Khronos Group Announces Vulkan, OpenCL 2.1, SPIR-V

[smitty3268]

I know its too soon to ask something like that, but, if Vulkan API ends up being really close to the Gallium3D one (as already stated here), would be interesting for mesa developers to change Gallium interfaces to match Vulkan ? Is there some internal discussion about it ?

It would definitely be interesting to update Gallium to use SPIR-V directly instead of TGSI. I don't know if it'll happen or not, though - there's a lot of code relying on TGSI at this point that would all have to get ported, and I'm sure everyone is busy enough trying to support the new APIs even without adding additional work.

The rest of the Gallium interfaces probably won't change much. They're still going to want it to be able to support misc APIs, such as OpenGL, Direct3D, etc, and that will all require pretty much the same functionality it has now. The new Vulkan features might need a few additional methods added, but that should be it.

[smitty3268]

this is not true. you can, for example (to unquestionably proof my standpoint) search for games on sourceforge, you will rarely find any game that requires anything above the classic 1.x opengl api.

LOL.

Judging by all the games on GOG, nobody uses anything past DX7. Why does anybody even bother with OpenGL or DX9?

[smitty3268]

No, Vulkan is a programming language like Java. Just like Java gets translated to a jar file, Vulcan and OpenCL 2.0 get translated to SPIR-V.

That's actually false.

GLSL is the programming language, which gets translated to SPIR-V. OpenCL kernels also get translated to SPIR-V (while the non-kernel OpenCL code does not).

Vulkan is an API. It stays in C, and doesn't get translated to anything.


The difference is that shader code running on the GPU = SPIR-V. Driver code running on the CPU = Vulkan.

SPIR-V is comparable to a jar file with java: it only gets translated to x86 at run-time. This ensures cross-platform compatibility.

Yeah, that bit is correct.

[Geri]

LOL.

Judging by all the games on GOG, nobody uses anything past DX7. Why does anybody even bother with OpenGL or DX9?

your funny analogy actually shadows a serious technical question.

why we use directx9, and not dx7, if we would have bigger compatibility with direct7?

the answer is: false we would not get better compatibility with directx7.
directx9 have actually better compatibility, than directx7.

directx7 is not properly supported (or not supported at all) on newer windows versions.
this means, using the dx7 api, you will not be able to deliver your software for a wide-enough community.

while directx9 have better compatibility, and the ability, to cooperate with dx11, 10, 9, 8, 7, 6, and 5 (or maybe even dx3) cards properly.

by using opengl or directx9 on windows, this is not an issue. with opengl, you can scale down your application even to support most graphics chips from 1995 to forever.
the same happens, when you are using directx9. directx9 can scale down to support all of the previously released directx compatible graphics cards, you basically even can use the newer vertex paths (with some limitations in some cases). for example, you can use an S3 Virge 4 mbyte with directx9 properly, without any problem. of course you will not be able to use shaders, and large textures on it, but you still can support extremely old graphics cards from both opengl and directx9, if you want - and if there is opengl and/or directx compatible drivers are available for that system at all. i dont want to tell with this that supporting an s3 virge has anything to do with 2015, this was just an example. but the fact that your software can run from a simply pentium1+3d graphics card computer TO the newest computer hardwares, so basically on every windows machine, even without modificiation of the code, by using opengl or directx9, results that using these apis are a better choice than using directx7, wich is not supported on newer computers, only on older operating systems.

therefore, there is no point using directx7.

[clementl]

previously, you calimed significant opengl 3/4 demand. so now you are telling that there is basically almost no opengl 3.3 game. which is true, thats what i told before.

Looking around in the steam store, most of the more graphically demanding games require OpenGL 3.3 or more.

also your imagination about the number of directx11 games is flawed.

http://en.wikipedia.org/wiki/List_of_games_with_DirectX_11_support

there is actually a very small number of PC games supporting directx11.
most directx-fan PC programmers actually use directx9 in they game, becouse they want to sell it, dx11 is still not yet widely supported enough (and probably never will).

You call that a small number?
Assassin's Creed belongs to the better selling AAA games of this generation. Same counts for Battlefield, Call of Duty, Civilization and a whole lot of other games in there.
We're also going a bit offtopic: we're talking about the demand for a high-performance API. Looking at the wiki page you posted, it's obvious a lot of more graphically demanding games seem to be utilizing newer API's

converting opengles to desktop opengl its not so terrible, probably a few day of work.

