AMD Mantle API [updating]

Was Mantle ever supposed to be a cross platform API? AFAICT, it was always about bringing a console-style low-level API to the PC. There was never anything saying that it was cross platform, just that using it on the PC is like using the APIs on the consoles that are more direct-to-the-metal.
 
How many different ways are there to make an API that is more efficient at drawcall throughput? Even if Mantle is not compatible with other APIs, I'd expect it to be similar enough to make porting pretty easy. The consequence of that is, of course, that it wouldn't be too hard for other companies that make a similar, Mantle-like API...

Maybe Mantle is about more than just drawcall throughout, but since it works on existing GPUs, it can invent something radically new.
 
silent_guy said:
Maybe Mantle is about more than just drawcall throughout, but since it works on existing GPUs, it can invent something radically new.
Click to expand...
If it's just about draw call throughput then while it might be a good industry forcing function, it's not doing anything that you couldn't do with a GL extension or some improvements to DX (or WDDM). If the intention is for it to provide something compelling long-term it really needs to do something that wouldn't be portable to other architectures at all. I'm actually hoping for the latter personally, as that's a lot more fun for developers like me :) I also don't really care if we have really fast or really big draw calls or both in the long run as long as I have the flexibility to pull increasingly dynamic state on the GPU.

So if I see anything resembling binding slots in Mantle given what we know of the GCN hardware from the ISA docs, I'm going to facepalm ;)
 
Last edited by a moderator:
Andrew Lauritzen said:
If it's just about draw call throughput then while it might be a good industry forcing function, it's not doing anything that you couldn't do with a GL extension or some improvements to DX (or WDDM). If the intention is for it to provide something compelling long-term it really needs to do something that wouldn't be portable to other architectures at all. I'm actually hoping for the latter personally, as that's a lot more fun for developers like me :) I also don't really care if we have really fast or really big draw calls or both in the long run as long as I have the flexibility to pull increasingly dynamic state on the GPU.

So if I see anything resembling binding slots in Mantle given what we know of the GCN hardware from the ISA docs, I'm going to facepalm ;)
Click to expand...

Well we heard about thoses GL instructions since a good long time now ( what game is running under windows with thoses extenstion ? and it is the same for the DX improvement, but maybe Mantle is because they was tired to wait to see them in use ? Everything have done AMD thoses last 10 years have been ported to other architecture available to other, this have not stop them to provide them.. they are included in DX, they are included in OpenGL, on professional softwares.. I dont recognize any gamers who even know from who is coming thoses features ( not that they have invent them, but they permit to bring it in gaming developpement or DX, or OpenGL. ).. AMD who have push OpenCL, could have not support directcompute as they have done, but they have do it, even TressFX is using DirectCompute instead of OpenCL ..

Today we have MS who do a loong speech about DX11.2 memory management, tile culling, Tiled ressource .. lol.. WHo have bring them to DX 11.2 ?

I dont know what will bring, give, to us Mantle, if it will be or not a success... a failure. But i can bet it will bring easely a new direction on how the API for games are developped today.. Maybe thats the only thing we should retain now, and maybe later.
 
Last edited by a moderator:
Andrew Lauritzen said:
If it's just about draw call throughput then while it might be a good industry forcing function, it's not doing anything that you couldn't do with a GL extension or some improvements to DX (or WDDM). If the intention is for it to provide something compelling long-term it really needs to do something that wouldn't be portable to other architectures at all. I'm actually hoping for the latter personally, as that's a lot more fun for developers like me :) I also don't really care if we have really fast or really big draw calls or both in the long run as long as I have the flexibility to pull increasingly dynamic state on the GPU.

So if I see anything resembling binding slots in Mantle given what we know of the GCN hardware from the ISA docs, I'm going to facepalm ;)
Click to expand...
The multitude of ways to do a thing in GL, even the modern one, is just off putting. Let's hope the Mantle API is as good as you hope.

Besides, the DX api is a real plus point for windows vs linux/ios/mac, if you ask me. I wish others would also make low level APIs.
 
lanek said:
Well we heard about thoses GL instructions since a good long time now ( what game is running under windows with thoses extenstion ?
Click to expand...
I must admit I'm having a hard time understanding your post, but I'm not sure you understand me here. I'm not speaking specifically of bindless and other current NVIDIA GL extensions, I'm just talking about the mechanism of exposing the optimized path. If it's at-all portable then there's no need for a separate API - we already have two perfectly good ones. It's only if it's not portable that it gets interesting.

rpg.314 said:
The multitude of ways to do a thing in GL, even the modern one, is just off putting. Let's hope the Mantle API is as good as you hope.
Click to expand...
Hey, I'm the first to say that GL is a bit of a mess, but AMD very well could have just made an "extension" that subsumes the majority of it if they wanted to. To reiterate, my point is just that making your own API really has only two goals I can see... 1) using it to force short term changes to the standard APIs or 2) making it specific to your architecture and not portable. The first is fine, but political and less interesting. The second is way more fun :)

(As an aside, I do really think that people are ignoring too much that a lot of the issues that AMD claims to want to address in Mantle need to be addressed properly in WDDM. That seems more like the "forcing function" approach and that's ultimately a good thing for everyone as well.)

Guess we'll see in November. With Johan heavily involved I have high hopes :)
 
Last edited by a moderator:
Andrew Lauritzen said:
Hey, I'm the first to say that GL is a bit of a mess, but AMD very well could have just made an "extension" that subsumes the majority of it if they wanted to. To reiterate, my point is just that making your own API really has only two goals I can see... 1) using it to force short term changes to the standard APIs or 2) making it specific to your architecture and not portable. The first is fine, but political and less interesting. The second is way more fun :)
Click to expand...

If it is 2), then I am not sure if you could get away with a GL extension. Even GL has serial command dispatch and many other things that one would want gone from low level APIs.

(As an aside, I do really think that people are ignoring too much that a lot of the issues that AMD claims to want to address in Mantle need to be addressed properly in WDDM. That seems more like the "forcing function" approach and that's ultimately a good thing for everyone as well.)
Click to expand...
I think it has more to do with trying to get custom sw written with minimal investment by going with the middleware vendors.
 
xDxD said:
http://www.worldsfactory.net/2013/10/11/ubisoft-amds-mantle-api-is-a-double-edged-sword
Click to expand...

I think AMD's proposal probably comes from the fact that the development environment for ps4 and xbone will be really similar, which in turn should also be really similar to PC's DirectX.
So instead of having 3 environments (PS3+360+ DirectX PC) that were very different from each other in last gen, developers will have instead 4 environments (PS4+XBone+Mantle+DirectX PC) that are very similar among them.
 
I think even if PS4 and XO didn't use something close to Mantle on principle, *any* layer that gets close to the hardware should be similar among the three devices when that hardware is the same.
 
I agree with that (on the basis that it seems logical) but even if it wasn't, is there anything stopping AMD from just extending Mantle so it is closer to both anyway?
 
I wish somebody would just come out and say what Mantle is and what it isn't and clearly describe the benefits. All we have now are vague one liners and ensuing rampant confusion :)
 
You must log in or register to reply here.

Similar threads

S
SDL_Gpu API, SDL 3.0 ... New cross-platform graphics API.
Replies
1
Views
1K
Lurkmass
L
D
DX12, DX12U, DXR, API bias, and API evolution ..
2
Replies
37
Views
6K
Lurkmass
L
R
State of GDeflate *spawn
2
Replies
25
Views
2K
Remij
R
F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
D
Comparative consideration of DirectX 12 in games *spawn
5 6 7
Replies
124
Views
15K
troyan
T
Back
Top