested the VAT ROP on the version 509 but the next part is for all the versions since the production version 460.
I figured out that for a cube in fluid mode and so with an exploded view of my cube as a mesh, the number of pixels between each frame is not constant.
With an exploded view, I'm not having any optimization by indexed vertices and so I normally have by frame/line:
6 (faces) * 2 (polygons) * 3 (vertices) = 36 pixels per frame.
I forced my “Target prim count” to 12 and not 6 as the “Calculate Optimal Prim Count” feature gave me (which is probably a bug in a cube case). In fact, I dit it to have for an UE4, a visual cube and not a retopologized one with half its polygons in the fluid case.
What I'm not understanding here is that for a cube with a transform that move the yaw 30 degrees by 30 degrees on 3 frames and a texture size setted up to 1024 pixels, I have a texture with:
1st row: 36*4 = 144 texels used
2nd row: 36*3 = 108 texels used
3rd row: 36*2 = 72 texels used
I also saw the GDC 2017 conference you've done for your implementation of the VAT and the attached ressources you provided contained “constant” frames so there is obviously something that has been changed in the VAT Textures since this GDC.
I guess it's probably something like an additional interpolation information but how to disable it if its somehow possible and can you explain what it realy is?
Thanks