Work nonetheless. And now the other way around: how much work is it to port a PC game to mobile devices? A ton when you're making use of complex engine's.

correct. many need opengl es2 - and many are still using opengl es1.x, becouse its just fine for them, and they dont see any point of converting it to opengl ES2 - and they will never convert it to Vulkan/whatever. They actually didnt even ported it to directx, thats why microsoft falled miserably the Metro apps market.

And this is where you seem to be missing the point. OpenGL will still be around. Vulkan isn't targeting low demanding games. Those don't need thorough optimizations.
Vulkan is for Battlefield 5, COD 14, The Elders Scrolls 6 etc. Graphically demanding games.

i would like to see, where is the money in an API'that does not even supported on 90%-99,99% of the computers (OpenGL 3.x, 4, Vulkan, OpenCL)...
you users actually want to run your software. So Narnia API-s canot be used for such purposes.

77.5% of Steam users has a DX11 capable GPU. That's 77.5% of 125.000.000 users. Since every DX11 capable GPU out there is OpenGL ES 3.1 compatible, it means those 96.875.000 users have an Vulkan capable GPU.

see, you start getting the point.
(most) games dont need more modern API-s any more (or they dont even need an API at all). So, whats the point of the whole Vulkan thing?

Some certainly do. And for those there's Vulkan.
There will always be a demand for high performance 3d-renderering. If virtual reality becomes a succes even more so.

not really... developers does not really using anything. They does not even using threads, becouse it was not required when they learn the hello world 20 years before. Only in the last 2-3 year, in very demanding applications, they started to use threads. So now, in 2015, we can say that the most machine demanding softwares are actually using multiple threads. In 2012, this was still not true. It take 10 years for programmers to start using thads actively - and thats just basically ONE LINE IN THE CODE a (createthread/pthread_create). imagine, what color they vomit, if they see any gpgpu-narnia api requiring ,,compute shaders''. also, it does not even works in the reality for the most cases. we can be thankful for having multiple cores, and for having multiple threads. thats something that works now, and will probably the only thing, that we will have in reality in the next 30 years. so yeah, the computing/directwhatever/opencl/opengl3/4 universe does not really exist outside the laboratory.

Nice that not multithreading works for your solution, a high demanding graphical application does certainly need them.

and now, we get another bullshit api, called Vulkan, that looks like some drug addict randomly mixed opengl functions with random shader dialects after consuming some heorin, and is just as unusable like the previous ones. programmers are tired of this non-neumann architectures, so i hope, it cost brutal amout of money for the manufacturers to release this api, becouse every forint on this is just a waste of money, a self-deception in a dreamworld.

Honestly, the Mantle and Vulkan code I've seen so far is much, much more readable and usable than any fixed-function or OpenGL 3.3 system I've built so far. And no, I'm not going to do 3d rendering on a CPU.

(i dont want to ruin your chain of (ill)logic, but ogles1.1 neither has that function, just to be precise, only types of the vertex arrays/elements. )

Are we starting to throw insults around now? If something I said doesn't equate you should be more concrete.
Also, you're not ruining anything. I just overlooked the fact that OpenGL ES 1.1 also for rid of the fixed-function pipeline. Doesn't help the case you made earlier.

we alreday having opengl and opengles as crossplatform graphics api.

And a lot of developers hate it. With good reason. I'm sorry to say it: DirectX is just much handier. It's shown we can do better, so now we are actually doing something to better it.

that lot game is actually an insignificant tiny, countable number.

Not when it comes to high-resource AAA games. Yes when comparing to tiny mobile apps.

dont misunderstand me, i am not against innovations, but the Vulkan API is just not one.

Not for your average 2d sidescroller no. There are higher demanding games out there that really push the limit of higher-end hardware. There's also the hate towards OpenGL's legacy functions. The Vulkan code snippets we've seen so far show it to be much simpler.

please, notify me if the Vulkan (or Mantle) API is reaching the 50.000th game as the main API, and the api will run on 99,5% on the computers, 99,9% from starting computers sold from 2000.

