Re: [Opensg-users] SceniX

[Victor Haefner]

Thanks a lot! I didn't expect that much :)
--

___
Opensg-users mailing list
Opensg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensg-users

Re: [Opensg-users] SceniX

[Dirk Reiners]

Hi Victor,

Victor Haefner wrote:
 Hi
 
 I am curious about SceniX  (
 http://developer.nvidia.com/object/scenix-home.html )
 
 ..its advantages / disadventages in comparison to OpenSG
 
 I'm thinking SceniX might be closer to the video card, but can it match the
 clustering abilities of OpenSG?

All of my comments are based on looking at the Scenix API a while ago and 
discussion that I've had with the Scenix team lead usually every year at 
SIGGRAPH. I've never practically used it, though.

IMHO the main advantages of Scenix are

- tight CGFX and MetaSL integration
- support for Optix (the nVidia Ray Tracer)
- supported by a commercial company (whether that's a pro or con is a loaded 
question. Ask the Cosmo3D or the Fahrenheit early adopters...)
- written by nVidia people with access to driver developers and source code, so 
they definitely know what not to do

It is a pretty well designed system that focuses on getting the best 
performance 
out of nVidia hard- and software. In that respect it is very much like 
Performer. It also integrates a whole laundry list of effects (see 
http://developer.nvidia.com/object/scenix-details.html), but I'm not sure how 
hard-coded and one-shot they are, or if there is an extensible backend behind 
them.

I'm not aware of any distribution support, and it's multi-threading is just 
based on node locking, which IMHO is just wrong. It can do multi-GPU rendering 
using CompleX, but that's it. So in that respect I don't see it as competition 
to OpenSG. It's also not really designed to be as extensible, so integration of 
new features is going to be a challenge.

But again, I've never used it myself. I'd be interested to hear from people who 
have.

 (I hope I'm not pissing someone off by bringing up another scene
 manager, Just interested in the technicalities. )

As long as you're not mindlessly bashing OpenSG not at all. :)

Yours

Dirk

--

___
Opensg-users mailing list
Opensg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensg-users

Re: [Opensg-users] SceniX

[Carsten Neumann]

Hello Victor,

Victor Haefner wrote:
 I am curious about SceniX 
  ( http://developer.nvidia.com/object/scenix-home.html ) 
 
 ..its advantages / disadventages in comparison to OpenSG
 
 I'm thinking SceniX might be closer to the video card, but can it match 
 the clustering abilities of OpenSG?
 
 (I hope I'm not pissing someone off by bringing up another scene 
 manager, Just interested in the technicalities. )

no, but hopefully you are not expecting unbiased opinions or detailed 
knowledge of it ;)

That being said, I don't know that much about it, just been looking at 
the API docs/headers/examples a bit, so the following is to be taken 
with a cartload of salt:

access to objects is done with locking (you create a guard object 
ReadableObjectT/WriteableObjectT and access data through it). 
They've probably tweaked this a lot already, but going through a lock on 
each (group of) access still sounds expensive
I don't know how well they protect against mistakes here, but it looks a 
bit like the trouble that beginEdit/endEdit in OpenSG 1.x caused when 
you forgot it - their paradigm is pretty close to what the FCEdit macro 
provided in 1.x
Having used 2.0 now for a while I find the whole beginEdit/endEdit dance 
quite tedious, so I would not want to go back to it ;)

they seem to have a set of useful utility traversers, we might want to 
look at those and see if we can implement similar functionality for 
OpenSG. I just saw SeachTraverser that lets you find e.g. all nodes of a 
specific type.

the traverser system seems fairly easy to understand and starting to 
extend it is not very difficult. I'm not sure how easy it is to keep 
different traversers orthogonal, e.g. how one would go about using the 
ShadowMapTraverser (from one of their examples) and make it work with 
say a HDRTraverser. In 1.x we have a similar problem with shadows being 
implemented as a special viewport, 2.0 decouples this more with the 
stage system, although it is not quite at the point were you can really 
mix any arbitrary stages with each other in a meaningful way.

i've not seen anything that suggests there is cluster support.

it's binary only - for linux this means you need to match gcc/libc 
version that was used compile the sources (or hope for the best with 
different versions), which I consider an annoying limitation

Cheers,
Carsten

--

___
Opensg-users mailing list
Opensg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensg-users