Hi Gustavo,

> I'm wondering if you have any enhancements/optimization
> regarding meshing in V3.

No nothing so far.


> Mostly crossing my fingers for a
> better continuation poly-flow across patches.

Do you have a simple example that you could give of a situation that you'd like to have improved?

Getting things to match across patches is a pretty tough problem to deal with - basically the way the mesher works is that surfaces are subdivided in UV space into smaller pieces. If you have 2 surfaces that have totally different UV parameterization between them then that will cause them to have somewhat different subdivisions between them.

Then there is also a pretty big problem that surfaces may be trimmed in numerous different directions not aligned with uv space as well - like for example imagine a plane that is trimmed by a 5 point star - there isn't any arrangement of uv-aligned quads that lines up with those edges since the trim curves are going in various different directions from the natural uv space directions.

So it's a pretty big problem to solve. Working in UV space tends to be the most natural way from an algorithmic standpoint to work with surfaces.

- Michael

Hi Michael,

I understand the general constraints, but I do have many cases where I model clean patches (untrimmed) that have matching UV flow, although varied UV count. So maybe there's some tricks regarding your UV spacing of your tessellation that can get us there? I'll send you some examples offline. The use-case here is to rough our something in Nurbs before taking in to SubDs for further refinement.

G

Hi gustavo - if you have untrimmed surfaces that share an edge, you could try using Rhino's MergeSrf command to merge them together into just one single surface. That would get you a unifed mesh.

But if they remain separate, it's not too likely that they actually have the same parameterizations between them, each surface probably has different spacing for its knot vectors. In Rhino you can see this if their isocurve wires do not line up with each other.


> The use-case here is to rough our something in Nurbs
> before taking in to SubDs for further refinement.

The MoI mesher is not really designed at all for that use case - that use case requires special constraints on topology and using all quads.

The MoI mesher is more oriented towards making a low polygon count with n-gons.

Those are really different kinds of things - it's not really feasible to just tweak the current MoI mesher to get that totally different kind of sub-d friendly topology result, it really requires a totally different mesh algorithm, probably one that works more in 3D space (like primarily hanging quad pieces from 3D edge curves) rather than one that operates in UV space.

It could be possible that I can make some improvements in a few situations, but just in general the kind of result that you want needs a different kind of meshing mechanism.

- Michael