Again: not talking about the 10.000 Zelda clones in the Google Play store. This API is for higher-demanding games. And there are certainly not 50.000 of those.

users rarely actually playing with the api settings, and will rarely even actually switch the rendering to the Vulkan. If the Vulkan would be the default API, they would probably use that, but why the Vulkan, if there is OpenGL alreday?

No, but often a manual switch isn't neccsary. The reason Mantle had be so is because it's officially still in beta.

also, it would be nice to see real games made with these engines, however, i didnt seen any.
there is basically no real games exist on these engines, just some BloodSteve14's WASD mapflyers and pleasegivemework level designers showcases.
this is far from gaming like the Moon-Earth distance. no player (does not matter that casual, or serious gamer) will EVER use anything made with these kind of engines in reality.
but this is an offtopic.

You haven't seen any games made in Unity of the Unreal Engine? Are you actually playing any high-end games?
If there really weren't any games shipping with them, Epic Games and Unity Technologies would've been bankrupt already.

its nice to see 2017 predictions for some of the games on the list, or basically just ,,no information'' for most of it imagine, how many of these will default to the Vulkan API. Probably 0.

Hard to say now. Even then, they might not need to default, just offer an alternative renderer.

compatibility is terribly, and vulkan will have the same compatibility problems. as i alreday mentioned, its like a male opengl creature raped a famel amd gpu running and opencl in narnia, and the little child is become a Vulkan, emitting lava silently, far away from any civilization.

I do not see how colorful metaphors help you communicate your viewpoint in any way.

well, if you feel that you should immediately go, and develop a vulkan game/engine/whatever, i will not stop you.

only the vulkan itself will stop you: its so broken that they wasnt even had the balls to release the sdk and the specifications publicly yet

They officially announced the API yesterday. The presentation is the 5th of march. Be reasonable.

it is so broken that they want the journalists first to write they confusing brainwashing on gamer sites, they are afraid if the real-world developers actually try it so early, they will early reveal, how bad is it in every mentionable ways.

Bold statements with no base or proof of any kind. It seems unlikely to me that Valve ported Dota 2 to a broken API. That would be too much of a gamble for them.

actually, imgtec and its allies DELETING all technical comments (!) caliming that vulkan is does not makes any step forward, or not satisfyed with the new api.

Where? Their blog post is still up.

this will be one of the biggest disaster in the history of API's, and will succumb everybody who invests in this technology.


its built on the same principles.



oh, and since when these are programming/game development companies?

and who asked them to do this api? becouse the developers isnt. who asked them, to do an api based on these principles? becouse the developers isnt. who asked them, to try releasing a new api together? becouse the developers isnt. and, finally who cares, what they want? becouse the developers isnt.

(also, imagination is just a little laughable insect in this list without the ability of further development.)

That list was a response to your comment: "corporations without the ability to make real cpu-s".

i, as a developer, personally removed every piece of gpu/thirdparty api dependent code from my works since 2013.
on android, however, i still need sdl to paint on the screen, becouse i was lazy to learn how to do it properly.
but, in reality, there is no need for such craps. cpu have enough power now to do anything that we previously did with igp-s/gpu-s.

the era of graphics acceleration ended.

Well, if you can do without GPU acceleration, Vulkan certainly isn't a solution for your product. However, there are tonnes of high-fidelity games out there that do squeeze the last bit of power out of your system.


You clearly are not a "hardcore" gamer, and so Vulkan is not for you. But that also means you'll never make accurate predictions about the industry.

[clementl]

That's actually false.

GLSL is the programming language, which gets translated to SPIR-V. OpenCL kernels also get translated to SPIR-V (while the non-kernel OpenCL code does not).

Vulkan is an API. It stays in C, and doesn't get translated to anything.


The difference is that shader code running on the GPU = SPIR-V. Driver code running on the CPU = Vulkan.


Yeah, that bit is correct.

Oh yes indeed, my bad. Got kinda caught up looking for an analogy.

[ahlaht]

Question

Is SPIR-V like some kind of shared, cross-platform binary BLOB interface?
How likely can it be used to take total control over the computer?

Still, Vulkan seems like great progress. I used to do OpenGL 1.2 long ago. I'm now glad I never learned newer versions. I suspect a minimalistic userspace library implemented on top of Vulkan would do everything I want.

[clementl]

Is SPIR-V like some kind of shared, cross-platform binary BLOB interface?
How likely can it be used to take total control over the computer?

