How does CDT compare to Triangle #29
Replies: 1 comment 3 replies
-
TL;DR: If you only need to do constrained Delaunay triangulations CDT is a great replacement for Triangle. Give it a go: I would really love to hear how it worked and see some results! LicensingAs you mentioned, Triangle lacks a comprehensive/commonly used/permissive license.
I tried to reach Prof. Shewchuck regarding licensing Triangle, but my attempts were left without answer. Even Shewchuck's code for robust predicates being "in public domain" is a problem for licensing: public domain is not a software license and arguably is not open source. That's why CDT uses a different BSD-3-Clause licensed implementation. Triangle needs patching for x64Original Triangle code does not work properly on x64 and some patching is needed to make it work. For example libigl is using patched Triangle. FunctionalityCDT can definitely be used to replace Triangle for computing constrained Delaunay triangulations (I've done it). However Triangle is a swiss-army knife for triangulations: it can also do conforming Delaunay triangulations and re-meshing via Steiner-point insertion (Ruppert's algorithm) to produce quality meshes for applications like finite-element analysis. PerformanceLastly Triangle positions itself as being very fast. CDT was definitely designed with performance in mind and we optimized things where we saw potential for it. For example boost's r-tree and flat_set are used to speed things up. But we never benchmarked CDT against Triangle. So Triangle may potentially be faster than CDT. Benchmarking is on my TODO list, but any help or input is appreciated. Maybe when you have libigl adapter for CDT working you could run some numbers? :) |
Beta Was this translation helpful? Give feedback.
-
Original post by @RotateMe:
Beta Was this translation helpful? Give feedback.
All reactions