Announcement

Collapse
No announcement yet.

The State Of OpenGL 3.x in Mesa Core

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

  • The State Of OpenGL 3.x in Mesa Core

    Phoronix: The State Of OpenGL 3.x in Mesa Core

    While ATI R600 users only recently received OpenGL 2.0 hardware support within the open-source Radeon 3D stack and there is many more OpenGL extensions to be implemented just not for the ATI Mesa driver but the other DRI drivers as well, Brian Paul has published a document that lays out the current state of OpenGL 3.x within the classic Mesa core. This document lays out what core Mesa supports and not necessarily that any of the drivers are implementing the said support at this time...

    http://www.phoronix.com/vr.php?view=Nzg1MQ

  • #2
    The focus for OpenGL from now on will probably be something that can be used for virtual GPU's.

    Comment


    • #3
      Hopefully the newer 3.x specs won't be as difficult as 3.0 support as a lot of infrastructure will be in place making it easier to add new functions

      As with GLSL, large chunks of this are being re-written from scratch, hopefully there will be increased interest from 3rd parties to get this support added too.

      Comment


      • #4
        Expect the extensions that people actually use to be done first.

        Comment


        • #5
          MESA Project needs a better infraestructure and a lot development power.

          This project have been dying over time, not catching new 3D stuff on time and losing relevance. That's why propietary graphic drivers won momentum.

          Now they have Gallium3D, but things are running a lot slower than desired. OpenGL is becoming a faster target than before due to the Kronos Group, a good thing in my opinion as DirectX is monopolizing everything and having a quite big feature set.

          They lack development power for making the MESA disaster to go on the proper direction. They are wasting time by not being totally centered in Gallium3D and making this nightware the shortest possible.

          MESA needs to found a proper Foundation just like any other project does. Tungsten Graphics, Inc. can be the leader force as they are the majority of developers involved on the code, but that's not enough and needs to be a more Free Software project.

          Linux Foundation should care a lot more of the weaker parts of the Open Source ecosystem (and the Linux ecosystem too because is mostly the same), instead wasting resources on stupid efforts like a niche Linux distro project targeting only a CPU family from a company (you know what one is...) . The most important stuff in the non-server area are the Open Source graphics drivers, because there is a very potential growing zone for Linux marketshare and propietary graphic drivers are really a big stone in the path for it.

          There are also other stuff like scanners drivers (SANE project is becoming a total joke), printer drivers (Apple owning CUPS scares me, Apple is a totally oportunist company in Open Source), OCR, technical proffessional tools (graphical, video, audio, scientifical...), clustering... This should be explained other time instead.

          Comment


          • #6
            Yeah I can't figure out why Intel isn't all over Gallium3D. They're the only ones around here who have graphics drivers done before launch day.

            Comment


            • #7
              Originally posted by timofonic View Post
              MESA Project needs a better infraestructure and a lot development power.

              This project have been dying over time, not catching new 3D stuff on time and losing relevance. That's why propietary graphic drivers won momentum.

              Now they have Gallium3D, but things are running a lot slower than desired. OpenGL is becoming a faster target than before due to the Kronos Group, a good thing in my opinion as DirectX is monopolizing everything and having a quite big feature set.

              They lack development power for making the MESA disaster to go on the proper direction. They are wasting time by not being totally centered in Gallium3D and making this nightware the shortest possible.

              MESA needs to found a proper Foundation just like any other project does. Tungsten Graphics, Inc. can be the leader force as they are the majority of developers involved on the code, but that's not enough and needs to be a more Free Software project.

              Linux Foundation should care a lot more of the weaker parts of the Open Source ecosystem (and the Linux ecosystem too because is mostly the same), instead wasting resources on stupid efforts like a niche Linux distro project targeting only a CPU family from a company (you know what one is...) . The most important stuff in the non-server area are the Open Source graphics drivers, because there is a very potential growing zone for Linux marketshare and propietary graphic drivers are really a big stone in the path for it.

              There are also other stuff like scanners drivers (SANE project is becoming a total joke), printer drivers (Apple owning CUPS scares me, Apple is a totally oportunist company in Open Source), OCR, technical proffessional tools (graphical, video, audio, scientifical...), clustering... This should be explained other time instead.
              Yeah, no.

              Mesa is more relevant today than ever before. Plenty of chipsets are accelerated with it. It's important enough that fglrx dropped support for all pre-HD chipsets and deferred to Mesa and Xorg for drivers.

