Announcement

Collapse
No announcement yet.

Mesa Developers Discuss Branching Off Old Drivers, Including R300g & i915

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

  • #11
    They should branch off r600 and these elder drop altogether

    6 year old chips should be branched off and 10 year old should be dropped. Last chip which works with r300 was released right there 10 years ago, so good bye - just wrote something like "this is last mesa verison who support these, source is available... ciao!"

    If community (on average Dave) wanna prolonging it for RHEL he can do it, but really who else care . Average Joe can use different distro with longterm kernel and so on, for these bravest ones literaly can be prolonged to 15 years even when it is dropped
    Last edited by dungeon; 26 May 2017, 11:09 AM.

    Comment


    • #12
      Originally posted by dungeon View Post
      They should branch off r600 and these elder drop altogether
      6 year old chips should be branched off and 10 year old should be dropped.
      2 years ago I bought an A4-4000, that was released 4 years ago and uses the r600 driver
      I hope it will last a little bit longer than dropping it now

      Comment


      • #13
        Originally posted by trek View Post

        2 years ago I bought an A4-4000, that was released 4 years ago and uses the r600 driver
        I hope it will last a little bit longer than dropping it now
        That one should be branched off, which is not a drop... drop is entire drop from newer mesa releases (it might live even in same git still for some time, but for releases anything older than 10 years should be ditched) which for your chip should happen after another 4 years or so.
        Last edited by dungeon; 26 May 2017, 11:15 AM.

        Comment


        • #14
          Nothing should be dropped until it is 100% complete with its intended functionality. There are no ATI/AMD GPUs that are 100% done, though R200 looks to be the closest:

          If a "mesa-legacy" were created, it wouldn't matter if that keeps up with "mesa-current" - they could both be on their own update schedule.

          Comment


          • #15
            Originally posted by schmidtbag View Post
            Nothing should be dropped until it is 100% complete with its intended functionality.
            It can live in git somewhere, but releases should not include these. So that is still not a classic drop, like blobs drop it

            There are no ATI/AMD GPUs that are 100% done, though R200 looks to be the closest:
            https://www.x.org/wiki/RadeonFeature/
            When you look at papers anything looks fine, but i still have that r200 supported card somewhere... About 4 years ago was last time i tried that, and i hit 3-4 obious bugs, in GL even lighting was broken, so for sure no one used that nor tested that with GL ... and there was another couple bugs with rendering...

            OK i bisected that and fixed it, played Angry Birds in Wine that worked But HyperZ was not supported there, in UMS i remember it was and with KM it is not, so it is slow literally half the speed everywhere. If that is not enough i tried Firefox and literally upon first big enough loaded image entire machine locked up

            So i say OK, lets try to watch some video mplayer gl... nope even that does not work fine, hit ttm regression with 3.14 kernel, used CPU look crazy, bisected that too maintainer reverted it, etc...

            Basically it was a crap, reading commits i found Dave patched something year or two befor ethat, just to hide things that it does not looks like so much broken on the first sight but underneath it it is can of worms

            Etc... just stop reading that feature crap, in reality old chips are severely broken long ago and no one really care

            Believe me, it is much better to drop something while it still working even if it is not total completed, than to reinvent a wheel and to break everything And r600 state exactly smells like that to me, better do not add anything just branched it off
            Last edited by dungeon; 26 May 2017, 12:19 PM.

            Comment


            • #16
              Originally posted by Tomin View Post

              Well, Gnome requires at least OpenGL ES 2.0, if I remember correctly. I think that's reason enough to have acceleration on r300. I know some people using cards with r300 drivers even though I don't personally have any of those anymore.
              I do have some, and play around with them sometimes. Not only are they weak for Gnome, the surrounding hardware (AGP era, P4s and stuff like that) can't even run a modern linux distro properly. It is a corner case. More fit for a special git branch, than software usually distributed for recent hardware.

              Comment


              • #17
                Originally posted by schmidtbag View Post
                Nothing should be dropped until it is 100% complete with its intended functionality. There are no ATI/AMD GPUs that are 100% done, though R200 looks to be the closest:

                If a "mesa-legacy" were created, it wouldn't matter if that keeps up with "mesa-current" - they could both be on their own update schedule.
                Since none is arsed to maintain or improve code for ancient cards and none is payign $$ for that to happen, you are basically saying "Nothing shuld be dropped EVER", which is plain wrong.

                Comment


                • #18
                  That is exactly a thing, Tarceri works for Valve... Valve wanna new drivers only, what Tarceri wants looks exactly what is more handy for Steam, SteamOS.

                  The motive for moving out these old graphics drivers is it will become much easier to clean-up the current core Mesa code, a lot of features can then be enabled by default that apply to newer hardware than continuing to add more run-time checks, and removing older code relevant to only these vintage GPUs. These old drivers barely receive any testing as-is on Mesa Git and thus are already subject to breakage and regressions.
                  I am sorry, but if that is "the motive" i must use more smiles

                  Neither as opposite what Bridgman said discussion actually goes nowhere, as there is no discussion They should say "Valve wanna this, can we do it?", that is more honest question than writting crap around

                  Even without that Mesa should apply general rule like - any 6 years hardware goes branch off and any 10 years old hardware go out of releases. Code for elder still seats off of it somewhere else, so if you care - it is there

                  And if older RHEL and Dave does not care, who care about these i dunno... just do not tell me he can't use git
                  Last edited by dungeon; 26 May 2017, 01:05 PM.

                  Comment


                  • #19
                    Originally posted by starshipeleven View Post
                    Since none is arsed to maintain or improve code for ancient cards and none is payign $$ for that to happen, you are basically saying "Nothing shuld be dropped EVER", which is plain wrong.
                    Keep in mind the definitions of the terminology being used. I feel this older, outdated, and neglected hardware should have it's own branch independent of current mesa. So yes, I want it literally dropped from current mesa, but it shouldn't be dropped entirely (in other words, I don't think these drivers should cease to exist). These drivers should still be maintained and should still get some patches from current mesa. Keep in mind many of them are still technically capable of running modern systems comfortably; whether they actually do is a different story, and separating them from current mesa seems to be a good start in fixing that.

                    I'm not saying things should never be dropped, because I agree that is ridiculous. For example, I was in-favor of the original Pentium being dropped from the Linux kernel, because you could probably go back to the 2.4 kernel and still have all the functionality you need on such a platform.
                    Last edited by schmidtbag; 26 May 2017, 01:15 PM.

                    Comment


                    • #20
                      This is one of those situations where I think "I still have that hardware around somewhere" but in reality I very rarely use it. Having it supported in a legacy branch hardly affects most users considering the hardware in question is 8+ years old now (my 9600 is 15 years old). AFAIC, that's *really* good as far as support for consumer hardware goes.

                      It's likely to be more of a problem for distro package maintainers who may have to maintain another set of package but otherwise, it looks like a good compromise to allow better support for modern hardware.

                      BTW, some of this older hardware still is capable of running a desktop and I did so quite comfortably with LXQt on an Intel 945 GPU until recently.

                      Edit: Actually, the 15 year old 9600 card uses r300g and would still be supported if the responses are anything to go by.
                      Last edited by ResponseWriter; 26 May 2017, 01:42 PM.

                      Comment

                      Working...
                      X