Still, Vulkan seems like great progress. I used to do OpenGL 1.2 long ago. I'm now glad I never learned newer versions. I suspect a minimalistic userspace library implemented on top of Vulkan would do everything I want.

You might still need Newer OpenGL knowledge if you ever try out WebGL.

[Geri]

Looking around in the steam store, most of the more graphically demanding games require OpenGL 3.3 or more.

almost all users of steam is a hardcore gamer. this makes this to be an insignificant statistical fact.

You call that a small number? Assassin's Creed belongs to the better selling AAA games of this generation. Same counts for Battlefield, Call of Duty, Civilization and a whole lot of other games in there.
We're also going a bit offtopic: we're talking about the demand for a high-performance API. Looking at the wiki page you posted, it's obvious a lot of more graphically demanding games seem to be utilizing newer API's

this is like telling that Estonia is a black and asian contry, becouse there is 500-1000 black, and 2500 asian, or so...

actually i would say above 50,000 game, as i alreday mentioned, that it is a serious number. these numbers are just nothing.

Work nonetheless. And now the other way around: how much work is it to port a PC game to mobile devices? A ton when you're making use of complex engine's.

how vulkan will change on this? with that, you will have to port on vulkan api too. making it in vulkan still makes much less sense than originally making it to be prepared for opengles porting.

And this is where you seem to be missing the point. OpenGL will still be around. Vulkan isn't targeting low demanding games. Those don't need thorough optimizations.
Vulkan is for Battlefield 5, COD 14, The Elders Scrolls 6 etc. Graphically demanding games.

me too assume that vulkan is mostly for high demanding softwares and games, and i seriously dubt that this is a viable conception for a graphics api.

77.5% of Steam users has a DX11 capable GPU. That's 77.5% of 125.000.000 users. Since every DX11 capable GPU out there is OpenGL ES 3.1 compatible, it means those 96.875.000 users have an Vulkan capable GPU.

steam users are mostly AAA gamers. of course, they need a decent graphics card, to play crysis 42. today if you buy a computer, you will find a directx11 compatible graphics card in it. as you say, these are facts. the problem is that if somebody got a computer 10 years ago, he is not necessary changing it, becouse maybe he does not even need a more powerfull computer. maybe currently there is a 10-20% of vulkan capable graphics chips on the market in total. thats too few. not speaking that there is no drivers for vulkan yet at all. maybe another 10 year, and we reach that 70-80% of users have a vulkan capable graphics chips. assuming that these chips are also dx11 and opengl3 (es?wtf, you quoted a desktop statistics), but still, you can see, that opengl3 and dx11 games are basically nowhere, just precisely on this small segment called aaa gaming. these apis are unable to walk out from this segments, limitated into this very tiny market segment that probably produces 1-2 percent, or even less of the total gaming industry.

Some certainly do. And for those there's Vulkan. There will always be a demand for high performance 3d-renderering. If virtual reality becomes a succes even more so.

and if you want these magic apis from narnia, you still can use directx11, opengl 3/4, opencl, and the other theocretical worldbeater apis, its not need for n+1 world saver solution for this.

Nice that not multithreading works for your solution, a high demanding graphical application does certainly need them.

of course it needs them, the question is, that it also needs the vulkan-named graphical solution, and the answer is probably negative.

Honestly, the Mantle and Vulkan code I've seen so far is much, much more readable and usable than any fixed-function or OpenGL 3.3 system I've built so far. And no, I'm not going to do 3d rendering on a CPU.

well, i bet that your own rendering code would be much readable, or maybe even faster and more stable than randomly copypasting or inventing a randomly failing shadercrap code to shine the fur on a heroine's leg.

of course its only your decide how you code, i dont want to persuade you from anything, but if i would be you, i would avoid to do even a a line of for vulkan.