              Tungsten is no more. You must have missed the memo. They got bought out by VMWare. LunarG is largely their replacement, but as before, their mission isn't to provide open-source drivers; it's to provide Unix-based drivers leveraging common open-source APIs. There are several closed-source Gallium drivers out there, and there will continue to be more.

              As far as development manpower, well, patches are welcome. Otherwise, get off my back; I don't get paid for this, and I'm on vacation. Get off everybody else's back, too; they're not paid for this either.

              All the features that you see in Mesa are being paid for or contributed by volunteers, and it's been a while since a volunteer added a GL extension.

              Originally posted by wswartzendruber View Post
              Yeah I can't figure out why Intel isn't all over Gallium3D. They're the only ones around here who have graphics drivers done before launch day.
              Intel is far more pragmatic than you might think. Their kernel-side drivers are open-source because that's the only way to get them in-tree, and the user-space drivers are open-source because that's the only way to get the kernel drivers accepted. (Kernel policy forbids drivers with specialized interfaces if there is no open userspace for testing.)

              The two Intel-based Gallium drivers are being done by VMWare, and Intel will not move to Gallium until they can be certain that it will confer a serious advantage.

              Comment


              • #8
                Originally posted by MostAwesomeDude View Post
                Yeah, no.

                Mesa is more relevant today than ever before. Plenty of chipsets are accelerated with it. It's important enough that fglrx dropped support for all pre-HD chipsets and deferred to Mesa and Xorg for drivers.
                The majority of the current IHVs closed source drivers are not OpenGL 3.2 nor GLSL 1.50.11 compliant yet as of today.

                The Mesa developers has released a lot of updates in the past two days and their OpenGL 2.1 API implementation is one of the best ones out there. They need to update the DRI drivers, like the R600 DRI driver, into being more OpenGL 2.1 compliant.

                As mentioned, they just need more volunteers in testing and providing feedback - if not developing code.

                Comment


                • #9
                  Originally posted by MostAwesomeDude View Post
                  Yeah, no.

                  Mesa is more relevant today than ever before. Plenty of chipsets are accelerated with it. It's important enough that fglrx dropped support for all pre-HD chipsets and deferred to Mesa and Xorg for drivers.

                  Tungsten is no more. You must have missed the memo. They got bought out by VMWare. LunarG is largely their replacement, but as before, their mission isn't to provide open-source drivers; it's to provide Unix-based drivers leveraging common open-source APIs. There are several closed-source Gallium drivers out there, and there will continue to be more.

                  As far as development manpower, well, patches are welcome. Otherwise, get off my back; I don't get paid for this, and I'm on vacation. Get off everybody else's back, too; they're not paid for this either.

                  All the features that you see in Mesa are being paid for or contributed by volunteers, and it's been a while since a volunteer added a GL extension..
                  But you confirmed part of my thinkings: MESA project is a half-running one, it needs a proper project organizational infraestructure.

                  The community only benefits if they provide Open Source, not just a nice Open Source API implementation. VMWare only cares of those developers making the VM more interesting for OpenGL/DirectX apps, LunarG only cares of getting paid for drivers. There's an obvious problem here, an obvious interest conflicts will merge to the surface in time.

                  So MESA needs to be backed by a powerful Foundation that not only cares of one part of the project but the combination of all. If not possible, probably it will end eventually replaced like other pieces of software. Live with the community or get out.

                  A proper MESA Foundation could make the efforts of both Tungsten and LunarG easier, and even able to have more paid developers too. Those devs would work on the main MESA important stuff instead focused interests from those companies, so both sides win.

                  Some people think MIT license for drivers and API stuff is a failure, I agree on it. But depending on corporations instead foundations is another big problem too.

                  Anyway, I getting more skeptic about MESA future. Lots of promises from ATI on their drivers, lots of promises about Gallium3D... but time goes and only hopes are there. Little real results, the path seems too long for them and people are tired of waiting for it.

                  Comment


                  • #10
                    Originally posted by timofonic View Post
                    MESA Project needs a better infraestructure and a lot development power. This project have been dying over time, not catching new 3D stuff on time and losing relevance. That's why propietary graphic drivers won momentum.
                    Have to disagree here - there seems to have been more visible progress on Mesa in the last couple of years than at any time in the last decade, at least on the hardware acceleration side. Brian and the other TG folks have been steadily pushing the core and software renderer ahead, but HW acceleration has really jumped ahead recently.

                    I don't think the state of open source drivers has that much effect on proprietary drivers, other than that people will *use* proprietary drivers when open source drivers are not available. I don't think the lack of open source drivers causes more effort to be spent on proprietary drivers, for example.

