Announcement

Collapse
No announcement yet.

Developers Still Hoping For AMD DAL Support In Linux 4.7

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

  • Developers Still Hoping For AMD DAL Support In Linux 4.7

    Phoronix: Developers Still Hoping For AMD DAL Support In Linux 4.7

    Open-source developers working on the Radeon Linux graphics driver stack remain hopeful that their massive "DAL" code-base will be ready for merging with Linux 4.7...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    DAL will help my notebook with A10 carrizo + r7 m360 to have lower temperature with AMDGPU driver?
    catalyst reach max 55° or 60° for some seconds, instead AMDGPU reach also 70° and with the minimum effort, temperatures reach 60+°.
    i hope this problem will be fixed

    EDIT: i've tried kernel 4.4 kernel 4.5 and kernel 4.6
    Last edited by Dea1993; 30 April 2016, 01:42 PM.

    Comment


    • #3
      It's really nice to see that both Dave and Jerome are trying to help to get the code ready for merging. And by the way, I never thought AMD had that many devs working on the driver.

      Comment


      • #4
        Originally posted by Dea1993 View Post
        DAL will help my notebook with A10 carrizo + r7 m360 to have lower temperature with AMDGPU driver?
        catalyst reach max 55° or 60° for some seconds, instead AMDGPU reach also 70° and with the minimum effort, temperatures reach 60+°.
        i hope this problem will be fixed

        EDIT: i've tried kernel 4.4 kernel 4.5 and kernel 4.6
        I only got sad performance from AMDGPU driver with my carrizo laptop (the same apu + gpu configurations of yours).
        I tested kernels 4.4, 4.5 and 4.6rc5. Using the latest 4.6 kernel, aside from 2d/3d performance, when I disable the discrete gpu using vga_switcheroo I get kernel panics at shutdown/reboot.

        The most annoying behavior is the screen backlight continously going off and on when the screen is in off mode, actually increasing temperatures and battery drain. It looks like the gpu is continously resetting.
        Also sometimes all of a sudden the display goes black and doesn't come back, so I have to reset the machine.

        I had to go back to legacy catalysts and stay stick to kernel 4.2 because the user experience with amdgpu driver on my laptop is awful

        Comment


        • #5
          As much as I would like to have open source support for Polaris I think the chances are slim. DAL is a massive code drop that has been very heavily criticized by DRI maintainers for being questionable quality and, more importantly, not using common DRI code.
          The most positive sounding comment mentioned that it maybe could be staged for 4.7 and from there used as a source of truth about hardware and bit by bit rewritten/pulled into mainline.
          Despite the hard work I've seen in the DAL branch (a few HUNDRED commits every week) I'm seriously worried that we have another Android mainlining on our hands..

          Comment


          • #6
            Originally posted by blackshard View Post

            I only got sad performance from AMDGPU driver with my carrizo laptop (the same apu + gpu configurations of yours).
            I tested kernels 4.4, 4.5 and 4.6rc5. Using the latest 4.6 kernel, aside from 2d/3d performance, when I disable the discrete gpu using vga_switcheroo I get kernel panics at shutdown/reboot.

            The most annoying behavior is the screen backlight continously going off and on when the screen is in off mode, actually increasing temperatures and battery drain. It looks like the gpu is continously resetting.
            Also sometimes all of a sudden the display goes black and doesn't come back, so I have to reset the machine.

            I had to go back to legacy catalysts and stay stick to kernel 4.2 because the user experience with amdgpu driver on my laptop is awful
            Why do you use vga_switcheroo? That should be an muxless system.

            Comment


            • #7
              Originally posted by blackshard View Post
              The most annoying behavior is the screen backlight continously going off and on when the screen is in off mode,
              Funny enough I have this issue with my AMD R9 390x and a 4k screen, in windows. I gotta turn off my monitor each night or this occurs and is annoying. I recently tested a OLD 4870 ATI card and it didn't exhibit this problem, which leads me to believe the current crop of AMD drivers have a powerstate problem (at least for windows, I haven't tested it in AMDGPU/Linux yet).

              Comment


              • #8
                and is a lot of code opened up from the Catalyst proprietary driver.
                I may be mistaken, but didn't even the devs state that this code was not ripped from fglrx/Cat./Crimson/whatever?
                Stop TCPA, stupid software patents and corrupt politicians!

                Comment


                • #9
                  Originally posted by c2h5oh View Post
                  DAL is a massive code drop that has been very heavily criticized by DRI maintainers for being questionable quality and, more importantly, not using common DRI code.
                  Of course it is... that's what happens when you take a big blob of closed code, and make it open. You almost always start with something in pretty bad shape, which doesn't do things in any kind of standard way and duplicates a ton of stuff for whatever internal reasons. The question is simply, what happens next? The fact that AMD has developers actively working on getting this into the upstream kernel suggests a willingness to make the necessary changes - it mostly comes down to how much those improvements have to be done before it can be accepted, and how much can be done over time afterwards.

                  Comment


                  • #10
                    I see a lot of "kill off x/y/z" commits in the log there, any many of them are by airlied/glisse. Looks like pretty good progress to me.

                    Comment

                    Working...
                    X