Rich Geldreich Leaves Valves; Points To More "End Of OpenGL" Articles

[wizard69]

Well, one, there is no working draft of OGL5 yet, so we have no idea what it will be like. If they do it right, it could be a new profile of extremely low overhead functionality that can release concurrently with GLES4 to provide similar workflows for both. If they do it wrong, they could just add another few dozen functions rendered useless in 5 years besides the need to implement them for standards compatibility even if nobody uses them.

The problem with OpenGL is that it has a lot of stake holders not even remotely involved in the gaming industry. They do not want and can't afford rapid changes to the APIs nor the reliability of OpenGL.

Second, while I think concepts like Metal and Mantle are stupid, that doesn't mean they don't have validity

If the have validity then they can't realistically be stupid. These technologies are realities tic ways to address the fact that OpenGL won't change in the way that game developer need. For the same reason we have developers working in C++ and Python the two platforms can exist with no problem. Each serves a specific need and you might see Metal as the C++ of the Graphics world whole OpenGL is the Python of the graphics world.

Yeah I know the comparison above isn't perfect but I think it highlights the major reality here Metal is about performance.

- most platforms have single-platform toolkits already, and also effectively have single platform languages (objc, java on dalvik, c# on .net, hell Vala with GTK on Linux) so it stands to reason that having platform specific graphics APIs doesn't mean they instantly fail, since you are often writing platform specific code already.

The value of and the desire for cross platform development is often massively over exaggerated. Unless you have an extremely trivial app you really want to leverage the platform you are running on. The especially in the context of portable devices where you literally need to get every once of performance out of the processors.

But it does mean your value proposition has to be really good to match up against cross platform write once debug lightly everywhere deploy with very little per platform modification solutions like Qt, Mono, or HTML5.

Why would you bring QT, Mono or HTML into a discussion about OpenGL and fast 3D libraries?

In the same sense, any non-open graphics standard has to be significantly better for either the developer or user than OGL / ES to woo investment over.

That is rather easy to do as performance means everything to se developers. Even if you aren't a developer that needs performance you may end up using the likes of Metal due to third party libraries. In a nut shell it doesn't have to be significantly better, it just needs reasonably quick adoption.

And yeah, it can happen. Like I said, it happens with toolkits all the time. But you would have a real hard time persuading me, if I were investing in a software project, to not use a cross platform toolkit because to be platform locked means to vastly reduce my potential audience size. Same with graphics API commitments.

You know I hear this nonsense all the time and frankly I'm beginning to wonder about some of these developer spouting this crap. The last thing we need more of in this world is cross platform tool kits! Is it really that hard tout your app logic into a bit of portable C++ and then skin it with a GUI of your choice?

[log0]

Praising metal seems like a sound thing to do.. riiiight..

So lets see, It is apple ONLY and ONLY for A7 arm, to be more specific, it will probably be PowerVR specific and only a subset of those chips will be supported.
To get the same kind of low cost API call access that AMD has achieved with Mantle, PowerVR will probably need hardware changes to be able to communicate in that way.

Not only does Imagination tech need to add this hardware feature to PowerVR for metal to work, they also are going to need a driver that targets metal (I would expect Imagination tech devs not being all that happy with apple atm..), unless of course apple is going to write drivers for them. Whoever writes these drivers, it WILL take a while before they work well. Because that's how things go.

Now, anyone that is not mantle or imagination tech, will probably not be wanting to use metal, just because it makes no damn sense to. Metal, and Mantle as well, while their APIs have a chance of becoming open, will probably carry a bias towards their first user.

And since some programmers seem convinced that the same effect can be achieved in OpenGL(the hardware support for which would have already been added in GL4.3 or 4.4 or something, can't remember), it is probably entirely possible to do it in OpenGL.
Now the problem is probably that OpenGL is buggy and strange in some ways, but as long as ONE developer can use it in a good way, to the wanted effect of mantle and metal, it serves its purpose, and everyone else just have to adopt to that best practice of using OpenGL.

Afaik the driver is already there, see ue4 engine demo.

But you are correct, those are highly specific apis:
mantle/windows/gcn
metal/ios/a7

So the only future I see are the next dx, gl api versions.

[gamerk2]

I'm still really trying to understand WHY openGL is so much worse. It seems several developers lately complain about it but don't give specifics. They just give the most vague statements. I understand it might not be as polished as directX, but what I really don't get is if openGL is open source, couldn't you just simply modify it at the source to fix where it fails? That being said, why is something like OGL5 necessary when you could just theoretically re-write openGL to have less CPU overhead while retaining the exact same functionality? To put it another way, why re-invent the wheel when you can take the exact same model and just use a different material?

I don't know enough about how OGL works on the lower level, and especially the hardware level. But there just seems to be too many "plot holes" for the claimed problems of OGL.

Rich gave a number of examples here:

http://richg42.blogspot.hu/2014/05/things-that-drive-me-nuts-about-opengl.html

For the most part, the issues with OGL isn't the performance, it just a PITA to actually code with.

[nilssab]

Afaik the driver is already there, see ue4 engine demo.

But you are correct, those are highly specific apis:
mantle/windows/gcn
metal/ios/a7

So the only future I see are the next dx, gl api versions.

