Next-Gen OpenGL To Be Shown Off Next Month

[johnc]

Ah yes, AMD's "closed API"... the one they gave 100% of over to Khronos to use in OpenGL-Next. So closed. Much proprietary. wow.
The point is that AMD revealed a new API that they were going to make Open (and thus contend with OpenGL) as soon as it was actually ready (a thing all too often overlooked these days...) and NVidia didn't like that. They like to be able to bend the rules in their favor.

Ohh yeah. Mantle was going to be open. Some day. At the end of 2013. No wait, mid-2014. No wait, end of 2014.

Don't worry... it's an open API though.

So far we have nothing... but it's open.

[grndzro]

Here is to hoping Khronos has all their ducks in a row because DX12 is coming and with it a lot of MS$$$ to get it adopted as the standard.

[justmy2cents]

Ah yes, AMD's "closed API"... the one they gave 100% of over to Khronos to use in OpenGL-Next. So closed. Much proprietary. wow.
The point is that AMD revealed a new API that they were going to make Open (and thus contend with OpenGL) as soon as it was actually ready (a thing all too often overlooked these days...) and NVidia didn't like that. They like to be able to bend the rules in their favor.

that was actually first open move with Mantle.

meanwhile, for anyone but Khronos there is page 43 on http://de.slideshare.net/DevCentralA...s-amd-at-gdc14 with highlights
- beta sdk access in april
- NDA required

how do open API and NDA connect? hopefully, that is only for beta access.

[SXX⁣]

Ah yes, AMD's "closed API"... the one they gave 100% of over to Khronos to use in OpenGL-Next. So closed. Much proprietary. wow.

You're likely missing what "Open" part of "OpenGL" mean. It's not about openness of specification, but about GL being industry standard developed by multiple vendors in cooperation and fact that everyone interested able to join Khronos.

In this meaning Mantle will be never open standard just like CUDA aren't.

[richgel999]

Just speculation:

I'm guessing NVidia and/or AMD have taken the AZDO extension bits, gotten rid of the old crusty stuff (i.e. most of what we know today as OpenGL 3/4), and rolled what's left plus a new fresh coat of paint (like the recent NV command list extensions stuff) into the new API. I would be surprised if the new API didn't make it relatively easy for AZDO enabled apps to be mostly copied & pasted into the new paradigm. A relatively small driver team could do this work because the AZDO details have already been proven and worked out between the vendors. Devs could add support for AZDO right now (using extensions) and know they're mostly prepared for the new API. Most vendors made it pretty clear they were all behind the AZDO stuff last year.

Also, kinda related to this: There are a lot of subtle backroom politics going on here. I believe the existence (and threat) of both Mantle and Metal (and less D3D12) had a lot to do with Khronos starting up GL Next. Also, it's in Valve's best interest to make sure GL Next goes somewhere, otherwise Apple could just take their ball home and push a Metal-like API onto OSX (just like MS does with Direct3D) and let OSX GL support mostly rot on the vine.

The reason: A whole bunch of the Steam Linux ports that we've seen are downstream of the OSX ports companies make (or pay others to do). Porting to OSX makes reasonable financial sense -- the market is small but there. Once you're on OSX (i.e. after switching to POSIX+OpenGL instead of Win32+D3D) doing that final, easy port to Linux for that extra nudge (.5-1% or whatever) in sales and exposure is a relatively easy pitch. But, if Apple ditched GL for Metal-OSX then porting to Linux potentially becomes that much more expensive and harder to justify.

[Marc Driftmeyer]

Just speculation:

I'm guessing NVidia and/or AMD have taken the AZDO extension bits, gotten rid of the old crusty stuff (i.e. most of what we know today as OpenGL 3/4), and rolled what's left plus a new fresh coat of paint (like the recent NV command list extensions stuff) into the new API. I would be surprised if the new API didn't make it relatively easy for AZDO enabled apps to be mostly copied & pasted into the new paradigm. A relatively small driver team could do this work because the AZDO details have already been proven and worked out between the vendors. Devs could add support for AZDO right now (using extensions) and know they're mostly prepared for the new API. Most vendors made it pretty clear they were all behind the AZDO stuff last year.

Also, kinda related to this: There are a lot of subtle backroom politics going on here. I believe the existence (and threat) of both Mantle and Metal (and less D3D12) had a lot to do with Khronos starting up GL Next. Also, it's in Valve's best interest to make sure GL Next goes somewhere, otherwise Apple could just take their ball home and push a Metal-like API onto OSX (just like MS does with Direct3D) and let OSX GL support mostly rot on the vine.

The reason: A whole bunch of the Steam Linux ports that we've seen are downstream of the OSX ports companies make (or pay others to do). Porting to OSX makes reasonable financial sense -- the market is small but there. Once you're on OSX (i.e. after switching to POSIX+OpenGL instead of Win32+D3D) doing that final, easy port to Linux for that extra nudge (.5-1% or whatever) in sales and exposure is a relatively easy pitch. But, if Apple ditched GL for Metal-OSX then porting to Linux potentially becomes that much more expensive and harder to justify.

The market for OS X is a 22-25 million new hardware and climbing yearly market, and then the 100 million plus install base. And Linux is nowhere near it. Put in iOS and it's average 200 million yearly market explains how come Apple is the largest company around.

[haagch]

Ah yes, AMD's "closed API"... the one they gave 100% of over to Khronos to use in OpenGL-Next. So closed. Much proprietary. wow.
The point is that AMD revealed a new API that they were going to make Open (and thus contend with OpenGL) as soon as it was actually ready

Nope. They put it on their website: "a public release of the specifications later in 2014". http://support.amd.com/en-us/search/faq/184

They didn't even bother correcting this. They'll still release it later in 2014.

So far it is secret and proprietary windows-only technology as far as we are concerned. Only AAA developers and industry consortiums can even look at the specification.

[mufasa72]

I guess we have a long way to go before we get full OpenGL support in Mesa. Maybe things will speed up once NIR matures and AMDGPU is finished and the new AMD cards are enabled.

Supposing the glNext is a complete new slate for the OpenGL API, wouldn't it be possible to just start writing support for glNext since it won't depend on OpenGL x.y?

[Daktyl198]

Nope. They put it on their website: "a public release of the specifications later in 2014". http://support.amd.com/en-us/search/faq/184

They didn't even bother correcting this. They'll still release it later in 2014.

So far it is secret and proprietary windows-only technology as far as we are concerned. Only AAA developers and industry consortiums can even look at the specification.

That might have been the original plan, but since they decided to give Mantle to Khronos to use in OGL-Next they probably delayed it to see what Khronos would do. If Khronos decided to use Mantle as a base, then there'd really be no reason in AMD releasing it, would there?

[BSDude]

I could think of a "real-world application" Valve could show to demonstrate OpenGL Next.
It would invovle rendering lots of headcrabs.....

HL3 confirmed!