Announcement

Collapse
No announcement yet.

Radeon VRAM Optimizations Coming, But Help Is Needed

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

  • #81
    (can't edit) which of them were ok?

    Comment


    • #82
      Originally posted by curaga View Post
      If you refer to the libudev crash in current master, you can build mesa from my tree (the stats7 or stats-rebased branch). They haven't yet been synced to include the commit causing that breakage, so if you build from there, you should be able to run Steam games fine.
      I did use that branch.

      Originally posted by curaga View Post
      (can't edit) which of them were ok?
      In Rust the moon was missing and in Postal 2 people and some things inside of a building where completely black. The rest was fine.

      Here a short description how I did the porting:
      1) clone the git repository 2 times, one for master, one for stat7.
      2) diff the two.
      3) Manually apply the diff to mesa 10.0.2 - Some functions where in other files and a radeonsi file was missing (I just didn't apply the diff for that radeonsi file as I'm using r600g anyway).
      4) Compile.

      I just tested the two games without R600_DEBUG=bostats and the rendering issues where gone. You don't have any idea what this could be, do you?

      Comment


      • #83
        0 A.D.

        Maybe you already have this, but it would be nice if someone can submit trace for 0 A.D., since it was a game affected by a performance regression I reported some time ago here.

        Comment


        • #84
          @oibaf

          No, no 0ad trace yet.

          @TAXI

          If you can consistently reproduce the issues by setting the env var, then it's either my code or your port that's buggy.

          Comment


          • #86
            hey,

            some minutes of Kerbal Space Program (but demo version): https://anonfiles.com/file/c744a946e...79422083f2a4f9
            debian unstable system, kernel 3.13, xorg/mesa/radeonsi stack completely from git, radeon 7950.

            Comment


            • #87
              Thanks mmstick.

              Fragmentation simulation results are up, analysis will take a couple days.

              Comment


              • #88
                Originally posted by curaga View Post
                If you can consistently reproduce the issues by setting the env var, then it's either my code or your port that's buggy.
                Yes, it's reproduceable.

                Please note that I'm not a developer (at least not a C developer) so the code is way over my head.

                Comment


                • #89
                  Cogs is another example of very good memory behavior. It's obviously a custom engine, and following best GL practises to boot.

                  Comment


                  • #90
                    Originally posted by curaga View Post
                    Cogs is another example of very good memory behavior. It's obviously a custom engine, and following best GL practises to boot.
                    Do you mean the puzzle game which is one of the worst GL apps ever written?

                    Comment

                    Working...
                    X