Are we starting to throw insults around now? If something I said doesn't equate you should be more concrete.
Also, you're not ruining anything. I just overlooked the fact that OpenGL ES 1.1 also for rid of the fixed-function pipeline. Doesn't help the case you made earlier.

i told that there is ~half million game that needs some versions of opengl. vulcano and mantle is nothing compared to that. you replyed -by trying to collude it with previous, different insigths- that there is the opengl es2 with programmable-only-pipeline.

yeah. and?

they are using on mobiles in notable numbers, nobody told it in different ways. still, the fixed-pipeline is much more popular. i seen some statistics on android, from years before, that calimed that the numbers of opengles1 game seriously outperforms the opengles2 games by 8:1. maybe this is not a trustable source, and i dont have it any more, maybe its significantly changed now, but as we seen on PC, where the primitive opengl applications very brutally outnumber the ,,modern'' opengl solutions, this can be simply believed even when now the number of es2 games are significantly incrased. (also almost all opengles2 applications just displaying textured planes, thats what tutorial the developer found, and thats all folks, modern programmable api, here we come.)

And a lot of developers hate it. With good reason. I'm sorry to say it: DirectX is just much handier. It's shown we can do better, so now we are actually doing something to better it.

so they hating the glBlablablaUniform54245f(bla,bla,bla,bla,bla,bla,b la,&bla,&bla,bla,GL_BLA,GL_BLA,GL_SHIT,GL_CRAP,GL_ GARBAGE,);, but they will not hate the vkBlablablaUniform54245f(bla,bla,bla,bla,bla,bla,b la,&bla,&bla,bla,VK_BLA,VK_BLA,VK_SHIT,VK_CRAP,VK_ GARBAGE); for some magical reason? the two api is crap for the same resons, for the same design flaws.

Not when it comes to high-resource AAA games. Yes when comparing to tiny mobile apps.

those tiny mobile apps are actually much more significant than the pc (or maybe nowdays even the console) AAA games. both by money, both by numbers, both by user base.
back then 15 years ago,the pc aaa was like the 40-50% of the total gaming market. in 2009, it was just around 8%. Nowdays, its probably 1-2%. The mobile and the casual cathegory harvested everything. And there basically there is no need for Vulkan in every ways.

Not for your average 2d sidescroller no. There are higher demanding games out there that really push the limit of higher-end hardware. There's also the hate towards OpenGL's legacy functions. The Vulkan code snippets we've seen so far show it to be much simpler.

you can keep your code in opengl to be extremely simply and minimalistic, if thats your goal. you dont need a new api for just this purpose.

Again: not talking about the 10.000 Zelda clones in the Google Play store. This API is for higher-demanding games. And there are certainly not 50.000 of those.

the problem is that the industry nowdays are ganging up on your 10.000 zelda clone. those are the significant both by revenue, both by numbers, both by popularity. and if they just using a 100 line long rendering engine alltogether in opengl, to display textured images, or even 3d polygons, then this is the mass-demand.

okay, that you calim that vulkan is for aaa games. (it is even totally unviable elsewhere at the moment.) there it can get maybe into maybe 10-100 title within this years. but just for that tiny number of games - even if its aaa - it is worth to maintain a new api? opengl is a mass-api, it covers the whole visualisation, thats why it reached its current popularity. i dont see such changes for vulkan. they will produce that few aaa titles, and thats all. vulkan will become obsolote too within 1-2 years, like everything else, and then it would be just nothing more than an obsolote api that used by a few aaa games that nobody remembers any more. there is no future for such apis.

No, but often a manual switch isn't neccsary. The reason Mantle had be so is because it's officially still in beta.

okay, well, vulkan is officially not even exist yet, we are discussing based on about leaked NDA-ed documents, specifications, and sdk snippets, and puzzling the things together with our own mind. (and mantle was not even public. )

You haven't seen any games made in Unity of the Unreal Engine?

i even attempted to run some. for example, 1.5 year ago, there was a gamedev competition, with 10 game, 3 used unity. all 3 crashed my computer (not only the application crashed, the computer had to be rebooted, mouse and numlock freezed, etc). (radeon 4200 + Athlon2 x4 3,3 ghz)
i also tried to run unity games on mobile, downloaded from google play. none of them worked properly, they stuttered too much (2-3 fps), they crashed at loading, or it had design flaws, so advertises filled the screen and i was not able to click on the play button even. (sony xperia mini pro, snapdragon cpu 1 ghz, 512 mbyte ram, adreno205)

so thats how unity and other miracle-engines ,,work'' nowdays
and thats, what most of AAA titles doing on a non-aaa gamer computer, resulting rapid uninstall, and keeping the person in the real gaming industry.

and by real, i mean not the shaderaddict #45245 fps-type games, instead, the simply, real and innovative games.

i never attempted to run unreal engine-based game, becouse i dont care, so i cant tell anything about that, however, i seen it in videos.

