OpenGL ARB notes are up

[Humus]

Just saw this:
http://www.opengl.org/developers/about/arb/notes/meeting_note_2002-12-10.html

Haven't read yet, just about to do that.

 

[Humus]

Ãœber buffers sounds cool

 

[Panajev2001a]

Apple's feedback is that their customers don't need more than 10 bits/pixel integer components today.

this doesn't sound uber-cool however... no color banding (quasi ) free 3D for gamers on Macs ( no FP frame-buffer ) ?

 

[WaltC]

Apple's feedback is that their customers don't need more than 10 bits/pixel integer components today.

this doesn't sound uber-cool however... no color banding (quasi ) free 3D for gamers on Macs ( no FP frame-buffer ) ?

*chuckle* And all this time Apple's been funning us all by claiming to be a professional graphics solution....

 

[Panajev2001a]

they are not ????

 

[Rodéric]

Interesting, many things in the work...

 

[demalion]

Floating point framebuffers, not floating point processing.

That's the way I understood it, though I don't know what you'd use a floating point framebuffer for.

 

[LeStoffer]

... though I don't know what you'd use a floating point framebuffer for.

Carmack is all over it. See last piece of his plan...

 

[Joe DeFuria]

Actually, I think it needs further clarification....

I know carmack has been a proponent of high precision floating point back buffers, but not necessarily the display frame buffer.

It's not clear to me if the ARB is talking about the back buffer or the display buffer. (I thought the NV30 and R300 already supported floating point back-buffers....)

 

[Joe DeFuria]

I found this interesting:

Revisited issue 23, ability to read from the framebuffer. After considerable discussion, including the performance costs of this in deeply pipelined architectures and desirability of overturning WG decisions in the full ARB, we revoted and decided this would not be supported by a large margin.

I think there have been some discussions here on this before...(Humus?) on the merits or lack therof being being able to read the framebuffer. Apparently, IHVs don't like it.

 

[Joe DeFuria]

Heh....score one for my memory:

http://www.beyond3d.com/forum/viewtopic.php?p=11232&highlight=frame+buffer#11232

The question is, is the ARB talking about the back or front buffer? (or both?) I've always seen "Frame-Buffer" as an all-encompassing term to mean both (or all three buffers in a triple-buffer case.) But I don't know how "the industry" defines "frame buffer."

Does this mean the ARB has decided to not allow back-buffer reads?

 

[archie4oz]

*chuckle* And all this time Apple's been funning us all by claiming to be a professional graphics solution....

this doesn't sound uber-cool however... no color banding (quasi ) free 3D for gamers on Macs ( no FP frame-buffer ) ?

It *is* rather funny as Apple has their own cross-vendor float-pixel extension... The same with fences, vertex array objects and ranges...

 

[Jallen]

I found this interesting:

Revisited issue 23, ability to read from the framebuffer. After considerable discussion, including the performance costs of this in deeply pipelined architectures and desirability of overturning WG decisions in the full ARB, we revoted and decided this would not be supported by a large margin.

I think there have been some discussions here on this before...(Humus?) on the merits or lack therof being being able to read the framebuffer. Apparently, IHVs don't like it.

You can actually do this right now using ARB_render_texture (at least on NVIDIA cards). You just bind the pbuffer as a texture before drawing to it, then access the pbuffer texture while you are drawing to it.

No guarantee that this will work in the future since the ARB_render_texture spec doesn't allow it but interesting that it works and could probably be useful for testing some algorithms.

 

[Joe DeFuria]

Hrumph...

You can actually do this right now using ARB_render_texture (at least on NVIDIA cards).....[however] the ARB_render_texture spec doesn't allow it.

On a side note...I'd just like to point this out as an example of how code "works" on architecture A, "break" in another architecture B, and then is subsequently blamed on "bad drivers on architecture B." I don't know if ATI's drivers allow the frame buffer reads, but if they don't, then one's code that runs perfectly fine on nVidia hardware will fail on ATI hardware....though no fault of ATI. (I'd be very surprised if there are no examples in the other direction as well.)

These are the kinds of pitfalls that you run into when developing using a single architeture....

 

[Jallen]

Yes, it should definitely never be used in a production environment

 

[Humus]

Does this mean the ARB has decided to not allow back-buffer reads?

It means the shader doesn't have direct access to framebuffer values. This means no special tricks ala DeltaChrome and no custom blending in the shaders.

 

[Humus]

I don't know if ATI's drivers allow the frame buffer reads

According to some ATi reps it works for ATi cards too, at least right now, no garantuee it will work tomorrow or on future cards.

 

[UberLord]

Quick Q - what does ARB stand for and what does it do/mean?

 

[Doomtrooper]

Architecture Review Board

 

[Joe DeFuria]

http://www.opengl.org/developers/about/arb.html