OpenGL3.0.. I mean 2.2

_the_phantom_ · 2008-09-20T22:49:05

Well, someone has linked to the following on OpenGL.org; http://opengl.org/registry/doc/glspec30.20080811.pdf And, well, thanks but no thanks. All those glorious changes? Gone. The rewritten API? Gone. What we are left with is OpenGL2.2. To quote Eddy Luten from the Opengl.org forum; Quote: For those who don't feel like digging through the spec, OpenGL 3.0 Equals: API support for the new texture lookup, texture format, and integer and unsigned integer capabilities of the OpenGL Shading Language 1.30 specification (GL EXT gpu shader4). Conditional rendering (GL NV conditional render). Fine control over mapping buffer subranges into client space and flushing modified data. Floating-point color and depth internal formats for textures and renderbuffers (GL ARB color buffer float, GL NV depth buffer float, 455 N.2. DEPRECATION MODEL 456 GL ARB texture float, GL EXT packed float, and GL EXT texture shared exponent). Framebuffer objects (GL EXT framebuffer object). Half-float (16-bit) vertex array and pixel data formats (GL NV half float and GL ARB half float pixel). Multisample stretch blit functionality (GL EXT framebuffer multisample and GL EXT framebuffer blit). Non-normalized integer color internal formats for textures and renderbuffers (GL EXT texture integer). One- and two-dimensional layered texture targets (GL EXT texture array). Packed depth/stencil internal formats for combined depth+stencil textures and renderbuffers (GL EXT packed depth stencil). Per-color-attachment blend enables and color writemasks (GL EXT draw buffers2). RGTC specific internal compressed formats (GL EXT texture compression rgtc). Single- and double-channel (R and RG) internal formats for textures and renderbuffers. Transform feedback (GL EXT transform feedback). Vertex array objects (GL APPLE vertex array object). sRGB framebuffer mode (GL EXT framebuffer sRGB) Plus deprecation of older features. As he said, where the hell are the objects? Frankly, this is crap. I said it was a sink or swim moment for the ARB and it's just sunk without a trace. It appears the reason is they don't want to break the API because of all the CAD apps out there (J. Carmack, QuakeCon2008), and in doing so have finally put the nail into the coffin games wise. I'd like to congratulate MS for winning the 3D API 'war' on Windows, turns out they didn't need to sink the goodship OpenGL, the captains ran it into an iceberg for them.

Graphics and GPU Programming Programming OpenGL

Started by _the_phantom_ August 11, 2008 08:56 AM

336 comments, last by JMNightmare 15 years, 7 months ago

_the_phantom_

11,263

Author

August 15, 2008 04:32 AM

Quote:Original post by Dmytry
Quote:Original post by phantom

still no 10% link from you btw. Don't assert "isn't something we've just pulled out of the air" if you dont have a link, ok? I didn't ever notice nvidia opengl people give any % values in articles on subject, too.

Promit's link with 5% trumped the 10%, I was also working on the (incorrect) assumption that the person before had checked the figure (I knew there was one and 10% seemed familar).

Based on the link given the presentation was a joint one between ATI and NV as such the figure can be applied to them both.

_the_phantom_

11,263

Author

August 15, 2008 04:49 AM

Quote:Original post by dorbie
Well I like to view this as good news, it gives me hope that there's the prospect of some change that amounts to something like a real OpenGL 3 still pending. Let's say for example that the deprecation mechanism kicks in with some teeth in 3.1 and we get some key features promoted to core. The landscape would look very different, assuming you're still on planet OpenGL.

Personally, I think the object model is dead in the water now. Introducing it at this stage would be the same as breaking the API, something this release was trying to avoid and would introduce a 3 api land where you have gl2.x, 3.x with no objects and the direct access API and 3.x (4?) with the object API which would be different again.

imo, what should have happened was that this was release was left as 2.2 as that spec implies it was going to be and an update was given in Jan saying the new object model was taking longer than expected and an interim release was coming in August to catch the API up a bit. There would have been annoyance but people tend to be less annoyed when they aren't kept in the dark and know whats going on.

Spoonbender

1,258

August 15, 2008 05:42 AM

Quote:Original post by dorbieLet's say for example that the deprecation mechanism kicks in with some teeth in 3.1 and we get some key features promoted to core.

Yeah, or let's say for example that Santa Claus visits us all with a new graphics API. Sure would be nice, but somewhat unlikely.
Same goes for deprecating functionality. Of course it would be nice if it meant anything, but I'll need to see it before I believe it.

