Announcement

Collapse
No announcement yet.

Mesa's Rate Of Git Development Is Slowing

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

  • #11
    As a Developer, Gallium looks far more attractive then MESA. Having one unified standard for N number of API's is frankly every OS should have done a VERY long time ago.

    Comment


    • #12
      Originally posted by uid313 View Post
      Could not find a roadmap.

      But if you want to contribute:
      http://www.mesa3d.org/helpwanted.html
      Join mailing list: http://www.mesa3d.org/lists.html
      Read OpenGL documentation: http://www.opengl.org/documentation/
      it seems like a good and logical thing to start adverting at least this "help wanted" message in public areas where outsiders can actually read and perhaps get incentive to actually do something with it, so reproduced here for convenience.

      for sure this is the first time iv ever realized it even existed.
      http://www.mesa3d.org/helpwanted.html

      "Help Wanted / To-Do List

      We can always use more help with the Mesa project. Here are some specific ideas and areas where help would be appreciated:

      Driver patching and testing. Patches are often posted to the mesa3d-dev mailing list, but aren't immediately checked into git because not enough people are testing them. Just applying patches, testing and reporting back is helpful.
      Driver debugging. There are plenty of open bugs in the bug database.

      Remove aliasing warnings. Enable gcc -Wstrict-aliasing=2 -fstrict-aliasing and track down aliasing issues in the code.
      Windows driver building, testing and maintenance. The Visual Studio project files aren't always updated in a timely manner when new source files are added or old ones are removed. Fixing these tends to delay new Mesa releases.
      Maintenance and testing of lesser-used drivers. Drivers such as DOS/DJGPP, GGI, etc that aren't being maintained are being deprecated starting in Mesa 7.3.
      Contribute more tests to glean.

      Automatic testing. It would be great if someone would set up an automated system for grabbing the latest Mesa code and run tests (such as glean) then report issues to the mailing list.
      If you want to do something new in Mesa, first join the Mesa developer's mailing list. Then post a message to propose what you want to do, just to make sure there's no issues.

      Anyone is welcome to contribute code to the Mesa project. By doing so, it's assumed that you agree to the code's licensing terms.
      Finally:

      Try to write high-quality code that follows the existing style.
      Use uniform indentation, write comments, use meaningful identifiers, etc.
      Test your code thoroughly. Include test programs if appropriate."

      you should probably also add/emphasise routinely benchmark your changes and publicly post your results, write and add to a new and comprehensive test suite to automate testing for correctness, auto bench marking and auto reporting of these results to a single central location these and other changes in the future
      Last edited by popper; 06-25-2012, 03:52 PM.

      Comment


      • #13
        Maintenance and testing of lesser-used drivers. Drivers such as DOS/DJGPP, GGI, etc that aren't being maintained are being deprecated starting in Mesa 7.3.
        So...even if the driver works 100% (which I can't say one way or the other), if there aren't any updates, it gets depriciated?

        Am I the only person who realizes how idiotic that is?

        Also, the fact MESA needs so many drivers for different architectures is why I hope Gallium replaces it. One interface, one backend.

        Comment


        • #14
          Probably devs taking well-deserved vacations, or planning their next moves.

          Comment


          • #15
            Originally posted by WorBlux View Post
            or planning their next moves.
            I would think more than planning. In mainline mesa there seem to be only features that actually work. Maybe for some few days something breaks, but mostly mainline mesa git is extremely well usable. So I think the unstable development is probably in different branches or maybe even local repositories and gets pushed to master as soon as it is ready.

            I have not looked at the article very well, but I think it only counts the commits to master and not to other branches.

            Comment


            • #16
              Originally posted by asdx
              Agreed, we also need patents like ST3C to become invalid and integrate all that stuff into Mesa by default.
              S3TC is owned by HTC now. It's hard to know if they'll aggressively pursue users of this patent. HTC doesn't strike me as a particularly evil company, at least not on the scale of Oracle or Microsoft.

