Re: [Algorithms] Dynamicbounding volume hierarchy creation and maintenance

[Jamie F. <ja...@qu...>]

Mark Wayland wrote:
> I'm curious as to how you exploit visibility when the parent box is
> wholly visible if the child box can overlap (in the conventional sense,
> that is, taking centre AND extent into account)? Right now, if the
> parent box is wholly visible I don't perform any more tests and throw
> the contents of the subtree into the render queue. I'm unsure how you'd
> make this basic assumption when the child boxes aren't wholly contained
> within the parent box? Or, as per usual, have I misunderstood? ;-)

The child boxes are wholly contained by the parent box, so it all works 
the way you'd expect :) The children of a node may overlap each other, 
but both children are wholly contained within the parent node's box.

Jamie

> Cheers,
> Mark
> 
>> -----Original Message-----
>> From: gda...@li... 
>> [mailto:gda...@li...] On 
>> Behalf Of Jamie Fowlston
>> Sent: Thursday, 28 August 2008 2:58 AM
>> To: Game Development Algorithms
>> Subject: Re: [Algorithms] Dynamicbounding volume hierarchy 
>> creation and maintenance
>>
>> Jon Watte wrote:
>>> Jamie Fowlston wrote:
>>>> That kind of fights a bit with the definition of a node's box, 
>>>> particularly as you then just consider the node's box to 
>> be expanded 
>>>> by the upper bound for performing any queries, which seems 
>> to lead to 
>>>> the same definition as overlapping below.
>>> I think "overlapping" is when the boxes for centers may overlap, 
>>> whereas "loose" is when there is a well-defined single node 
>> that maps 
>>> to a specific center/extent tuple.
>> Sounds like a plausible definition, now we just need to get 
>> everyone to agree to the terminology ;p
>>
>> Although by that definition, the tree I described is not 
>> overlapping and only trivially loose:
>>
>> - It isn't overlapping, as the centre of a box always falls 
>> on one side or the other of a plane, so the boxes *for 
>> centres* don't overlap, it's only the boxes containing the 
>> boxes that overlap.
>>
>> - It's only trivially loose because the extent has no effect 
>> on where the box ends up, it just ends up wherever its centre goes.
>>
>> All the best,
>>
>> Jamie
>>
>>> Sincerely,
>>>
>>> jw

-- 

Jamie Fowlston
Program Manager
Tools & Technology
Qube Software

www.qubesoft.com