Quote:The landscape would look very different, assuming you're still on planet OpenGL

What would be so different?
We'd still have an OpenGL based on the same 15 year old API, and we still wouldn't have gotten the new object model, or anything like it. We'd still have an GL3.x which required features not available on DX9 hardware, ruling out what, 70% of the market?
How exactly would the landscape look different on planet OpenGL?

Quote:still no 10% link from you btw. Don't assert "isn't something we've just pulled out of the air" if you dont have a link, ok? I didn't ever notice nvidia opengl people give any % values in articles on subject, too.

So why did you give the 1% value? That doesn't seem much more convincing.
How about this compromise:
ATI and Nvidia claimed that the new object model would improve performance of their drivers. Call it 1%, call it 10%, or 10% in certain use cases or 3-5%, does it really matter? Pulling figures out your ass just because you don't believe the figures other people pulled out of their asses isn't all that constructive.
Drivers would be able to perform better with the new object model, had it been adopted. Any objections to that?

Can we move on now, please?

Dmytry

1,151

August 15, 2008 05:48 AM

Quote:Original post by Spoonbender

Quote:still no 10% link from you btw. Don't assert "isn't something we've just pulled out of the air" if you dont have a link, ok? I didn't ever notice nvidia opengl people give any % values in articles on subject, too.

So why did you give the 1% value? That doesn't seem much more convincing.
How about this compromise:
ATI and Nvidia claimed that the new object model would improve performance of their drivers. Call it 1%, call it 10%, or 10% in certain use cases or 3-5%, does it really matter? Pulling figures out your ass just because you don't believe the figures other people pulled out of their asses isn't all that constructive.
Drivers would be able to perform better with the new object model, had it been adopted. Any objections to that?

Can we move on now, please?

I agree that it would be a great thing. Also, i didnt give "1%" value. I did give "likely less than 1%" performance improvement for applications, which is really what i think is the case.
Value of this refactoring is in cleaning up mess.
Its really pathetic that to gaming crowd, you have to pull out various speedup figures to justify anything. Whats about cleanup refactoring that as direct effect would slow things down by 1% but cleanup things a big deal ? How to promote those if all gaming crowd cares about is marginal speed changes, which they would not ever notice?

Additionally, what i think is going on is that it (new object model) is being implemented in drivers, tested, etc, and when its implemented, there will be opengl 3.1 or 4.0 or whatever, and that it was left out of 3.0 spec coz theres no drivers yet. Correct me if i'm wrong, but there exist draft specifications for new object model.

_the_phantom_

11,263

Author

August 15, 2008 06:08 AM

Quote:Original post by Dmytry
Additionally, what i think is going on is that it (new object model) is being implemented in drivers, tested, etc, and when its implemented, there will be opengl 3.1 or 4.0 or whatever, and that it was left out of 3.0 spec coz theres no drivers yet. Correct me if i'm wrong, but there exist draft specifications for new object model.

As far as i can gather the LP Object model was dropped in favour of a "2.2 update" (until a marketing drone decided 'hey, lets call it 3.0!') and the direct state access function thus if a draft spec of any kind exists it was outdated as of Jan08.

Based on ARB comments 3.1 is meant to enforce the deprecation stuff, the only problem being if they bring in 'new features' such as the geo shader this is going to take us right back to the original 'breaking code bases' issue which was thrown around this time. (As in anyone wanting the new features will have to use a 3.1 code base, which is clean, and can't use any legacy 2.x code which is deprecated thus requiring a rewrite)

I'd also assume (if I hadn't already leant that the ARB+assumptions of sanity = lulz) that the direct state access extension, which is currently for 2.x bind points only and not 3.0, will be updated... I mean, it would be logical as not doing so with basically make it useless, but you never know.

gold

412

August 15, 2008 09:49 AM

1) The DSA extension had nothing to do with GL3. It wasn't written by the ARB. The spec went public around the same time, and that's the extent of it. It may or may not ever become core, as the ARB is continuing to discuss object model changes and this obviously overlaps with that work. Nonetheless, for certain use cases it can yield significant benefits for 2.1-based codes.

2) I was the person who originally claimed a 3-5% overhead of name lookup. This wasn't meant to represent the total perf gain from a new object model, since changing the way names/handles are managed was but one aspect of the object model needing to be revamped.

