Announcement

Collapse
No announcement yet.

Legacy Mesa Drivers Receive Their Death Sentence

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

  • #31
    Originally posted by PsynoKhi0 View Post
    Darn... Still using savage, tdfx and r128 here... I actually find it more satisfying to resurrect old systems than to participate in FPS bragging contests.

    Oh well, there are a couple of distros aimed at old HW anyway, there's hoping legacy Mesa will be in maintenance mode through those.
    Or use a distro that uses an older Xorg/Mesa that has the driver for whatever old GPU you have on whatever old machine. It may be possible that some people might provide patched "legacy" drivers for newer Xorg versions though, or if you are adventurous you can probably hack the driver code yourself to make it work on newer Xorg's

    Comment


    • #32
      Looking forward to a cleaner source tree.


      Some thoughts/ideas on additional clean up and improvements.
      Another area of mesa which is in need of a good overhaul is the output during the compile. Looking through 99.9% compiler options rushing by on the screen isn't what I would call informative. Keep it simple and only output errors/warnings and file names. And please finish it off by telling the user that the build finished successfully.

      Comment


      • #33
        Originally posted by Silverthorn View Post
        Another area of mesa which is in need of a good overhaul is the output during the compile. Looking through 99.9% compiler options rushing by on the screen isn't what I would call informative. Keep it simple and only output errors/warnings and file names. And please finish it off by telling the user that the build finished successfully.
        I'd love to see that as well; it would make warnings a lot easier to spot.

        The Mesa build system certainly is a disaster; especially the fact that we now have four of them (autoconf/plain Makefiles/SCons/Android). We've long wanted to clean it up and make one build system that works well...it's just a lot of tedious work, and we're all pretty swamped already. In the meantime, we've been cleaning house, removing or splitting out a bunch of old code that no one cares about, as well as code that basically never changes. In particular, we split out Mark Kilgard's GLUT into its own repository (hardly changes and most people just use FreeGLUT), libGLw (OpenGL widgets for Motif/Athena...seriously, who cares?), and are going to do the same for libGLU (used, but never changes). This means that when we do rework the build system, we won't have to deal with those components...just core GL and drivers. Deleting the DRI1 drivers will help as well.
        Free Software Developer .:. Mesa and Xorg
        Opinions expressed in these forum posts are my own.

        Comment


        • #34
          Originally posted by Silverthorn View Post
          Another area of mesa which is in need of a good overhaul is the output during the compile. Looking through 99.9% compiler options rushing by on the screen isn't what I would call informative. Keep it simple and only output errors/warnings and file names. And please finish it off by telling the user that the build finished successfully.
          I always use:

          make -j4 >/dev/null

          The output of that contains only useful info (I don't consider printing every file name useful).

          Comment


          • #35
            Originally posted by gilboa View Post
            The question is - what is the greater good.
            By time Intel decided to remove the cruft, most of the user base of **new** distributions were already using $new_chips instead of obsolete i810 supported devices.
            Keep in mind that if you are using an old machine, you're most likely -far- better of with using older, long term distribution (such as SL/CentOS/RHEL as old as 4.x) instead of using Ubuntu and Fedora.

            E.g. my PII/333Mhz/MACH64/256MB laptop will most likely use CentOS 5.x till it dies.

            - Gilboa
            My 800Mhz G4 Powermac w/ Radeon 7000(R100) does not agree...

            Ubuntu 11.04 runs just fine for anything non gaming or HD video, All non HD and youtube video plays fine in Totem, FF6 runs fine with 50-70 tabs open across 3-4 windows with no flash anything installed and a bunch of privacy, security and anonmysing addons running, Thunderbird runs fine, Libre Office, Rhythmbox, Pidgin with 10 IRC rooms and chats with 20 or so friends across 5 different protocols are still all good, even when all of that is running I can still manage to pop open Totem and play a video without frames being dropped. The only time that happens is when running Transmission, tie up the IDE/PATA bus with all the reads and writes happening.

            So yeah, while this box may fail to run Doom properly, let alone OpenArena, it still remains a very viable desktop machine. Though if I do need to game I can just boot into OS X and play AvP, Sacrifice, Star/Warcraft2-3:FT, Fallout 2, THPS2, EV:Nova, Majesty and so on and so fourth.

            Since Apple dropped Rosetta and I have allot of PPC OS X games and even allot of even older OS 9 games I have no intention of ever getting rid of this old Mac and if I could afford it, would even upgrade it to a dual 7448 1.8Ghz, FireGLX3 AGP(flashed to Mac X800XL ROM) and 1.5Gb of PC133, though I remember seeing 1Gb sticks years ago, nobody seems to have any anymore, so no seeing if it maxes out at 3Gb. I could go even bigger with the GPU to a 7800GS AGP by flashing the ROM with that of a Mac 7800GT, but the Nouveau NV49 drivers are still very rough compared to drivers for the R400 based X800.

            Comment


            • #36
              Originally posted by Kivada View Post
              My 800Mhz G4 Powermac w/ Radeon 7000(R100) does not agree...