                    Proprietary drivers make it possible for HW vendors leverage large portions of their driver code across 100% of the PC market rather than having to write and maintain an entire driver stack specific to Linux. That, in turn, makes it possible to offer similar levels of functionality and performance across all OSes, at least on the 3D side.

                    Originally posted by timofonic View Post
                    Now they have Gallium3D, but things are running a lot slower than desired. OpenGL is becoming a faster target than before due to the Kronos Group, a good thing in my opinion as DirectX is monopolizing everything and having a quite big feature set.

                    They lack development power for making the MESA disaster to go on the proper direction. They are wasting time by not being totally centered in Gallium3D and making this nightware the shortest possible.
                    Not sure why you call it a disaster - it seems to have delivered a lot of results relative to the development effort expended, so I would call it a success in that respect. Also, a lot of effort in the last year has been spent on lower level functionality, particularly a kernel memory manager, which was a pre-requisite for further progress on the level of 3D support.

                    As MostAwesomeDude said, this is a volunteer project and more volunteers would always be welcome. There are a lot of users calling out for ever-increasing complexity and performance, however very few of them seem to be rolling up their sleeves and helping with the work. 3D is not a simple area, but it is probably the "least unfriendly" portion of the driver stack for new developers so there's no excuse for people not jumping in and helping.

                    Originally posted by timofonic View Post
                    MESA needs to found a proper Foundation just like any other project does. Tungsten Graphics, Inc. can be the leader force as they are the majority of developers involved on the code, but that's not enough and needs to be a more Free Software project.

                    Linux Foundation should care a lot more of the weaker parts of the Open Source ecosystem (and the Linux ecosystem too because is mostly the same), instead wasting resources on stupid efforts like a niche Linux distro project targeting only a CPU family from a company (you know what one is...) .
                    I have always felt that foundations should be funding more active development, but when you look closely at the economics they don't really have the money to fund very much - so they focus on what they feel to be "high leverage" or high visibility projects. These are all volunteer organizations, so the solution for anyone who doesn't like their direction is to get involved and CHANGE the direction.

                    Tungsten Graphics has done a fantastic job of operating a business while doing great things for the open source community, but that is an extraordinarily difficult management challenge and does not seem like a particularly scalable model unless some new and sustainable business models can be developed. Everyone talks about desktop Linux being the bright future and the big growth market, but so far nearly all the revenue streams are from the server business. That needs to change.

                    What desktop Linux needs is either a business model which will fund the required development effort -- which is probably going to require money coming from users -- or a user base which accepts the current volunteer-driven rate of progress and celebrates the successes rather than complaining.

                    The current model, with a few hard-working volunteers and an ever-more-demanding user base, is not really good for anyone. Saying "the beatings will continue until the morale improves" may sound funny from a distance but that is exactly how users are treating developers today.

                    Originally posted by timofonic View Post
                    The most important stuff in the non-server area are the Open Source graphics drivers, because there is a very potential growing zone for Linux marketshare and propietary graphic drivers are really a big stone in the path for it.
                    IMO the days of proprietary drivers being an obstacle are behind us, albeit only recently. Distros can now ship with open source drivers for all major GPU vendors - not full featured in all cases, but enough to allow distro developers to push ahead with architectural changes without having to wait for the proprietary drivers. That's why the recent Nouveau milestones were so important.
                    Last edited by bridgman; 01-01-2010, 12:21 PM.

                    Comment


                    • #11
                      Originally posted by timofonic View Post
                      Anyway, I getting more skeptic about MESA future. Lots of promises from ATI on their drivers, lots of promises about Gallium3D... but time goes and only hopes are there. Little real results, the path seems too long for them and people are tired of waiting for it.
                      Excuse me, but exactly what are these promises you are talking about ?

                      We promised to write and release documentation and to fund some initial development effort to jump-start the process. AFAIK we are doing all that and more.

                      Nobody is making promises about Gallium3D as far as I know -- if they are I would like to see them as well.

                      Comment


                      • #12
                        What desktop Linux needs is either a business model which will fund the required development effort -- which is probably going to require money coming from users
                        Where do I send the money?

                        I'd gladly pay $100/year to fund development. That's more than I would be paying Microsoft for an upgrade once every 4 years, even. $100/year is nothing if it meant having a fully-usable Linux desktop.

