Announcement

Collapse
No announcement yet.

AMD's Mantle Graphics API For Linux?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Mantle isn't an OpenGL alternative.

    Mantle will be a set of frameworks to encapsulate specific parts of their Hardware via API calls, while in conjunction to interfacing with OpenGL and OpenCL. Are people truly that dense to think it's a replacement? It'll be an API that supports DirectX, OpenGL, OpenCL, not to mention other stuff by accessing the most common APIs without having to directly write strictly in each.

    Sorry, but AMD did a truly piss poor job discussing Mantle. I took it as a sound set of APIs to access assembly calls in C/C++ [most likely C++].

    AMD is fully committed to OpenGL and OpenCL moving forward. This Mantle opens up access to hardware level optimizations.
    Last edited by Marc Driftmeyer; 09-27-2013, 02:03 PM.

    Comment


    • #32
      Originally posted by Marc Driftmeyer View Post
      Mantle isn't an OpenGL alternative.

      Mantle will be a set of frameworks to encapsulate access to OpenGL and OpenCL. Are people truly that dense to think it's a replacement?
      So you're saying that Mantle is an API into OpenGL?

      You realize Mantle is a low-level console API, right?

      Comment


      • #33
        Originally posted by chris200x9 View Post
        Not saying it was clearly a shot against Valve's ambitions I'm just saying it's a bit ironic that a user of a system with ~2% market share is saying 15% market share is no big deal.



        No tell me about the work involved, I'm genuinely curious. I wouldn't think going from one api to the same api on another platform would be that difficult, wouldn't it just me like going from XNA to iOS game using monogame?
        Porting to XB1 to Mantle (easy)
        Now you have moved from XB1 which is a fixed stat platform to the PC which has different processors, RAM, graphics, drivers, etc. I would think because of the fixed platform, that API code wouldn't be very robust as using an API for PCs. Lots of latent bugs will be found just from the different clock speeds and shader counts. To put this in perspective, most large Java programs I have encountered don't run without issue on Linux when it was developed for Windows. In the same way on Android, you can test your code on 8 devices with different hardware and OS versions and everything runs perfectly, release it and have then have a 25% failure rate total. One of the reasons Rage isn't ported to Linux or other OpenGL games.
        Testing, you pretty much have double the scope of testing since you are supporting two APIs in Windows. Otherwise, you are a screwing your consumers and why id Software has traditionally said Linux is unsupported when they release binaries.

        Again it will come down to performance improvements on Windows. I hope it does 25% better, AMD has a Linux version in a year, and I can play frostbite 3 titles on Linux.

        Comment


        • #34
          Originally posted by johnc View Post
          So you're saying that Mantle is an API into OpenGL?

          You realize Mantle is a low-level console API, right?
          You have two low level APIs. OpenCL and OpenGL.

          You create an API that leverages the most commonly used areas with specific calls to AMD hardware. You have a unified GCN architecture from the console, to the desktop, to the laptop, to the embedded space, to the workstation and server spaces.

          You have a lot of code reuse across all levels.

          You then use:
          • OpenGL 4.x/OpenCL1.2 profiles for the Desktop/Laptop/Workstation/Server
          • OpenGL ES 3.x/OpenCL 1.2 profiles for the Embedded Space/Console Space

          where Mantle reuse doesn't happen.

          AMD is 100% in on OpenGL and OpenCL. It's their future. Mantle is a means of encapsulating most of the heavy lifting for you. It's a means of accessing AMD customizations where possible and keeping your `agnostic' OpenGL/OpenCL code separate.

          The lack of information is a means for AMD to create `buzz' by the sheer amount of speculation on what it is or is not.

          Comment


          • #35
            Originally posted by Marc Driftmeyer View Post
            You have two low level APIs. OpenCL and OpenGL.

            You create an API that leverages the most commonly used areas with specific calls to AMD hardware. You have a unified GCN architecture from the console, to the desktop, to the laptop, to the embedded space, to the workstation and server spaces.