              Ubuntu 11.04 runs just fine for anything non gaming or HD video, All non HD and youtube video plays fine in Totem, FF6 runs fine with 50-70 tabs open across 3-4 windows with no flash anything installed and a bunch of privacy, security and anonmysing addons running, Thunderbird runs fine, Libre Office, Rhythmbox, Pidgin with 10 IRC rooms and chats with 20 or so friends across 5 different protocols are still all good, even when all of that is running I can still manage to pop open Totem and play a video without frames being dropped. The only time that happens is when running Transmission, tie up the IDE/PATA bus with all the reads and writes happening.
              The r100 driver will still be around, it's only the really old drivers that are going away.

              Furthermore, even if it was going away, there is nothing stopping you from sticking with Ubuntu 11.04 on that machine. And if you really wanted to upgrade to a newer distro, you could always just port Mesa 7.11 forward onto your distro (or hope that someone has created a PPA to do that). Old versions of Mesa generally shouldn't stop working just from a distro upgrade, so it shouldn't be too much trouble to keep around.

              Comment


              • #37
                Originally posted by smitty3268 View Post
                The r100 driver will still be around, it's only the really old drivers that are going away.

                Furthermore, even if it was going away, there is nothing stopping you from sticking with Ubuntu 11.04 on that machine. And if you really wanted to upgrade to a newer distro, you could always just port Mesa 7.11 forward onto your distro (or hope that someone has created a PPA to do that). Old versions of Mesa generally shouldn't stop working just from a distro upgrade, so it shouldn't be too much trouble to keep around.
                And now for some facts...

                intel i810 was releast in 1999. ATI R100 was released in 2000. The first VIA unichrome was announced in 2002 and i got it in the summer of 2003. Actual age has absolutely nothing to do with it.

                Now, have you at any time in the last 3 or so years updated either of kernel, xserver or mesa with anything more than bugfixes? Do you realize the tight dependency that exists between those three parts and the libdrm in between, a dependency that has existed pronouncedly since libdrm grew driver dependant bits.

                Apart from the concerns i raised (i still did not see an answer to my questions about which criteria make a driver suitable for continued inclusion, nor about which drivers then actually remain -- both are questions to which apparently the answers are not helpful to the proposal): where does this leave the X drivers or the drm drivers for these mesa drivers? Are they really thinking about _just_ the mesa bits, or is this the start of something bigger (it is of course something bigger, but they do not want to own up to it yet). And where will this end: will we only have 3 manufacturers cards supported in future across the board, only for linux, and only if the hardware is just old enough to have support, but not old enough to have its support broken again?

                Comment


                • #38
                  Originally posted by libv View Post
                  i still did not see an answer to my questions about which criteria make a driver suitable for continued inclusion
                  How about an actual user base that uses the open-source 3D stack and a healthy dev community supported by documentation (or enough market share/demand to do reverse engineering in the case of nouveau)? Yeah, it's sad that only the big 3 currently have a viable open-source dev community built, but it's not like other manufacturers' products don't/didn't have a chance. For example, if VIA chips get modern driver features, they can join the big boy club too. Maybe you should get working on that...

                  Comment


                  • #39
                    Originally posted by libv View Post
                    And now for some facts...

                    intel i810 was releast in 1999. ATI R100 was released in 2000. The first VIA unichrome was announced in 2002 and i got it in the summer of 2003. Actual age has absolutely nothing to do with it.

                    Now, have you at any time in the last 3 or so years updated either of kernel, xserver or mesa with anything more than bugfixes? Do you realize the tight dependency that exists between those three parts and the libdrm in between, a dependency that has existed pronouncedly since libdrm grew driver dependant bits.

                    Apart from the concerns i raised (i still did not see an answer to my questions about which criteria make a driver suitable for continued inclusion, nor about which drivers then actually remain -- both are questions to which apparently the answers are not helpful to the proposal): where does this leave the X drivers or the drm drivers for these mesa drivers? Are they really thinking about _just_ the mesa bits, or is this the start of something bigger (it is of course something bigger, but they do not want to own up to it yet). And where will this end: will we only have 3 manufacturers cards supported in future across the board, only for linux, and only if the hardware is just old enough to have support, but not old enough to have its support broken again?
                    ^This. All of it.

                    Dropping these old drivers is just straight up lazy, I've given out lots of junked "ancient" boxes loaded with Linux to various charities, should all of those people be unable to update their software ever?


                    I'm really not liking this new belife system being taken up by the OSS devs that everyone has to run the latest and greatest everything or they can die in a fire. Things like Gnome2 and KDE3 worked great and only needed to stay bugfixed and security patched, there was no reason to force everyone to jump on to a new stack that felt like we had taken a decade long step backwards just so they could reinvent the wheel.

                    So now we're going to see this with the GPU driver devs as well? And their answer is stick with outdated and insecure versions of a distro? Sorry but no, most of us don't need to rat race the FPS wall or couldn't possibly give less of a shit about whatever useless whizbang wobbly window effects you and think are cool, but anyone with the mental maturity past that of a 10 year old doesn't care about.

                    We just want software thats up to date with security and just fucking works without having to rewrite it ourselves. Its the reason some of us started using OSS software in the first place, we wanted something that could do those things things and not get in the way and that was true up until very recently.