3) Any actual perf gain from changing object models is conjecture at this point since it has never been implemented; however personal experience indicates that significant driver overhead could have been eliminated. Problem is this is based on familiarity with one implementation, and other vendors may/may not benefit to the same degree. We would need to see the same application on both the existing and the LP object model to determine the actual gain. Then again, some vendors aren't concerned with achieving 100% optimal CPU utilization relative to the tradeoffs of backward compatibility and flexibility. (LP is less flexible because of object immutability).

Alpha_ProgDes

6,963

August 15, 2008 11:58 AM

A question... thought.

If DX10 is not widely used (from what I understand most developers use DX9) and so everyone is waiting for DX11, doesn't that make the wait for OGL3 not as bad?

OGL 2.x is DX9 level and most (windows) developers are waiting for DX11 anyway. So there's still some time for the ARB to come out with the real OGL3 (per the spec of last year) and not be wiped out completely.

Maybe the ARB is planning to release OGL3 around the same time as DX11. Who knows?

*DX = D3D

Beginner in Game Development? Read here. And read here.

_the_phantom_

11,263

Author

August 15, 2008 12:09 PM

Quote:Original post by Alpha_ProgDes
Maybe the ARB is planning to release OGL3 around the same time as DX11. Who knows?

Erm, OGL3.0 is out. If you mean Longs Peak that's not likely to happen; for the last 8 months efforts have been put towards the 3.0 spec and DX11's tech preview is out in November and many aspects of it work on current generation DX10 hardware.

pbjgame

100

August 16, 2008 01:13 AM

As one who love(s/ed) OpenGL for it's cross-platform abilities and ease of use...I am not the happiest to hear this news (or lack thereof).

They are lagging so far behind DX. But I don't want to switch to DX because of the distribution problems in the DX9 series, and because of Microsoft's apparent need to force people to upgrade. (For example, making Windows Media Player 10 to be only compatible with systems from XP on, and now releasing DX10 for Vista and on only.) Cash obviously isn't on my good side and I don't think people should be forced to upgrade to such an obscene extent just to play a video game or use a cad app. What will we do if we get stuck with practically nothing but big brother's DX? Why has Microsoft taken a liking for my favorite 3D company (Caligari), bought out the company and started giving away their software and training videos?! Will Truespace become some underfunded side-project and eventually force us with no alternative to compete in the advancing market but Maya and 3DsMax? Does anyone have any idea how we can prevent Microsoft world domination? I probably sound like I'm whining (and I kind of am) but I just want a non-Microsoft environment with Caligari (not Microsoft) Truespace and the REAL OpenGL 3.0 that I hear was promised.

Krohm

5,051

August 16, 2008 02:30 AM

They could have least called it "OpenGL 3", dropping the min-version notation. I think there was a rumor in this regard.
After all, the versions are basically arbitrary strings now. It could be "OpenGL 666" for what I care at this point.
[rant]
It's just lame that some things proposed for GL2.0 by 3DLabs still aren't there.
They cannot just say that they went over-schedule. They cannot just say that they are unsure how to manage the alpha ref in rasterizer objects.

I believe most people is missing that point: what makes the thing terrible isn't the fact it misses this or that but the point that it misses at least most goals after so much time. Sounds like a terrible internal mismanagement to me considering the big names and surely budgets involved. Not to mention the communication issue. I agree maybe someone pushed to proof they're getting the money for something.
Obviously if this is how the things go, I don't know why I should hold my breath another 3-6-X months.

And no, it isn't as good as it was yesterday. Yesterday the spec was quite old. It was just acceptable. It isn't acceptable a new spec to not carry GS considering DX10 cards (albeit at low performance) are now commodity items at how much? 40 currency units? (Intel admittedly not included)
...but since GL users are used to extensions, there isn't much point in putting stuff in core in the first place. Everybody knows nobody uses plain Core-GL, I wonder why they released (see above).

I don't even know why NV puts resources in GL at this point. They could just say goodbye at least when it comes to win32, other vendors are doing this since years and did you notice nobody is really complaining besides us?
Apple delivers their drivers, they seem to be working fine for me, users seems to be ok with them(I expect somebody to tell me this is not always the case), linux users seems to be happy spending tens of hours trying every possible driver installation method or lagging behind a few generations... As for other non-mainstream systems, I'm rather sure they don't iterate on a mass-marketing basis. It's ok. It's just ok.
(???)
[/rant]

Previously "Krohm"

OpenGL3.0.. I mean 2.2

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

OpenGL3.0.. I mean 2.2

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines