Announcement

Collapse
No announcement yet.

Blumenkrantz Picks His Next Battle: Mesa's DRI Interfaces

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

  • #11
    Third and last observation. Has AMD become the "Boeing" of the chip world ?

    Comment


    • #12
      Originally posted by Jumbotron View Post
      Curious that Google feels compelled to use DRI with AMD based Chromebooks at least internally. Why ? Does this dovetail into what was discussed by the creator of Tinybox Neural Net devices to the crappy state of not only AMD drivers and firmware but ROCm as well ? What's going on at AMD and why are multiple vendors using AMD tech being hamstrung like this ?
      Minigbm is generally a somewhat strange piece of software. From experience I can say that the AMD implementation there got some stuff right that others didn't, such as certain DRM modifier handling. And IIRC that was precisely because they did poke into the Mesa internals, while other implementations like the Intel one didn't and therefore had to re-implement certain stuff (which implies: needing to keep up to date with Mesa driver features).

      Comment


      • #13
        I can't wait for the code purge + deep refactor when only wayland, vulkan and zink will be needed in the main release!

        Comment


        • #14
          Originally posted by Hazel View Post
          We need more Blumenkrantz and less avis.
          I've had a few heated disagreements with him in the past.
          I've also encountered him in various repos helping devs and projects out by reporting issues with detailed reproduction steps and background details and going back and forth trying various fixes out for the devs. And I have a ton of respect for the latter. Not saying one thing makes up for the other, but I just want to point that out.

          If he has his future disagreements and arguments closer to room temperature than in the past, I really don't see an issue moving on from the old days. It's what Linus did.
          Last edited by Mitch; 13 March 2024, 07:40 PM.

          Comment


          • #15
            Originally posted by rmfx View Post
            I can't wait for the code purge + deep refactor when only wayland, vulkan and zink will be needed in the main release!
            This !

            Comment


            • #16
              What happens to those who want to play games on a *BSD like FreeBSD or OpenBSD using X11?

              Comment


              • #17
                Originally posted by Jumbotron View Post
                Slightly off topic but still I think in the wheel house of issues with AMD, it's been confirmed by them that VCN 4 is broken at the hardware level. This was discovered with the new release of OBS as certain encoded video on AMD hardware with VCN 4 adds padding to the top and bottom because it can't conform to standard pixel sizing. It was also confirmed that it will NOT be fixed until VCN 5 arrived....in brand new hardware. The only thing that AMD can do is when presented with say a standard HD sizing regime is to basically keep the black padding to a minimum when the encoder goes outside the normal sizing of say a 1080p sized video.
                Can you give some source for VCN 4 problems. I cannot find this info on a rapid search.
                Thanks.

                Comment


                • #18
                  Originally posted by Hazel View Post
                  We need more Blumenkrantz and less avis.
                  I, for one, welcome our avian overlord(s).

                  Comment


                  • #19
                    Originally posted by Bisu View Post

                    Can you give some source for VCN 4 problems. I cannot find this info on a rapid search.
                    Thanks.
                    Sure. It’s actually embedded two stories down from this one on the Phoronix homesite. Look for the first story under the date stamp March 12 entitled “ OBS RELEASED WITH AV1 SUPPORT “. Click that article and look in the comments for the poster “ shmerl “. He’s the first to leave a comment. Click on the link he (she) embedded in their reply in the bug with AMD VCN 4 . That will take you to Freedesktop.org and an issue page for Mesa. It concerns VAAPI and encoding AV1 video at a prescribed size but AMD cards with VCN 4 can’t get it right because it’s broken at the hardware level so it introduces a black pad top and bottom. Read the whole thread and you will eventually get to a poster named “ Ruijing Dong “. He’s the one who states AMD VCN 4 is broken in hardware and that he confirmed with “our HW/FW” team that VCN 4 is broken and won’t be fixed until the release of VCN 5 hardware.

                    Comment


                    • #20
                      Originally posted by Jumbotron View Post
                      Third and last observation. Has AMD become the "Boeing" of the chip world ?
                      Once whistleblowers start dying, maybe.

                      Comment

                      Working...
                      X