Ah, didn't see the demo, but I think it will still take time, mantle had a tech demo long ago, but still only a beta driver afaik.

yeah, people seem to forget at how many levels things need to change and how (un)likely it is that all those layers would be working before OpenGL can fix it's driver and API layer problems if the same effort were to be put towards that instead.

[yogi_berra]

I'm in kind of a special position to know what Rich Geldreich has been up to. He was not fired from Valve. Did not join Apple. Is currently not working for anyone and taking a break. Did not turn his back on OpenGL because he wanted to. OpenGL is flawed in many ways and he pointed out those flaws to save everyone a headache down the road. He either had very good or very bad timing on his blog posts. What he's really up to is cleaning out the shed before I strangle him! Now you know... The truth shall set you free! Everyone can relax now and Rich will be able to say, I told you so one of these days, when everyone figures out he was right.

Yeah, don't take anything said on Moronix seriously.

[Akdor 1154]

Yeah, don't take anything said on Moronix seriously.

Haha, it took three pages for an actual primary source to the scenario the article is about to even get acknowledged. And now everyone is going to go back to arguing over their pet APIs that they've probably never used.

Christ this place is a joke.

[wizard69]

I'm still really trying to understand WHY openGL is so much worse.

It isn't worst. The big problem I have with this discussion is people reading too much into what is happening here. In a nut shell Metal provides an API destined for a different user base. You have to remember Where OpenGL came from, it really wasn't meant to be a high performance gaming platform.

It seems several developers lately complain about it but don't give specifics. They just give the most vague statements. I understand it might not be as polished as directX, but what I really don't get is if openGL is open source, couldn't you just simply modify it at the source to fix where it fails?

Because there are many stake holders in the OpenGL world most whom have nothing to do with gaming. Those stake holders simply can't afford to have OpenGL evolve in ways that breaks their software.

That being said, why is something like OGL5 necessary when you could just theoretically re-write openGL to have less CPU overhead while retaining the exact same functionality? To put it another way, why re-invent the wheel when you can take the exact same model and just use a different material?

If they could don't you think that that would be happening already?

I don't know enough about how OGL works on the lower level, and especially the hardware level. But there just seems to be too many "plot holes" for the claimed problems of OGL.

Well you will have some involved in the marketing aspects of promotion of new technologies. However some of the problems with OpenGL have been around for a long time.

In the end I really don't understand what all the fear is about. Here we have nothing more than a new technology that on the face is very interesting. It is really up to the individual developer to get onboard or not. In the case of this thread maybe I should say get left behind or not.

[wizard69]

So what if Ubisoft supports Mantle? Pretty much all games revolve around DX9 level graphics. If it runs on Xbox 360 and PS3, then you aren't pushing your PC as much as you would think. Mantle will only work if AMD opens it up 100%.

Mantle works if only a few game developers adopt it. This idea that something has to be 100% adopted to be successful is nonsense.

From the looks of it, that'll be a cold day in hell. They can't even get it working on HD5000/6000 cards. Plus benchmarks show results to be iffy. Take a look for yourself.

Yeah not good now if you believe their testing but look back in time to the trash that was Direct X. It takes time to message everything into a high performance API.

These API's are only meant to lock developers onto a platform.

Nope, it is meant to leverage the available hardware more completely.

Apple is losing a lot to Android right now, and they figure Metal will give them the edge they need,

I see this posted often, apparently by people scared that their preferred platform isn't doing as well as they like. Right now Apple has 42% of the market that is not a sign that a company is losing. Beyond that things like Metal take years to develop, literally so if we believe you then we also have to believe Apple has a crustal ball to see this "losing" coming.

along with code that only works on their platform. Lets hope that OpenGL 5 isn't a pile of trash. Considering how confidant Apple/AMD are, it may just be.

OpenGL is not trash, it however has many users outside of the gaming world. As such it will never be a high performance gaming API.

[nilssab]

If they could don't you think that that would be happening already?

"They" are.. They being the driver teams of each and every GPU on the market that sports the OpenGL API. Because OpenGL is just an API protocol, not a piece of software, the API has to be implemented by the GPU drivers.
These *are* being worked on, and the "innards" of OpenGL are always subject to optimizations.
The problem is that for the linux drivers most GPU makers don't make drivers with decent OpenGL implementations, and the mesa drivers are probably working towards full OpenGL 4.4 compliance rather than optimizations. Even the windows intel drivers don't have full support for OpenGL, so games that use specific extensions will just crash.
THIS is what needs work, from GPU vendors and from the community, to get the OpenGL implementations working flawlessly, instead of making a new API that will see the same problems. Also, if the extensions people think will rival mantle and metal prove to not suffice, further extensions are needed, but at the moment no one knows if the current extensions will be enough.

[wizard69]

There is two things wrong with your post. 1. API's were not covered by copyright in the past. 2. The decision was overturned on appeal (at least for the moment google and software development has lost)

Google has lost but Software development has won! This means you can apply the type of license you want to an API to protect it. This is a good thing, especially in the open source world. It means if you are a crazy neck beard GPL believer your API along with your code shares the same level of protection.

If anything this is a strong decision in support of copyright and the ability of individuals to license their software.