Announcement

Collapse
No announcement yet.

The R300 GLSL Compiler Improvements Are Coming

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

  • #31
    (in my best Mr. Burns voice)

    EX-cellent.

    Carry on

    Comment


    • #32
      Actually, celestia is still slow with the most basic render path

      Ctrl-V immediately after starting celestia will send you to the "Basic" render path, and r300g suffers even here with my RV350 AGP.

      Comment


      • #33
        I've tested Celestia now and it is slower even with PCIe but only when I zoom in. How the hell to display framerate there? I cannot investigate without knowing the framerate.

        Comment


        • #34
          Type a [`] (= backslanted apostrophe = accent grave) to enable Celestia's fps (frames per second) display in the lower left corner. If you have a keyboard that supports diacritical marks, then you may need to type a [SPACE] after the [`].
          http://www.shatters.net/forum/viewto...p?f=15&t=14387

          Or you could use something like bugle.

          Comment


          • #35
            To give you some numbers...

            With my "right-drag the Earth while using the Basic render path" test:

            I am getting approx 25 FPS with r300c, but only just over 2 FPS with r300g.

            Comment


            • #36
              Originally posted by chrisr View Post
              I am getting approx 25 FPS with r300c, but only just over 2 FPS with r300g.
              If you set the environment variable RADEON_DEBUG to "notiling", does it change anything?

              Comment


              • #37
                No, "notiling" doesn't do anything

                Originally posted by marek View Post
                If you set the environment variable RADEON_DEBUG to "notiling", does it change anything?
                I'm still getting about 2 FPS unless I zoom so far out that I can't see any detail on the Earth.

                Comment


                • #38
                  So the problem is celestia is submitting lots of vertices using vertex arrays without use of VBOs. The driver spends most of the time in memcpy, copying vertices from a user pointer to a mapped vertex buffer. The copying takes place in u_upload_mgr. Currently I have no idea what to do about it.

                  Comment


                  • #39
                    Does this same copy not happen with classic?

                    Comment


                    • #40
                      More importantly, with fglrx and the NVIDIA blob too?

                      Originally posted by curaga View Post
                      Does this same copy not happen with classic?
                      The NVIDIA blob gives me 60fps with the OpenGL 2.0 render path. I don't see how celestia could manage that if the CPU is being forced to do a lot of memcpy operations. (This is on a 2.5 GHz dual core laptop, vs the dual 2.66 GHz P4 Xeon PC that uses my RV350.)

                      So the problem is celestia is submitting lots of vertices using vertex arrays without use of VBOs.
                      So either celestia is taking a different code path with the NVIDIA blob, or there must be a solution that cannot involve memcpy().

                      Comment

                      Working...
                      X