[sldev] Improving rendering sequence

[Matthew Dowd]

Ah ok.

However, I gather that the current flow (greatly simplified as I realise rendering is actually happening almost continuously):

download prim data
render
download textures of non-occluded prims
render

My impression is that the sculpt textures get downloaded alongside all other textures, on the basis tha tp-ing an area with lots of sculpts initially appears as a lot of floating blobs which eventually (depending on lag and how many textures to download) resolve themselves into their proper sculpt shapes.

Are sculpt textures prioritised over non-sculpt textures in the download queue?

A good example is a sculpt tree - initially these appear as a large floating blog (which you can see underneath), and some smaller disconnected blobs. When the sculpt textures have loaded the trunk touches the ground and links up with the branches. As such the sculpt mesh and culling *after* the sculpt texture has loaded will be different from the mesh and culling before the sculpt texture has loaded - so we may get some additional objects/textures downloaded which are occluded by the sculpted mesh but not the intermediate base prim mesh.

Naively, it would therefore appear to be an optimisation to pad out the difference between a sculpts volume and its base prim with a few extra prims. I say naively since I suspect in practice in the majority of cases any benefits gained from this would be offset by the overhead of the extra prims, particularly if the observer is moving.

Matthew


----------------------------------------
> Date: Mon, 7 Jan 2008 14:42:02 -0800
> From: davep at lindenlab.com
> To: baba at libsecondlife.org
> CC: matthew.dowd at hotmail.co.uk; sldev at lists.secondlife.com
> Subject: Re: [sldev] Improving rendering sequence
> 
> He's right.
> 
> Baba wrote:
>> Quoting John Hurliman here whose email server is on vacation.
>>
>> "the client occlusion culling is ignorant of things like prim types 
>> since it happens at the opengl level working on meshes using 
>> GL_ARB_occlusion_query, while the simulator physics engine uses mostly 
>> rough approximations of objects and the two share nothing at all in 
>> common"
>>
>>
>> Baba
>>
>>
>> On Mon, Jan 07, 2008 12:55, Matthew Dowd wrote:
>>> I would guess that they behave as if they were the unsculpted base 
>>> prim (which is normally a spheroid) for culling - the sculpted 
>>> surface playing no part itself in the culling in the same way it 
>>> doesn't take part in collision detecting.
>>>
>>> Matthew
>>>
>>>
>>> ----------------------------------------
>>>  
>>>> Date: Mon, 7 Jan 2008 12:21:59 -0800
>>>> From: tofu.linden at lindenlab.com
>>>> To: sldev at lists.secondlife.com
>>>> Subject: Re: [sldev] Improving rendering sequence
>>>>
>>>> Argent Stonecutter wrote:
>>>>    
>>>>> Question: how do sculpties interact with culling?
>>>>>       
>>>> The same as any other prim, I believe.
>>>> _______________________________________________
>>>> Click here to unsubscribe or manage your list subscription:
>>>> /index.html
>>>>     
>>>
>>> _________________________________________________________________
>>> Get Hotmail on your mobile, text MSN to 63463!
>>> http://mobile.uk.msn.com/pc/mail.aspx_______________________________________________ 
>>>
>>> Click here to unsubscribe or manage your list subscription:
>>> /index.html
>>>
>>>   
>>
>> _______________________________________________
>> Click here to unsubscribe or manage your list subscription:
>> /index.html
> 

_________________________________________________________________
Get Hotmail on your mobile, text MSN to 63463!
http://mobile.uk.msn.com/pc/mail.aspx