Announcement

Collapse
No announcement yet.

Mesa's On-Disk Shader Cache Revised For The 7th Time

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

  • #11
    Originally posted by starshipeleven View Post
    Reiser4 = they didn't want to fix reported issues
    I think it was more of a design difference, where kernel devs saw stuff in r4 that should belong in the vfs and the r4 guys disagreed for reasons I forgot, than just *issues*.

    Originally posted by starshipeleven View Post
    DAL = huge code drop from hell (windows)
    DAL didn't originate from the Windows stack.

    Comment


    • #12
      Originally posted by geearf View Post
      I think it was more of a design difference, where kernel devs saw stuff in r4 that should belong in the vfs and the r4 guys disagreed for reasons I forgot, than just *issues*.
      I remember there were major issues about code maintainability too, someone complained about the usage of deprecated functions or using kernel interfaces wrong or something, and the fact that the filesystem was a huge code drop and not an incremental change (i.e. pain to review).

      Then you add the fact that Reiser acted like an ass when receiving feedback.

      DAL didn't originate from the Windows stack.
      It comes from Catalyst or fxgl or whatever the closed driver's name was, which is mostly a port of the Windows driver (like NVIDIA's).

      Comment


      • #13
        Originally posted by starshipeleven View Post
        It comes from Catalyst or fxgl or whatever the closed driver's name was, which is mostly a port of the Windows driver (like NVIDIA's).
        I don't believe DAL comes from Catalyst or anything, I think it was wrote from scratch, but by non a experienced Linux guy/team.

        Here's a quote from bridgman:

        Originally posted by bridgman View Post
        That isn't what happened here (the DAL code is relatively new, not pulled from Catalyst) but we could do it if it made sense.

        Comment


        • #14
          Originally posted by geearf View Post
          edit: "Written" and "a non".
          With no edit I really need to re-read myself before posting

          Comment


          • #15
            DAL3 (which is what we are running in the amdgpu driver) was written largely from scratch for compatibility with the Linux kernel. It also needs to support other OSes as well so the interfaces can't change all that much relative to previous DAL, but the code is substantially new.

            The slides from Harry's XDC presentation have more info, see slides 6-8:

            Test signature

            Comment


            • #16
              Originally posted by bridgman View Post
              DAL3 (which is what we are running in the amdgpu driver) was written largely from scratch for compatibility with the Linux kernel. It also needs to support other OSes as well so the interfaces can't change all that much relative to previous DAL, but the code is substantially new.

              The slides from Harry's XDC presentation have more info, see slides 6-8:

              https://www.x.org/wiki/Events/XDC201...am/amd_dal.pdf
              welp, I got confused with DAL2 then. thanks.

              Comment


              • #17
                Any update on this? I've switched over to mesa git with an RX 480. Games such as Rocket League (unreal engine) pause for 5-20 seconds when switching through basic menus. Once the game loads up it is rock solid though.

                Still much better than amdgpu-pro overall.

                Comment

                Working...
                X