            You have a lot of code reuse across all levels.

            You then use:
            • OpenGL 4.x/OpenCL1.2 profiles for the Desktop/Laptop/Workstation/Server
            • OpenGL ES 3.x/OpenCL 1.2 profiles for the Embedded Space/Console Space

            where Mantle reuse doesn't happen.

            AMD is 100% in on OpenGL and OpenCL. It's their future. Mantle is a means of encapsulating most of the heavy lifting for you. It's a means of accessing AMD customizations where possible and keeping your `agnostic' OpenGL/OpenCL code separate.

            The lack of information is a means for AMD to create `buzz' by the sheer amount of speculation on what it is or is not.

            Mantle is just a damn Buzz Word OpenGL>Mantle the way they got there 5X of DX9 was removing the limits of DX9 as DX9 has a Cap on performance There is no way it's going to be way faster then OpenGL+OpenCL Buzz Buzz Buzz Buzz we can say OpenGL+OpenCL is 5x faster then DX9 as well but from what i can see there is no CPU+GPU's to date that can do this Buzz Buzz Buzz

            Comment


            • #36
              Originally posted by Marc Driftmeyer View Post
              You have two low level APIs. OpenCL and OpenGL.

              You create an API that leverages the most commonly used areas with specific calls to AMD hardware. You have a unified GCN architecture from the console, to the desktop, to the laptop, to the embedded space, to the workstation and server spaces.

              You have a lot of code reuse across all levels.

              You then use:
              • OpenGL 4.x/OpenCL1.2 profiles for the Desktop/Laptop/Workstation/Server
              • OpenGL ES 3.x/OpenCL 1.2 profiles for the Embedded Space/Console Space

              where Mantle reuse doesn't happen.

              AMD is 100% in on OpenGL and OpenCL. It's their future. Mantle is a means of encapsulating most of the heavy lifting for you. It's a means of accessing AMD customizations where possible and keeping your `agnostic' OpenGL/OpenCL code separate.

              The lack of information is a means for AMD to create `buzz' by the sheer amount of speculation on what it is or is not.
              Source...?

              Comment


              • #37
                A glimpse on Mantle docs suggest quite opposite, it looks like developers can work with some kind of virtual machine. I guess best analogy would be llvm (just it's internals closely match GCN architecture).


                [edit: I mixed up things a lil ]
                Last edited by Xeno; 09-27-2013, 06:29 PM.

                Comment


                • #38
                  oh come on people, you all know this is another 3DNow! thing that nobody but AMD will be interested in.

                  Comment


                  • #39
                    Originally posted by Mike Frett View Post
                    oh come on people, you all know this is another 3DNow! thing that nobody but AMD will be interested in.
                    EA is going pretty HAM on this. Frostbite 3 will use it, which means other EA games will also use it. This includes two Bioware titles and three DICE titles... yeah it has plenty of interest already I think.

                    Maybe it will never gain traction beyond EA. But just from looking at AMD's Gaming Evolved wins this past year I'd be surprised if there weren't more coming.

                    I'm still not getting an AMD card because FGLRX is terrible, I'm just saying.

                    Comment


                    • #40
                      Anybody who develops a game for PS4 or XB1 and is using the low-level API is probably going to be using this Mantle (or a slight variant of it), so they're in. Moving their games over to Windows should be somewhat trivial.

                      And whether it ever comes to Linux is still up in the air. All I know is that I don't think Linux anything was ever even mentioned Wednesday. So that's something to chew on.

                      Remember that consoles still lead the whole gaming industry. PC gets sloppy seconds.

                      Comment


                      • #41
                        Originally posted by Marc Driftmeyer View Post
                        AMD is fully committed to OpenGL and OpenCL moving forward.
                        This is just hilarious. AMD has the worst OpenGL drivers (both Windows and Linux) and AMD cares so much about OpenCL that their much-touted TressFX tech was implemented with DirectCompute. Gotta make sure it's Windows only, after all.

