11 replies to this topic

[roel]

According to the guys of Geomerics.

I must say that it looks quite impressive. In contrast with demo of the Fantasy engine, these guys show a way more real-world complicated scene.

[davepermen]

looks pretty.. but, as well, looks like static geometry only.. if so, not really that cool..

if not, very cool. i'm remembering an algorithm from .. hm.. adrian lark? who wrote an engine that essencially did the same as raytracing, just with frustums, and in hardware, wich resulted in automatic global illumination with all it's nice features (no automatic caustics and that, though.. but soft shadows, indirect lighting.. and this not only for completely diffuse surfaces (a.k.a. the radiosity equation style)).

possibly it's even he working there.. have to find out more infos about all of that, hehe.

[Jare]

Seems to be an improvement over typical PRT scenes, better precision and details but still not suitable for dynamic geometry. If runtime costs are reasonable, it seems a nice step forward.

[flipgimble]

I've been following the annoucements out of Geomerics since this years GDC. The several pieces of technology they are getting ready to market are all based on Geometric Algebra. There is some discussion of this on gd-algorithms mailing list. Richard Wareham (sp?), one of the developers, has developed a GA library.

I am very skeptical about the actual utility of their technology to real world game development, until they show up with more information. They do not address criucial concerns like what are the memory costs of their method, how much preprocessing is necessary, how does it handle dynamic objects, or massively dynamic scenes like the destruction you see in next-gen games, and large out door environments.

In my experience big advances in graphics have come from intelligent tradeoffs of what you're going to focus your engine on with advances in hardware capability. I'd love to be surprised but I don't think a different mathematical formulation will be radically revolutionary.

Oh, incidentally if Geomerics wants to prove their rendering technology they should license it immedietly to a proven team of game developers to integrate in a real world game. Just my opinion.

[Goz]

There is an interesting discussion on this over on the GDAlgorithms mailing list if ya wanna know more about it. Not a lot been given away about it (protecting IP and all that) but worth a read fer sure :)

[deiterate]

FROM THE Geomerics WEB SITE:

Quote

At Geomerics we have applied our experience with spherical wavelets to extend the wavelet PRT idea to a method that is genuinely real-time (up to 300Hz). The method incorporates a glossy BRDF and allows objects to move around. It can be extended to allow for moving light sources, and is every bit as compact as spherical harmonic lighting.

Highlighted the important part.... seems that it isn't JUST for static geometry...?

[roel]

The messages on the GDAlgorithms make me believe the opposite.

[Alex]

I had a quick look at some of the papers. If the GA can do what they claim for physics then that might be a realy nice formulation. But eventually it is a different formulation of things already in existance (with the additional bonus of a geometrical interpretation and extendability of geometric concepts into higher dimensions). Maybe I don't know enough about it but I don't see how that helps to solve the global illumination in a fundamentally new way. Using a different compression (waveletts instead of he harmonics) is all nice and might simplify the transformation of the local PRT data but you still need to calculate the PRT. What's the new algorithm that makes this so fast and glorious? Doing old stuff in a new fomrulation won't do the trick I fear...
They should really give some more details...

Alex

[flipgimble]

Here is the specific info from gd-algorithms about how Geomerics supports only static geometry for now:

[INDENT]cons:
- Static (-ish) geometry for the moment although that is mostly down
to our implementation (and some optimisation tricks inside it) rather
than something that can't be fixed in a later release. One can
certainly plonk a small number of dynamic objects in the scene but we
aren't 100% satisfied with our method just yet.[/INDENT]

Btw. calculating the PRT can be done offline and can be arbitrarily complex, meaning you can calculate subsurface scattering, radiosity, PRT glossy reflections etc. The trick is to somehow use the PRT results efficiently at runtime. Some implementations of Spherical Harmonics PRT use 9 to 16 coefficients per surface (stored at a texel or vertex), and they try to cluster similiar surface coefficients. Still the results are very poor (IMHO) and it takes too much memory to be useful beyond demo scenes.

There is not enough information to judge if its possible or not. They are giving the first presentation at the Develop Cnference in the UK (?) (http://www.developconference.com/). Is anyone attending that conference can give us more information?

[emanuele.russo]

I'm a student working on a thesis about real-time global illumination. I'm looking for an efficient tecnique to implement this. Do you think your Spherical Harmonic implementation could be the right method? Does this method manage only static object? in this case, how could I manage moving objects such as clouds, or cars? Should I add a less sophisticated shadowing system for those things, such as shadow maps?

Do you know something (implemented) that could handle both situations? I've found "Reflective shadow maps", or a newer "real-time radiosity", but no implementations are downloadable.

Thanks in advance for your answer,
regards,
Emanuele Russo

[Nicholas Christopher]

To all readers, I do recognize this being an old thread, but with a new post.

@ emanuele,

What is your criteria for this project of yours?

To start are you looking at a CPU / software only solution?
Or a hardware GPU(OpenGL or DirectX) / shader solution?
Real-time or not?

If you are looking for a hardware solution , you can google "real-time HDR" and find at least 3 projects you can download.

If you are looking into a software solution you may think of looking into a piece
of software called vrad.exe you can obtain all the code to this tool.
Because the game company who developed it released the complete source with their game SDK.

A game company that I personally will not endorse.

But whose work may hold some answers for you.

[emanuele.russo]

thank you nicholas, in fact I'm looking for a hardware shader solution... now I'm oriented on spherical harmonics for static plus shadow maps for moving objects... what do you think about?