                        I still 6 months later get constant corruption on my R700-based desktop with the very latest kernel, mesa, drm... it's like nobody gives a shit about quality in Linux land. I'd much rather have a stable desktop than GL3 right now. I've been consistently spending more and more and more time in Windows 7 just because it actually works, not to mention works better in some ways, especially ease-of-use for when I just want to get shit done and not dick around for 15 minutes doing things the Linux way. If there were someone I could help fund to reverse this situation, I would. Promise. I used to buy the Red Hat releases when they come out, but Red Hat stopped selling consumer OSes. Ubuntu... I'm not sure where their money comes from, I've never seen them ask for a cent for anything. I don't have the time or the interest to contribute to most of the things in Linux land that don't work well or at all, though I do contribute to some areas that hit my fancy (none related to graphics). It's like I'm stuck using Free Software that can't actually get the job done or I'm stuck using proprietary software.

                        So yeah, if a measly $100/yr could make a difference, I'm all for it.

                        Comment


                        • #13
                          I'd also pay a few dollars. But most devs are hired by a company, e.g. AMD, Red Hat, Novell, Intel and they do ATM even more than they have to do AFAICS, for example in their free time. And other unpayed developers do as much as they can do in their free time anyway.

                          I would like to support the linux graphics ecosystem in general, but don't know how to do this, and also would like to support special features, like overscan support in free drivers, so that I can use free drivers to use my TV (connected over HDMI, and I cannot disable overscan in the TV). I don't know how much work it would be to implement overscan in the graphics driver, but there are projects like http://www.kickstarter.com/ where one can spend money on projects, which need a special amount of money. And if you want to support this special project, you say "Yes, I will pay X $, if there are enough other people, who want to support, too." And only if enough people would spend the money, it gets transferred.
                          I'd like to support, but don't know how.

                          Comment


                          • #14
                            BTW Mesa OpenGL stack will still get used with Gallium3D drivers. It is not like Gallium3D is designed to replace Mesa or anything like that.. Gallium is dependent on Mesa if you ever want to have OpenGL support.

                            Maybe there is confusion going on in this thread?

                            Here is how it goes:

                            Xorg DDX = Device Dependent X. It's is X's own proprietary (as in their own special way, not closed source) way to do 2D driver. This provides EXA 2D acceleration.

                            Linux DRM = Linux Direct Rendering Manager. The Linux kernel's way to allow managed access to the hardware. This provides the DRI and/or DRI2 protocol to userland, which is a "Stable ABI" designed for graphics drivers. DDX does not use this interface and instead accesses the hardware on it's own.

                            DRI/DRI2 = The cross platform API in which userland graphics drivers can interact with the hardware through a kernel DRM.

                            Mesa = Open Source OpenGL implementation. They also provide DRI/DRI2-speaking 3D drivers.

                            Gallium3D = New driver model focusing on minimizing the amount of hardware specific code and is designed to support multiple APIs in the form of "state trackers".



                            Now... The long term goal is so that you can get rid of the need for DDX for 2D acceleration and a unique DRI driver for 3D. Instead you have a generic Xorg state tracker for 2D and a Mesa state tracker for 3D.

                            All the work going into things like kernel memory management, kms, mesa etc etc. is all still going to benefit Gallium drivers.

                            --------------------------


                            Now as far as Intel....

                            The reason Intel is focusing on the current drivers is because users don't care about the APIs used in their drivers. Instead they want stability and featureness.

                            Right now they have finished making the transition away from having X access the hardware directly and instead is now using the kernel for memory management for UXA (EXA on top of kernel memory management), which required a buttload of other changes.

                            At the beginning of 2009 the drivers were at a low point because they were in a state of transition from old stuff to the newer stuff. Now they at the point were they are stabilizing and doing performance enhancements.

                            Gallium is still too far off for Intel to care about going through another transition period just yet. Once the current drivers reach a satisfactory state were they can just sit and be used for a year or two then that is when the focus will again switch to Gallium.


                            --------------

                            The open source ATI drivers are just now getting to about the same point that Intel's were in 2009.

                            They are getting usable and they are now supporting the hardware and feature level needed by most _desktop_ type activities, but are going to be buggy and irritating to deal with.

                            But I think that ATI will transition faster then Intel due to the fact that only some of Intel's video hardware supports modern feature while pretty much all ATI hardware does. Also ATI users are more performance and feature oriented so they should end up putting more work into testing and helping out the developers. Especially when Gallium starts entering the picture in a big way.. because at that point it should be at least _possible_ to support the same level of features with OSS that are currently only available for proprietary Nvidia driver users.

                            Then at that point OSS drivers should start becoming a real viable alternative for people _needing_ more then just 3D desktop support.

                            Comment


                            • #15
                              Ironically, nouveau is developed exclusively toward Gallum3D...

                              Comment

                              Working...
                              X