                    But this is just the opinion of a non coding heathen, so as usual my opinion will be completely ignored.

                    Comment


                    • #40
                      But this is just the opinion of a non coding heathen, so as usual my opinion will be completely ignored.
                      Nah, it will be ignored because it doesn't make any sense.

                      The new driver stack was needed in order to progress past OpenGL 1.5. The drivers could never progress past it otherwise. End of story.

                      If there are 10-year old drivers that nobody is actively maintaining, then it's a problem.

                      Comment


                      • #41
                        Originally posted by Kivada View Post
                        Dropping these old drivers is just straight up lazy, I've given out lots of junked "ancient" boxes loaded with Linux to various charities, should all of those people be unable to update their software ever?
                        That's rather the point. The drivers that are being removed haven't been updated in years. Folks can still build Mesa 7.11 and play Quake2 on their Voodoo3 or Rage128 to their heart's content....

                        Comment


                        • #42
                          Originally posted by libv View Post
                          And now for some facts...

                          intel i810 was releast in 1999. ATI R100 was released in 2000. The first VIA unichrome was announced in 2002 and i got it in the summer of 2003. Actual age has absolutely nothing to do with it.
                          But how many people use them does. ATI R100 use was much more widespread the unichrome, hence it's not going away yet. Plus AMD is still willing to provide some (very limited) support for it through Alex Deucher.

                          Apart from the concerns i raised (i still did not see an answer to my questions about which criteria make a driver suitable for continued inclusion, nor about which drivers then actually remain -- both are questions to which apparently the answers are not helpful to the proposal)
                          I thought that was pretty obvious. The criteria is: any driver that has a maintainer gets to stay in. Any driver that doesn't have any developers willing to work on it, is in danger of being dropped. If you really don't want unichrome support to go away, then just announce on the list that from now on you intend to maintain it.

                          where does this leave the X drivers or the drm drivers for these mesa drivers? Are they really thinking about _just_ the mesa bits, or is this the start of something bigger (it is of course something bigger, but they do not want to own up to it yet).
                          Neither of us know. But we should focus on each proposal as it comes up, rather than trying to stir up fear, uncertainty, and doubt about future actions that may or may not ever happen. Focus on the present.

                          And where will this end: will we only have 3 manufacturers cards supported in future across the board, only for linux, and only if the hardware is just old enough to have support, but not old enough to have its support broken again?
                          Again, it just requires maintainers to do the work. Why should Intel be forced to subsidize via drivers? If Via or the community come up with someone to do the work, it will stay. Otherwise, tough luck. That's the open source way.

                          Comment


                          • #43
                            Farewell old drivers: http://cgit.freedesktop.org/mesa/mes...58f087bd8bf94d

                            Comment


                            • #44
                              Originally posted by Kivada View Post
                              But this is just the opinion of a non coding heathen, so as usual my opinion will be completely ignored.
                              You are missing the main point of OSS.
                              OSS is *NOT* a democracy, its a meritocracy - as in, those who are willing to do the actual work, get to make the decisions.
                              As long as no-one jumps in and offer to maintain the old drivers himself, these drivers will simply die. Period. End-of-story.

                              So in short, yes, unless you are willing to pitch in, you opinions don't really matter. (Nor do mine )

                              - Gilboa
                              DEV: Intel S2600C0, 2xE52658V2, 32GB, 4x2TB + 2x3TB, GTX780, F21/x86_64, Dell U2711.
                              SRV: Intel S5520SC, 2xX5680, 36GB, 4x2TB, GTX550, F21/x86_64, Dell U2412..
                              BACK: Tyan Tempest i5400XT, 2xE5335, 8GB, 3x1.5TB, 9800GTX, F21/x86-64.
                              LAP: ASUS N56VJ, i7-3630QM, 16GB, 1TB, 635M, F21/x86_64.

                              Comment


                              • #45
                                Originally posted by gilboa View Post
                                You are missing the main point of OSS.
                                OSS is *NOT* a democracy, its a meritocracy - as in, those who are willing to do the actual work, get to make the decisions.
                                As long as no-one jumps in and offer to maintain the old drivers himself, these drivers will simply die. Period. End-of-story.

                                So in short, yes, unless you are willing to pitch in, you opinions don't really matter. (Nor do mine )

                                - Gilboa
                                While that may be the case many coders make it sound like these kinds of things can be kept going in less then 5 lines of code that they could do in their sleep, if that is in fact the case then how much time could it possibly take them just keep gimping it along?

                                It's almost as bad as the coder that wanted to drop support for the ISA bus, when tons of devices in current gen hardware still uses it.

                                Maybe they can't possibly fathom why someone would be using old hardware, perhaps they've forgotten that the vast majority of people out there that are using a computer are using one that is more then 5 years old, that Linux has long been the champion of old tech users since Windows EOLs and newer versions run like crap on old hardware.

                                I thought the game plan was to gain more market share, so as to make these OSS coder's skills more in demand allowing them to ask for more money when being picked up by an Intel or AMD. Isn't that the end goal? To get paid to write OSS code?

                                Comment

                                Working...
                                X