Announcement

Collapse
No announcement yet.

Intel Drops Mode-Setting Rework Patch Bomb

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

  • #21
    Originally posted by Gusar View Post
    Intel switching requires work that is much better spent bringing new stuff to Mesa.
    Sounds like a valid point.

    Even IF intel comes to the conclusion that they were better off with Gallium now (state trackers, in particular OpenCL),
    this might just be no option (anymore). Departing from the classic mesa approach seems like rewriting large parts
    of the driver from scratch. Read highly unlikely.

    Btw, does anybody know why the intel team opted against Gallium?

    Comment


    • #22
      Originally posted by 89c51 View Post
      Wasn't Michael going to publish some Intel G3D vs Intel Classic benchmarks??
      They'd be pretty pointless. There's a pretty good Gallium i915 driver (mainly as it runs vertex shaders through llvmpipe as the hardware doesn't support them), where the classic driver is fairly poor and certainly hasn't received any attention in the last couple of years or more. As for i965, it's very well supported under classic Mesa but doesn't even have a Gallium driver: there was a rough draft of one, but it was neglected and eventually removed.

      Comment


      • #23
        Originally posted by Gusar View Post
        So the small work the radeon and nouveau teams need to do after the Intel guys have added functionality to Mesa core is too much for you, but the huge amount of work Intel would require to switch their driver to gallium is fine?

        I can turn your post directly against you and say: Intel switching requires work that is much better spent bringing new stuff to Mesa.
        Your argument is a bit weak since we know for a fact that they rewrote their driver recently and had the chance to jump onto the G3D train but chose not to do so. They have their reasons and i am fine with that. However i would prefer not having everybody do it their own way.

        Comment


        • #24
          Originally posted by daniels View Post
          They'd be pretty pointless. There's a pretty good Gallium i915 driver (mainly as it runs vertex shaders through llvmpipe as the hardware doesn't support them), where the classic driver is fairly poor and certainly hasn't received any attention in the last couple of years or more. As for i965, it's very well supported under classic Mesa but doesn't even have a Gallium driver: there was a rough draft of one, but it was neglected and eventually removed.
          Do we know if google -which was behind the 915 driver AFAIK- will develop a G3D for their next generation of chromebooks?

          Comment


          • #25
            Originally posted by 89c51 View Post
            Your argument is a bit weak since we know for a fact that they rewrote their driver recently
            I'd say it's the opposite, it gives my argument strength. They just finished a big overhaul, so why immediately start another? The benefits would need to be huge. I don't think Gallium provides them. Their driver supports the highest GL version Mesa can provide. It supports video decoding using their native hardware decoder as opposed to using shaders. So there was simply nothing to justify another overhaul.

            Comment


            • #26
              Originally posted by 89c51 View Post
              Do we know if google -which was behind the 915 driver AFAIK- will develop a G3D for their next generation of chromebooks?
              Why would they need to? For i915 the classic driver didn't support enough functionality, but for Ivy Bridge the default driver already does.

              Comment


              • #27
                Originally posted by Gusar View Post
                I'd say it's the opposite, it gives my argument strength. They just finished a big overhaul, so why immediately start another? The benefits would need to be huge. I don't think Gallium provides them. Their driver supports the highest GL version Mesa can provide. It supports video decoding using their native hardware decoder as opposed to using shaders. So there was simply nothing to justify another overhaul.
                This mode-setting overhaul was in the kernel part of Intel. It's not related to Mesa/Gallium.

                Comment


                • #28
                  Massive Progress in short time.

                  Last I checked Intel wasn't getting paid to hack on Linux, the sense of entitlement is pretty astounding. Consider Intel could just Nvidia up and ignore Linux. They're doing pretty well. The fact is Nouveau/Intel/ATi are well on track to support modern graphics APIs in a timely manner. We're starting to see Linux finally get a fighting shot against other platforms with regards to support. I predict in 6 months Linux will have better open source GPU support than Apple do for OpenGL. I also predict within 3 years the open source and closed source drivers will hit parity, eg OpenGL 5 on Linux will be available for both open and closed cards. Gallium3d unifying Nvidia/Ati efforts is enough to make this happen. Intel's contributions to making an easy implementation of OpenGL on Linux are helpful. Rather than having to clean room reverse engineer OpenGL they can just take the intel code and refactor it. It saves a crap ton of time.

                  Comment


                  • #29
                    Originally posted by AlbertP View Post
                    This mode-setting overhaul was in the kernel part of Intel. It's not related to Mesa/Gallium.
                    We're not talking about the modesetting overhaul. There was one in the userspace part too, I think it was to bring in glsl support.

                    Comment


                    • #30
                      Originally posted by timofonic View Post
                      Intel Open Source Team isn?t so OPEN...

                      You are all putting the work on your codebase so others can't benefit of it.
                      So once again it came to gallium vs non-gallium battle... I kinda don't have much more to say about it that I haven't already said before - we do not intend to work on Gallium because it does not seems the best way to write 3d driver to us. And I honestly don't think there is any advantage of throwing away the i965 driver that works, just for the sake of rewriting it from scratch with gallium backend.

                      However - all the code is open, all the documentation on how the hardware works is already available up to the Ivy Bridge generation GPUs, so nothing prevents someone from simple writing a i965g driver again. Please, feel free to do so - this is all open-source!

                      Comment

                      Working...
                      X