Announcement

Collapse
No announcement yet.

AMD GPU-PRO Hybrid Linux OpenGL Performance vs. RadeonSI Gallium3D

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

  • #71
    Originally posted by geearf View Post
    I don't think they are. First one is about their intent, ie what they wish will happen, yet not a promise that it will actually happen. The 2nd one means the same but using a different wording.
    Sounds like we are all aligned now. IIUC smitty3268 was saying "if you're trying to be specific and precise, you're not succeeding" (which is fair).

    Unfortunately for any project involving open sourcing a big whack of code written to be closed source the reality is more like the request from the Philosophers' Union in the first Hitchhikers Guide book.... "we demand clearly defined areas of ambiguity"... we have committed to work on it, we think it will succeed (although both have their risks & complications), but that's all we can say at the moment.
    Last edited by bridgman; 22 March 2016, 09:51 AM.
    Test signature

    Comment


    • #72
      Originally posted by azari View Post
      which appears to essentially be a hacked-together retroactive ICD system since Mesa historically never provisioned for this back in the 90s.
      mesa can't provision standard for others to use
      Originally posted by azari View Post
      the open source drivers can do the heavy lifting until native support makes it in
      open source drivers are native support

      Comment


      • #73
        Originally posted by azari View Post

        If all future Linux games start going Vulkan, then it's going to be a much easier decision, as the OpenGL stuff won't matter as much.

        Stability shouldn't be an issue anymore now that they're using amdgpu though, since a lot of the issues previously were with the kernel blob. From now on it should be as stable as the open source stack.
        Yes, That is what I'm personally hoping for. Although I have to admit that the moment mesa begins working with the Vulkan stack, I'm switching back.
        Last edited by duby229; 22 March 2016, 01:26 PM.

        Comment


        • #74
          Originally posted by smitty3268 View Post
          I figured you guys were trying to open source this stuff but couldn't definitely 100% commit to it, let alone have a hard date.
          no software project could have a hard date based on features. they could give you hard date by which they will opensource what will be ready by that date. if you need 100% answer, you should contact insurance company. every day some corporations die, what happened to their hard dates?

          Comment


          • #75
            Originally posted by bridgman View Post

            Sounds like we are all aligned now. IIUC smitty3268 was saying "if you're trying to be specific and precise, you're not succeeding" (which is fair).
            I should probably just let this die, but it's the internet so....

            That wasn't my point.

            My point was that i originally had the right idea, but when you replied to me and said something it made me question what i had previously thought. It seemed like you were contradicting me. And now it seems like i was right all along, and you didn't mean to say something that changed my mind.



            ...
            Originally posted by smitty3268 View Post

            I don't believe AMD has ever announced any kind of timeline, just that they would "try" to open source it.

            I'm hoping it will happen this year, too, but based on what bridgman has said I'm not sure they've even 100% promised to do it yet.


            ok - i have things correct here.


            And then you reply
            Originally posted by bridgman View Post
            IIRC we have said that we intend to do it but not given a timeline yet.


            and now this reply to me makes me think i must have gotten something wrong, and you are in fact guaranteeing it. but it seems like that's not what you meant - i guess you were confirming my original post rather than disagreeing with it. Perhaps it's all my fault.


            Last edited by smitty3268; 22 March 2016, 03:23 PM.

            Comment


            • #76
              Originally posted by smitty3268 View Post
              ... but it's the internet so....


              Originally posted by smitty3268 View Post
              That wasn't my point.

              My point was that i originally had the right idea, but when you replied to me and said something it made me question what i had previously thought. It seemed like you were contradicting me. And now it seems like i was right all along, and you didn't mean to say something that changed my mind.

              and now this reply to me makes me think i must have gotten something wrong, and you are in fact guaranteeing it. but it seems like that's not what you meant - i guess you were confirming my original post rather than disagreeing with it. Perhaps it's all my fault.
              I think the source of confusion is that "opening" can refer to either "working on making something open" or "delivering the open-sourced code", and I was using it both ways depending on context. When using the first meaning, I could say "we are definitely going to do it" but when using the second meaning I was adding notes like "if things go well".

              Something like that anyways.
              Test signature

              Comment


              • #77
                i want try this driver, please AMD add soon the capability with A10 8700p (carrizo) and rx 300m (r7 m360), i hope that with this driver i'll get a boost of performance (i've read that in this first release opengl isn't yet optimized)

                Comment


                • #78
                  Originally posted by bridgman View Post

                  The OpenCL, OpenGL and Vulkan drivers are the same as Windows except for the OS interface layer. There will probably be more changes as we progress towards opening up OpenCL and Vulkan, but the general idea is that all of the API code (which is most of the driver) will stay common.



                  It's not officially supported, but (a) a lot of people have asked for it, (b) in the short term it should be pretty hackable if you ignore GL/CL interop.

                  Rather than "hybrid with Mesa for OpenGL" it's more like "all-open with Vulkan and OpenCL" which Alex & Jammy discussed last fall.
                  late response, not been logged in for while, but thanks for you answer. Looks like a sollid idea to have such a deep integration. Sounds simple enough, but probably has a lot of technical challenges to get such an integration. Looking forward to a driver + hardware combo that satisfies my personal needs (due to past bad experiences I'm mostly sitting back right now, but really amd over the competitors as soon there the features/stability are comparable.

                  On a side note, great to see amd being used more on enterprise level, at least in my country I see it used by the biggest organisation in the country.

                  Comment

                  Working...
                  X