interactivecomputergraphics / treensearch Goto Github PK

How does performance compare to a kd-tree?

In particular, I am wondering if this data structure could be used for example for photon mapping.

Greetings.

I want to to bring an issue with your code to your attention. Specifically it is the issue that various variables are shadowing class members or even local variables shadowing other local variables. (For example here)
While this is technically not an error according to C++ standard, this is considered bad practice and compilation fails with '-Wshadow -Werror' on gcc or when using MSVC with /WX (or any warning is error flag enabled) and the conformance mode (/permissive-) enabled.

Cheers

• Your paper is wonderful， especially the brachless part, really get so much out of it

Background

• Recently, I implemented a rigidbody simulator with PBD, which learned from papers written by interactive graphics and nvidia

• But just as you metioned in your paper, uniform grid usually work with particle based method
• If I want put two rigid body into the simulator, the difference of size for them can‘t be too large
• Because the particle number is in a cubic relationship with object size，if we keep the same particle radius for all, memory will explode
• So maybe a hierarchy data structure + mixed radius size will help and that‘s exactly your paper do in sph
• While in your paper you also mentioned TreeNSearch works not very well in the situation that objects move rapidlly

The Question

• Is this data structure suitable for Position based dynamics , including rigidbody, cloth and fluid simulation