Announcement

Collapse
No announcement yet.

A Mesa Branch Provides GLSL IR To TGSI Translator

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

  • #11
    Originally posted by Plombo View Post
    I've just pushed to my branch again. Before, the translator still generated Mesa IR that wasn't ultimately used, but still depended on some information from it. Now, that information has been moved elsewhere in the translator, so I was able to remove the generation of Mesa IR instructions completely.
    I tested it with r300g but it crashes with:
    Code:
    tgsi/tgsi_ureg.c:801:ureg_emit_src: Assertion `src.File != TGSI_FILE_OUTPUT' failed.
    Click here for full backtrace.

    Comment


    • #12
      Originally posted by oibaf View Post
      I tested it with r300g but it crashes with:
      Code:
      tgsi/tgsi_ureg.c:801:ureg_emit_src: Assertion `src.File != TGSI_FILE_OUTPUT' failed.
      Click here for full backtrace.
      That looks like a general issue, not a driver issue. What program are you running?

      Comment


      • #13
        Originally posted by Plombo View Post
        That looks like a general issue, not a driver issue. What program are you running?
        0 A.D. game. It doesn't crash with master branch of standard mesa. Also note that my RV530 only support GLSL 1.20 (dunno if your branch require a GLSL 1.30 compatible card).

        Comment


        • #14
          OK, the crash only appears when compiling your code with --enable-debug . Then 0ad starts fine, however it's missing water (e.g. in Oasis map).

          Comment


          • #15
            Originally posted by oibaf View Post
            0 A.D. game. It doesn't crash with master branch of standard mesa. Also note that my RV530 only support GLSL 1.20 (dunno if your branch require a GLSL 1.30 compatible card).
            Okay, thanks. I think I know what's going on from your error message; it's the absence of a Mesa IR function that was removing reads of output registers. I'm working on it now, and I'll test it with 0 A.D. when I'm done.

            Comment


            • #16
              It should be fixed now by commit f259e14dbff03a79a343. Thanks for reporting it!

              Now to figure out what broke Nexuiz in the same commit that broke 0 A.D.

              Comment


              • #17
                Originally posted by Plombo View Post
                It should be fixed now by commit f259e14dbff03a79a343. Thanks for reporting it!

                Now to figure out what broke Nexuiz in the same commit that broke 0 A.D.
                OK, I confirm the crash is fixed. 0ad water is still not rendered.

                Comment


                • #18
                  Wow, I just spent several hours of my life tracking down the issue with the Nexuiz title screen and the 0 A.D. water (same issue), and the patch that finally fixed it is a whopping 13 lines long.

                  Anyway, both are now fixed by a14d10973e77895c9c4f.

                  Comment


                  • #19
                    Originally posted by Plombo View Post
                    Wow, I just spent several hours of my life tracking down the issue with the Nexuiz title screen ...
                    <OT>
                    Speaking of which, it's xonotic test weekend: http://forums.xonotic.org/showthread...1750&pid=23895
                    </OT>

                    Comment


                    • #20
                      @Plombo

                      Which apps/games look differently with your special mesa branch? How often do you update from mesa master? Should games run faster?

                      Comment

                      Working...
                      X