Announcement

Collapse
No announcement yet.

Intel Linux Graphics Stack Gets Another Speed Boost - Helping Civilization VI By ~18%

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

  • #21
    Originally posted by tildearrow View Post

    The problem is that Windows uses a completely different driver model. It is not easy to port a solution from Windows to Linux.
    I'm not asking them to port a solution as is. Problems such as power management, and other low level bits have more to do with the hardware. I know there are differences in kernel APIs for different kernels, but they don't prevent AMD dev's from finding out how the driver talks to the hardware (registers and values) to achieve a given goal - the interface or APIs for doing so may be different, but ultimately it's certain registers that must have certain values at certain times or steps, or in certain situations.

    For example, if you look at the article here, Intel devs didn't port over the exact code, they merely examined what the Windows driver was doing, and then re-implemented it in the Linux driver (well here they just removed code).

    That's all I ask AMD to do. If you don't know how to do something, look at your own damn code and find out the correct way to do it.

    Comment


    • #22
      Originally posted by sandy8925 View Post

      That's all I ask AMD to do. If you don't know how to do something, look at your own damn code and find out the correct way to do it.
      For AMD the problems can be proprietary just as much as they could be hardware related. For all we know they are doing just that and having issues changing proprietary based code on Windows to open code on Linux.

      Comment


      • #23
        Originally posted by dungeon

        Now you sound like LGBTQ+ muslim If you are, don't be afraid just say it.



        Cool, If you don't care, i care even less
        From one dumbass to another, you've done gone full retard

        Comment


        • #24
          Originally posted by Aryma View Post
          why are you keep forcing this force Subject down in my throat i dont care i come here to talk about linux not about anything elso
          and Bingo was his nameo

          (I'm sorry, I had to )

          Comment


          • #25
            Thread about Intel Linux Graphics ends up about LGBT and prison and countries.

            What the fuck did I just read?

            Comment


            • #26
              Originally posted by Weasel View Post
              Thread about Intel Linux Graphics ends up about LGBT and prison and countries.

              What the fuck did I just read?
              Yes, it happens often at Phoronix. And in my opinion, it adds so much fun to the forum. We talk about the topic, and then after a few pages the topic has changed.

              Same thing happened with 3 other threads.

              A thread about AMD power management turned into a conversation about climate change.

              A thread about AMDGPU code changes turned into a conversation about QNX, and a few pages later we were talking about car engines.

              And the Huawei thread was transformed into a talk about RISC-V and democracy.




              Also, don't worry, dungeon just likes to transform threads by adding unrelated topics to the thread.

              Comment


              • #27
                Originally posted by geondun

                你讀了你讀的內容

                It was english and not mandarin chinese nor in any else of about 6500 spoken languages, 2000 of languages have fewer than 1000 speakers... and some of these people are in prison, etc...

                Ha, ha, most likely because "what the flock did i just xyz" again
                Lorem ipsum dolor sit amet, something about consequence blah blah blah.




                OK, back to Intel. Is there any reason why when using Valgrind to safety-check my OpenGL application output more errors under Intel than under AMD?
                Is the Intel driver actually more unstable?

                Comment


                • #28
                  Originally posted by tildearrow View Post

                  What about the difficulty of writing such abstraction layer?
                  That's a one-off problem that solves the rest or most of the rest of your issues. After that you'll have a single codebase with single issue tracking and only an occassional OS specific case.

                  It's definitely worth it, but AMD has a track record of keeping 3+ driver variants for even the same OS just for fun. They're extremely bad at drivers.

                  Comment


                  • #29
                    Originally posted by Almindor View Post
                    Doesn't prevent Nvidia from using the same driver on both. All you need is a good abstraction layer.
                    ... and that abstraction layer would not be allowed upstream. Are you saying that we should only produce out-of-tree drivers like NVidia ?

                    Comment


                    • #30
                      Originally posted by tildearrow View Post
                      OK, back to Intel. Is there any reason why when using Valgrind to safety-check my OpenGL application output more errors under Intel than under AMD?
                      Is the Intel driver actually more unstable?
                      In general with graphics drivers valgrind doesn't know about CPU mappings of GPU memory unless you teach it. If you have the valgrind headers on your system when you compile Mesa, the Intel drivers hook into the valgrind API and let it know about these things. With that, it ought to be entirely valgrind clean. Distro mesa typically isn't build with valgrind support though.
                      Free Software Developer .:. Mesa and Xorg
                      Opinions expressed in these forum posts are my own.

                      Comment

                      Working...
                      X