RE: [Algorithms] R-Trees vs. Loose Octrees

[Robert D. <RD...@ac...>]

> I was probably not very clear on what the hgrid is used/good for.
> The primary use of a hgrid is for broad-phase culling in a 
> collision system.
> I don't immediately see how it would be useful for frustum 
> culling, but I haven't really thought about it.

Having just re-read your description I've suddenly realised that we used
this on a game I worked on ages ago, specifically for frustum culling.
Because of the hierarchical nature you can do culling of a large number of
cells at a higher level thus avoiding more detailed checks lower down.  On a
system where render speeds and poly counts are a significant problem,
anything which reduces the time spent rejecting stuff is a big gain.  The
other nice thing about this approach is that even having decided that a
large object in a large cell requires that cell to be rendered, you can
still reject small objects in the smaller cells it includes if they should
be rejected, which in this particular case was a major benefit.

Rob