The OpenGL Pipeline 004

[NocturnDragon]

The OpenGL Pipeline volume 004 is out.
The latest issue covers:

• Climbing OpenGL Longs Peak, Camp 3: An OpenGL ARB Progress Update
• Shaders Go Mobile: Announcing OpenGL ES 2.0
• Longs Peak Update: Buffer Object Improvements
• Another Object Lesson
• Transforming OpenGL Debugging to a "White Box" Model

http://www.opengl.org/pipeline/vol004/

 

[Rodéric]

good news, only need the specs (even a draft) and driver support (even beta) to make me happy !

 

[Demirug]

I agree but as I need full D3D10 feature support I have to wait even longer.

 

[compres]

This situation is really worrying me, and I am rarely worried about anything. The 3d market is really going to be owned by Microsoft if opengl does not get in the game.

 

[Anteru]

Well, it seems they are pushing it quite hard. However, I fear it is going to lack the coherence of DirectX (i.e. DX10, SDK, DXUT, FX, they all just fit together, and not to forget PIX) - I simply don't get why Khronos cannot provide a real OpenGL SDK which looks like it comes from a single company. With OpenGL, I always have the feeling nobody wants to take the lead and do something, be it tools or docs.

However, if they manage to bring out a real SDK, solve the per-platform windowing issues (no WGL etc.) and get together some tools - then I believe more people will look into OpenGL. After all, there is no other way to exploit DX10 cards on XP. Mac gaming might play a role, too.

 

[K.I.L.E.R]

The windowing issues and misc library stuff is easily solved via third party libraries. I would love to see the stuff inside an SDK for OpenGL but is it really necessary?

 

[Anteru]

I would love to see the stuff inside an SDK for OpenGL but is it really necessary?

It would give me this feeling that this is OpenGL, with everything included, just like DX so I don't have to look for 3rd-Party solutions for rather trivial problems. It's more an attitude thing: If you leave that out, you start to rely on other parties in other segments as well, and eventually you end up with an API that needs help from 3rd parties to get started. I don't have anything against the free solutions to that problem, but I don't understand why the OpenGL ARB or Khronos can't agree on one, make it standard and add it to the next OpenGL? After all, we all see how nice DX development is cause you don't need to 3rd party tools.

 

[Rodéric]

I agree but as I need full D3D10 feature support I have to wait even longer.

Already available on GF8 with OpenGL...

 

[Demirug]

Already available on GF8 with OpenGL...

I know. I am already using it but LP/ME would simplify some tasks.

 

[Rodéric]

I know. I am already using it but LP/ME would simplify some tasks.

It was for the sack of correctness, I completely understand you and agree that LP is really strongly needed at this time.

 

[Andrew Lauritzen]

Already available on GF8 with OpenGL...

Yeah let's not get too hasty... there are several features available in DX10 that are still not available in OpenGL, even with extensions, including some mind-blowing ones like rendering to 1/2-component FBOs, and rendering without a VBO bound (using vertex IDs).

I have hopes that the new GL versions will correct these issues and many others, but they really need to come soon... I really like GL, but it's becoming unuseable compared to DX, and that needs to be corrected ASAP.

 

[GLX]

Yeah let's not get too hasty... there are several features available in DX10 that are still not available in OpenGL, even with extensions, including some mind-blowing ones like rendering to 1/2-component FBOs, and rendering without a VBO bound (using vertex IDs).

One thing to keep in mind is that LP doesn't obsolete all the pre-DX10 hardware, it is going down a path that is shader-only and no fixed function, but not limited to the fourth generation shader hardware such as NV 8800 the way that DX10 is.

Can you elaborate on the DX10 features you mentioned and what makes then useful? If you have any URL's that could go into a little more detail I would enjoy reading them.

 

[Andrew Lauritzen]

One thing to keep in mind is that LP doesn't obsolete all the pre-DX10 hardware, it is going down a path that is shader-only and no fixed function, but not limited to the fourth generation shader hardware such as NV 8800 the way that DX10 is.

Sure but honestly I don't care about anything pre-DX10 although I can understand that some people certainly do Still I don't think it's unreasonable to target DX10 hardware exclusively if one is *starting* to develop a game today.

Can you elaborate on the DX10 features you mentioned and what makes then useful?

Oh man it's absolutely ridiculous that you can't render to one/two-component textures... that's useful for *so* many things, not the least of which being shadows. Rendering without input vertex buffers is very useful for R2VB-like functionality (via vertex texture fetch) and scatter (which is a fundamental part of *many* algorithms).

Unfortunately I don't have anu URLs handy right now (I'm not at home), but Google-ing you could probably find a few.

 

[Zengar]

The Mount Evans (DX10-only equivalent) is coming just some month later... Actually I am very exited about the new APi, as I believe it will be more clear then the DX (all that function and token names drive me crasy . Also, if the IHV(ATI, I am talking to you!) provide reasonable implementations, the new API will have some real potential! Maybe not now, but in several years for sure...

 

[K.I.L.E.R]

Layering the new API with classes will become much more natural, I'm looking forward to it as well.

 

[GLX]

Oh man it's absolutely ridiculous that you can't render to one/two-component textures... that's useful for *so* many things, not the least of which being shadows. Rendering without input vertex buffers is very useful for R2VB-like functionality (via vertex texture fetch) and scatter (which is a fundamental part of *many* algorithms).

Are either of these techniques possible on DX9-class GPU's ? Thanks for the supporting detail.

 

[Andrew Lauritzen]

Are either of these techniques possible on DX9-class GPU's ? Thanks for the supporting detail.

Yes, particularly the rendering to a one-component render target (you could probably do that in DX8.1 even, but definitely in DX9). And for the record those are just the first two examples of unforgivable annoyances - there are more (although DX9 has its own set as well, and DX10 has a few).

 

[Entropy]

Still I don't think it's unreasonable to target DX10 hardware exclusively if one is *starting* to develop a game today.

When would you say 90% of the game buyers will have switched to Vista?
How much would you be prepared to place on that bet?

 

[Rodéric]

When would you say 90% of the game buyers will have switched to Vista?
How much would you be prepared to place on that bet?

Vista does support DX9 too you know, Aero Glass for example is not DX10 bound, you can run it on GF7.
You would need most of your users to have Vista and have DX10 compatible hardware...
That's not gonna happen before (gonna throw a random number around it's unpredictable) 5 years.
(Must be around the time the next MS OS ships)

 

[Entropy]

Vista does support DX9 too you know, Aero Glass for example is not DX10 bound, you can run it on GF7.
You would need most of your users to have Vista and have DX10 compatible hardware...
That's not gonna happen before (gonna throw a random number around it's unpredictable) 5 years.
(Must be around the time the next MS OS ships)

The suggestion was that since DX10 requires Vista, we have the situation that the combined adoption rate of Vista and DX10 capable hardware is rate limiting. And given the adoption rate so far, Vista seems to be more of a limitation, and perhaps even more so among people who play games.