              Originally posted by asdx
              Why is this happening? How can we help with things?
              It's hard to know if the statistics reflect what's actually happening. For example, just ONE of the commits earlier this year pushed over 100,000 lines of code to get the ball rolling for HD7000 support -- which still isn't complete, but last I checked, you can successfully boot up a HD7000 in KMS mode, and use a 2D-only driver or llvmpipe to get working X.

              Originally posted by asdx
              Is Mesa being backed by companies like Canonical
              No.

              Originally posted by asdx
              Red Hat,
              Yes, very heavily. Much of the grunt work of supporting recent AMD GPUs is thanks to Red Hat. They also help out Intel on their driver, they also work on old GPUs and server GPU chipsets for KMS support, and they also help with the overall architecture. Also, one of their employees, Dave Airlie, is the "chief" for the DRM (direct rendering manager) subsystem of the kernel, which basically means that he is one "rank" removed from Linus Torvalds in terms of decision-making power about what does or does not go into the Linux kernel's open source graphics drivers. Dave's task in that role is to take all the DRM-related patches, review them, and determine whether they are really ready for mainline. That way, ideally, Linus Torvalds never sees graphics code that's totally atrocious and untested and bad.

              Oh, wait. Except that he does. Remember the "UNTESTED CRAP" fiasco?

              Originally posted by asdx
              etc?
              Yes, the "etc" role is quite significant. Don't forget VMware, which is the company that bought up Tungsten Graphics. Tungsten is the company that originally designed the Gallium3d architecture, which is being used by all open source graphics drivers except for Intel's. The one guy who initially started the Mesa project way back in the 90s, Brian Paul, now works for VMware.

              A lot of people who are either unaffiliated with a company, or working for some small business, are significant contributors to Mesa and/or the DRM stack. While a "majority" of the work may be done by large enterprises such as Intel, VMware and Red Hat, the individual / small business contributor is a huge part of Mesa.

              Originally posted by asdx
              Who are the major Mesa contributors?
              I'm going to widen your question to include the entire open source graphics stack, which includes Xorg, libdrm, the DRM kernel subsystem, and Mesa.

              In the approximate descending order by level of commitment (first one is most committed/best contributor IMHO): Red Hat, VMware, Intel, AMD, a few dedicated people who are unemployed or whose day job is unrelated to graphics drivers, Google, PathScale, Oracle, Apple, IBM, and a "long tail" of people from various companies and individuals who have between 1 and 10 commits (in other words, not a whole lot).

              Comment


              • #17
                Originally posted by allquixotic View Post
                a few dedicated people who are unemployed or whose day job is unrelated to graphics drivers
                "At school" sounds better than "unemployed". The most common pattern is "working on open source drivers is more interesting than finishing my doctoral thesis"

                Comment


                • #18
                  Originally posted by bridgman View Post
                  "At school" sounds better than "unemployed". The most common pattern is "working on open source drivers is more interesting than finishing my doctoral thesis"
                  I find it worrying that you say this the day after I start reading an OpenGL book and clone the xf86-video-ati and mesa repositories...

                  Comment


                  • #19
                    Originally posted by archibald View Post
                    I find it worrying that you say this the day after I start reading an OpenGL book and clone the xf86-video-ati and mesa repositories...
                    :O!

                    Archibald! The next great Mesa committer!

                    Marek! Alex! You must capture him, before he sneaks away...

                    Comment


                    • #20
                      Originally posted by asdx
                      Red Hat, Intel, VMware and AMD should sue the hell out of HTC over that fucking ST3C patent.
                      Hmmm, no.

                      But seeing as HTC likes to make android-devices maybe they should approach them asking if they can have a no-suing-over-this/these-patents-aggreement for the good of the Linux ecosystem. I think it would be a win for HTC too.

                      Comment

                      Working...
                      X