Horde3D

They're core now. Are there any plans to update Horde3D soon or whats going to happen?

I cannot speak for any horde3d dev, and myself am not good enough to do anything in this scale without help, but I would love to have geometry shaders in the engine, as well as a dx10 backend.

Geometry shaders probably would be done via a opengl 2.x extension and not via opengl 3.1 though. Also we might need some kind of simple fallback mechanism for materials to support such possibly not available functionality in shaders without manually needing to select between different rendering paths. Not necessary though.

Hmm, what would you need geometry shaders for?

My experience is that although they can be handy in certain situations, their use is rather limited, especially since performance is not that good. And as phoenix indicates, you need a fallback implementation anyway if you don't want to limit yourself to d3d10 hardware, so why bothering with geometry shaders at all?


I was under the impression you were a "hardcore" Linux guy

DX10...
ew...

Hi, I was always shouting "We need geometry shaders!" in the forum and a fan of DirectX 1x, but changed my mind after watching their demos. There is a small demo bundled with the latest DirectX SDK that includes a sample of geometry shaders which IMO is a lot boring than watching drying paint on the wall

A few notes about DirectX 10: Well, there is a very small improvement in visual quality [soft particles, better lighting, ...] and also makes the life easier for the developer [in comparison with DirectX 9] in cost of loosing about 50% of Windows gamers.

And about DirectX 11: Well, there is a great improvements in the visual quality [tessellation] that makes a good gaming experience in cost of halving the frame rate. Currently about 3 ATI graphic cards [5000 series] are 100% supporting DirectX 11 and tessellation [DirectCompute and hardware tessellation features are supported by older DirectX 10 compatible cards too] and you should run Windows 7 which means you will lost 80% of gamers.

Take a look at Steam Hardware Survey for detailed numbers and charts about gamers and their specs. Also have a look at the Heaven Benchmark to decide which backend [DX10/11 or OGL] you are going to use.

Currently the games that they are using DirectX 10.1 aren't more than fingers and Dirt2 only supports DirectX 11 [not in depth]. IMHO Horde3D + OpenGL 2 is the best/only option for now

Best regards, Siavash.

I've seen OpenGL drivers perform much worse than the DX drivers on the same system for the same scenes before, and there are quite some benchmarks out there telling the same. It's the best to at least provide both options for such cases. I don't talk of DX 9 here though because that really should be worse than OpenGL 2.x plus extensions, but DX 10 would be something nice to have.

Oh, and nearly everyone I know already runs either vista or 7. At least all gamers I know. I am the only one left with XP because I don't care about Windows for myself. I won't be the one writing the DX backend -.-


Just to clarify some bs here, DX 11 works on Vista just fine. :p

I had done a little in this regard, but pretty much just outright stopped. When I thought about for a second, I really couldn't think of anything useful to do with them. I couldn't figure out how to generalize it either. All of the really cool stuff is pretty implementation specific.

The only thing I actually found interesting was reading back geometry after it had been processed.

I'm not sure what I'd use geometry shaders for *right now*, especially seeing that their performance isn't very good on current hardware. But in the long-term they'd be on my Horde wish-list too, as I'm sure they will become important sooner or later.
The most popular example is using them to move stencil-shadow calculations off the CPU to the GPU (contrary to popular belief, stencil shadows are still "better" than shadow mapping in some cases). But seeing that Horde doesn't support stencil shadows (or application-side use of the stencil buffer), this obviously isn't needed.
As far as I can tell, most of this "visual quality" stuff is just Mircrosoft's propaganda... The Wii (which uses ~DX7/DX8 level hardware) can even do soft particles (although, yes, it's much easier for the developer to do it with DX10 )!

First post of 2010

The same issues keep coming up.

Regarding GS - There really isn't enough support or applicability to make this a priority. If anyone really wants to mess with GS, it's most likely that you will be doing small proof of concepts and writing all the code by hand to fit your specific project needs.

Regarding alternative backends - to be done properly this will require a new layer of abstraction to be added to the engine, it's a major design effort and since we only have two core devs: marciano and volker, and unfortunately they don't have that much time to be doing design work it seems unlikely that we will get alternative backends. Unless someone decides to step up to the task, perhaps doing this as a thesis or some other academic project that will reward the student. The ugly truth is that no one wants to do the boring plumbing code, instead everybody want to focus on their own projects. No wonder, it's tedious, hard work to get right and requires allot of testing and debugging but it does give you amazing insight into the guts of the engine. Anyone in his last year of college looking for a project

I've got my own project I am working on.
And youall are probably right. Geometry shaders might be useless for now.

I doubt that geometry shaders are difficult to integrate. But everyone seems to agree that they are not useful right now, so we will not spend any time on them.

Regarding D3D10: It would be nice to have but we just don't have the time for a full port. However, we will slowly work towards it by making the rendering backend more abstracted and removing all fixed function functionality.

Pretty much anything you can do with GS, can be done more efficiently with HistoPyramids - although they aren't quite as easy to wrap your head around.

_________________
Tristam MacDonald - [swiftcoding]