                        And when it came time to announce all those wonderful new AMD tablets, how many were fitted with OpenGL ES and Android? About zero. And how many with DirectX and Windows 8? About a whole bunch. That's a curious decision -- is Android not a viable path in the tablet market? Something to think about. "Fully committed to OpenGL" indeed.


                        Look, I'm not saying here that I haven't seen their slides and fancy talking points. All I'm pointing out is that there's a difference between talk and action. And the action ain't there, and that's what matters.

                        Comment


                        • #42
                          By the answers, great majority here at phoronix think AMD is acting in favour of DirectX/Microsoft. So I ask why John Carmack says Microsoft can be hostile to Mantle...? http://www.pcgamer.com/2013/09/27/ca...d-amds-mantle/

                          Comment


                          • #43
                            Originally posted by rudregues View Post
                            By the answers, great majority here at phoronix think AMD is acting in favour of DirectX/Microsoft. So I ask why John Carmack says Microsoft can be hostile to Mantle...? http://www.pcgamer.com/2013/09/27/ca...d-amds-mantle/
                            It's hard to extrapolate his ideas off a couple of short off-the-cuff tweets, but the way I understand what he's saying there is that MS and Sony would not want to see Mantle in a Steam Machine because a SM is, in some ways, a direct competitor in the living room. And a Steam Machine with both high-end hardware and Mantle would crush the PS4 and XB1. So they very well might use their influence to nip that idea in the bud.

                            And let's be honest here, in this relationship they have got a LOT of influence over AMD.

                            Let's not lose sight of the fact that AMD's financial troubles are to the point that they're selling their own buildings to raise cash and then turning around and renting them. AMD is in no position to be pissing off MS.

                            Comment


                            • #44
                              No tell me about the work involved, I'm genuinely curious. I wouldn't think going from one api to the same api on another platform would be that difficult, wouldn't it just me like going from XNA to iOS game using monogame?
                              I tried to imply the opposite, actually. Using the same API on different platform isn't that troublesome. Once you have a working graphic context, OpenGL is platform neutral, which means almost all of the code is still valid. The only platform specific code I can remember aside from context creation (which would probably be just a big function, with a matching context destruction one) is the buffer swapping code. Even easier, SDL can handle the differences between platforms for context creation and I believe for buffer swapping, too.
                              Supporting different APIs is almost like rewriting a big portion of the engine, except for the fact you could use the same design and interfaces, so that process can be skipped, but graphics is still a lot of the code, and you need to rewrite it all to support several APIs.
                              There is another part that need to be handled when working on support for different OSes, mostly in input and audio handling, but I believe those as big a big deal.
                              Consider then, writing some context creation and a few more things to support an extra platform (extra buyers), to write from scratch the renderer to get what? Better performance on 15% of the same amount of users buying your product than before you invest your time in such a port? I don't think anyone will avoid buying the game just because it's in OpenGL or DirectX instead of Mantle, so I don't see a developer writing things for several APIs if it doesn't enable new platforms, and writing only for Mantle would be probably crippling their own market. Of course, IFF consoles use Mantle (they need to use special, low level APIs, mostly because they make sense as they can squeeze all of the advantages having a fixed platform gives them), then using it BOTH enables support for the said consoles and for better performance on PCs supporting the API.

                              Please note, I'm basing my ideas on my work in a pretty much obsolete engine (a Wolfenstein engine I'm modifying as a hobby project, it supports Windows and almost everything that uses X11 and is neatly organized by its author so I can check how much shared code there is), so the proportion of code used in every area could be sizable different from what I believe it is.

                              Comment


                              • #45
                                Originally posted by entropy View Post
                                Funny, I was thinking roughly the same:
                                Wondering why he has not showed up so far.

                                Really, I'd be interested in what he thinks about it.
                                Short version: neither me nor our chief graphics dev even heard about it until this week. Could be cool, could end up another historical artefact, no idea.

                                Comment

                                Working...
                                X