Question on Whether RXMesh Is Better Suited for Intrinsic Topology Edits Than Hole Closing

[will-zzy]

Hello, sorry to bother you again.

I would like to ask whether RXMesh is better suited for intrinsic topological modifications, such as split and flip, while being less suitable for extrinsic modifications like hole filling. More specifically, if I want to implement closed holes, would a half-edge data structure be a more appropriate choice?

[Ahdhn]

It really depends on the algorithm you want to implement. I think RXMesh offers everything that a half-edge data structure would offer. The main question is: how much parallelism does your algorithm expose?

If you are building/processing a long chain of sequential steps, then there is little parallelism, and RXMesh would perform poorly. For example, if you are building a directed loop over the outer boundary of an open mesh, shown below as E0→E1→E2…, then there is little parallel work because we have to know E0 before knowing E1. However, if your algorithm is something like “do x work for all boundary edges at the same time,” then there is parallel work because all boundary edges can be processed simultaneously. In that case, RXMesh would be a good choice for your algorithm.

Image taken from here

[Ahdhn]

Another thing (that we also mentioned in our paper (under non-goal in page 3) is we do not support "non-topologically local areas of impact for update operations" like in mesh zippering. This is because the cavity here spans more than one patch and these patches are not neighbors.