We’re wondering if we should keep the Adaptive Kernel (formerly known as Anisotropic) meshing in the first release of Frost. Has anyone been able to get good results with it? Or does it need to stay in beta?
Yes and no. We’re getting good (very good) surface, but not good mesh. So I don’t think it’s the kernel’s fault. I know, I owe you some videos.
We’re also having a hard time (because it’s slow) debugging overstretched (I think) particles. There isn’t any feedback except for the mesh. BTW, Frost doesn’t update when you move a culling object around in a PRT Loader. We were trying to limit the processing of particles and mesh to problematic areas using culling objects. It required the Force Viewport Update button.
But anyway, if I knew I’d be able to get the results back into Frost, I might try duplicating the stretching using Box#3 shapes and see if we can make geometry proxies of the ellipsoids, like in the paper. Darn production deadlines getting in the way of fun testing!
EDIT: Hey wait, what’s this Geometry Meshing? If you (OK, so now maybe we’re talking beta again?) put a checkbox to debug the Adaptive Kernel mode using ellipsoid meshes showing the stretch and position smoothing, that would be a huge help.
BTW, I don’t like “Adaptive Kernel” as a name. Union of Spheres, Metaballs, Zhu-Bridson… I know exactly what those are. But Adaptive Kernel sounds like a “make better” checkbox name. Anisotropic Kernel sounds much more specific.
We’re also having a hard time (because it’s slow) debugging overstretched (I think) particles.
Is it slow at a particular progress % ? That will help us track down the problem. (eg is it slow from 5-15%? After 20%?)
Frost doesn’t update when you move a culling object around in a PRT Loader
Noted, thanks.
Hey wait, what’s this Geometry Meshing? If you (OK, so now maybe we’re talking beta again?) put a checkbox to debug the Adaptive Kernel mode using ellipsoid meshes showing the stretch and position smoothing, that would be a huge help.
I’ll add this to our wish list.
BTW, I don’t like “Adaptive Kernel” as a name. Union of Spheres, Metaballs, Zhu-Bridson… I know exactly what those are. But Adaptive Kernel sounds like a “make better” checkbox name. Anisotropic Kernel sounds much more specific.
Yeah, that’s fair enough… I was worried that “anisotropic” would be a confusing name. Does anyone else have an opinion on this?
I would suggest using whatever the authors of the paper called it. Just like with “Zhu-Bridson” where you immediately know what paper it came from 
Though “Sand as a Fluid” didn’t exactly catch on. “Yu & Turk” sounds cute, though.