Are you actually playing any high-end games?
If there really weren't any games shipping with them, Epic Games and Unity Technologies would've been bankrupt already.

no. i actually pressed a shift-delete on my game directory in 2005, when i realized, how boring these games are.
from that point of date, both as an user and as a programmer, if we are talking about gaming, i only interested in subcultural casual gaming, and i broke almost every connection with the AAA.

and yeah, most aaa studio is bankrupted, at least in my contry ALL of them is bankrupted. they made crap aniways. except some smart ones that scaled back to the casual market, recognizing that the #45245 fps clone is boring.

Hard to say now. Even then, they might not need to default, just offer an alternative renderer.

reminds me of the swan song of s3 metal

I do not see how colorful metaphors help you communicate your viewpoint in any way.

dont worry, i see it :P

They officially announced the API yesterday. The presentation is the 5th of march. Be reasonable.

well, yeah, then i officially annouce the world's first black hole generator in my garden, the presentation will be in in 2415 march 5. however, the specification is kept in so secret that i will sign nda with everybody who is interested in it, and i will remove every comments from my blog that calims that this solution is not working, becouse i can do it. Be reasonable. Vulcano, the next generation of graphics apis, fuck yeah.

Bold statements with no base or proof of any kind. It seems unlikely to me that Valve ported Dota 2 to a broken API. That would be too much of a gamble for them.

they alreday ported plenty of games on your so-broken opengl api. whats the difference? money talks, dog barks. once they get the money for porting it, they will probably place it in the recycle bin.

Where? Their blog post is still up.

for example, there. i also talked about comments, not from blog posts.

AND yeah, this is one of the blogs where they removed comments. i personally can confirm this.

That list was a response to your comment: "corporations without the ability to make real cpu-s".

intel makes good cpus - but they didnt made any notable propaganda for this api, even if they told that they might be writing drivers for them.

but the most lout propagandists are the responsible for this api, for example, imgtec. they are into selling gpu IP-s, they made a crap weakling mips cpu-platform for theyself. now they are shitting they pants, becouse there is no interest on it. i dont think they have huge future.

Well, if you can do without GPU acceleration, Vulkan certainly isn't a solution for your product. However, there are tonnes of high-fidelity games out there that do squeeze the last bit of power out of your system.

some of them actually having real time ray tracing. and i still run it on cpu, becouse it was better decide to avoid the crap drivers of the graphics chips, and seriously short the development time (months) by not wasting the time on googlins esoteric gpgpu dream-bullshit, and just purely implementing it on cpu. it looks very very ugly, and has small resolution, eats lot ram, etc, however, my previous opengl game engine was ugly too, eated a lot of ram, while offering MUCH less compatibility even if was written it to be able to scale back from opengl 2.1 to 1.0.

You clearly are not a "hardcore" gamer, and so Vulkan is not for you.

correct

But that also means you'll never make accurate predictions about the industry.

this only means that i see much further than gaming.

[clementl]

I don't really feel like writing a whole argument, we're not getting anywhere. Part of the post is even a statement of personal taste or are experiences I can't say anything about. We've also gone completely offtopic. We were talking whether Vulcan will fall into obscurity or not.

And I think this is where the mismatch in our argument lies:

almost all users of steam is a hardcore gamer.

Of course. And Vulkan will certainly not be used in casual games anytime soon.

Will that make Vulkan fall into obscurity? Even the "small" amount of DX11 games didn't hinder development or use of that API, did it?

for example, there. i also talked about comments, not from blog posts.

AND yeah, this is one of the blogs where they removed comments. i personally can confirm this.

Ahh, you mean they are removing very negative comments? Not surprising, I think most company websites with a comment section probably do that.

intel makes good cpus - but they didnt made any notable propaganda for this api, even if they told that they might be writing drivers for them.

I don't know what you view as "propaganda", but this might be something like it. They even expressed interest in Mantle.

this only means that i see much further than gaming.

Not sure what this even means, but Vulkan is an API for demanding 3d applications. A big chunk of that will be "hardcore" games (for which there certainly a market exists).
There's also Pixar in the sponsor list, tough. Seems Vulkan will be used in other sectors too.