Announcement

Collapse
No announcement yet.

Direct3D 10/11 Is Now Natively Implemented On Linux!

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

  • Originally posted by eternaleye
    I think he meant "jibe with". It means that a set of pieces of information are collectively consistent.
    That makes perfect sense now, thanks!

    Originally posted by Svartalf View Post
    Heh... Stuff I get told in confidence and can't disclose any of the details thereof as it'd get my contacts in trouble...
    Fair enough, fair enough.


    Originally posted by Starvalf
    Yeah, I know that doesn't count as proof. But what I've been shown corroborates this- and just because it's a power point slide, it WAS the one MS led with and they DO have a better handle than many on just how much "incursion" into the market we actually have. If you want other proof go digging through the Comes stuff- they give out similar figures in that mountain of actual Discovery documents from MS.
    OK, so we seem to agree on the logic and on what counts as solid information. For some reason I find the issue of Linux market share/usage to be quite fascinating, perhaps because there are so many uncertainties and lack of reliable information about it. Also, despite being quite familiar with the low figures it still is a hard pill to swallow, but I can't accept something just because it points to where I'd like the numbers to be.

    If I have to guess, I'd say that the two entities that have the most accurate grasp about it are Microsoft and Google. Unfortunately, Google doesn't disclose its data, and all we have from Microsoft is to be taken with two spoons of salt. Of course, without insider information such as the one you claim to have the question remains entirely open (regardless of what over-enthusiastic, questionable bloggers may want us to believe). I still think that the most reliable indicators are web browsing statistics: it's better having a number with an idea of where the pitfalls are than just having a number. That you and others have access to private information it's a different matter and I value your limited sharing of it, but I can't base my judgement on this for intellectual honesty reasons (which I'm sure you'd share if the roles were reversed). We'll see with time.

    Comment


    • Originally posted by yotambien View Post
      I still think that the most reliable indicators are web browsing statistics
      Considering that people change their browser agent string in Linux to sidestep bad javascript, etc. on pooly designed websites, coupled with the reality that most of the browser stats sites are covering stuff that is less likely to have a Linux browser show up...

      Your choice isn't ANY better than the ones I use...

      Comment


      • this may just be the alcohol talking, as i have had a fair few. However it seems that Michael has hit the linux hive as far as page views are concerned.

        Comment


        • Originally posted by Svartalf View Post
          And before anyone comments, YES, I know I said it's counter-productive and all. I still hold that it is- and if it weren't next to impossible to do the taxes any other way, I'd do without it because it IS a vote for more Windows software. But, much like trying to get a powerful high-end laptop with Linux is impossible, I'm kind of trapped on that score.
          I couldn't help but pick up on this one...

          Dell do sell Precision laptops with RHEL preinstalled, and I would have thought that Precision was pretty much the definition of high-end Dell laptop.

          Comment


          • Originally posted by baryluk View Post
            Wow, this commit is huge. Will see how good it will be.

            I still do not see why it is superior to the OpenGL API in terms of clean code or performance. But I actually do not know DX11 very well at all.

            But this also opens a way to use gallium as native DX drivers under windows, and virtualizing 3D of emulated Windows. Was there any testes on Windows with it? With native drivers maybe?
            Actually, virtualization is a major application of the Gallium3d architecture. Since Gallium's development is led by VMware, this area is actively being developed.

            If you owned a copy of VMware Workstation 7, you would be able to get an early look at the still-in-development "vmwgfx" graphics stack, which is a gallium driver that runs inside a Linux guest, and communicates with the hypervisor's virtual GPU for 3d acceleration. It's got a lot more smarts than something like GLX-over-IP or VNC. Of course, the downside is that a majority of the complexity (the layer that translates the guest GPU I/O into graphics stack calls host-side) is closed source, within the VMware Workstation product. But you can at least see how the vmwgfx driver communicates with it on the inside.

            I can't be sure, but I think they are either using or planning to use Gallium (or parts of it) to implement the guest WDDM driver for Windows guests. However, since Windows guests on VMware Workstation 7 already have full 3d accel and WDDM 1.1, anything they have been planning to do has already been accomplished -- I'm just not sure how they did it.

            As it stands, the Linux guest 3d accel in Workstation is about as stable and usable as r600g, but the Windows guest 3d accel is about as stable as Nvidia's binary blob for native hardware. So the Windows driver probably doesn't use Gallium now.

            Comment


            • Replying to myself...

              This may sound twisted, but one possible way for you to test out this Direct3d state tracker support today without running a hardware Gallium3d driver would be :

              0. Use a CPU with Extended Page Tables (Core i3/i5/i7) if you want decent performance, otherwise you probably won't be able to tell the difference between software rendering and hardware
              1. Use a Windows or Linux host OS
              2. Buy VMware Workstation 7
              3. Install a very recent Linux distro as a guest
              4. Get the vmwgfx graphics driver working well with OpenGL and stuff, to make sure that the hardware driver portion of it is working properly. At least make sure glxgears is 3d accelerated, and maybe enable the EXA state tracker (although EXA had corruption problems last I checked). Without EXA, 2d in the guest will be done in software within the guest.
              5. Turn on the D3D10 state tracker in Gallium and rebuild (or it enable it from the get-go if you're feeling adventurous!)
              6. Download or write a minimal Direct3d 10 application in C using as few calls as possible (to keep it as easy to debug as possible)
              7. Build it as a native Linux binary
              8. Run and pray!

              Comment


              • If you only want a fully 3D accelerated Windows guest in Linux, there's the theoretical possibility of using an IOMMU to assign a graphics card completely to the guest. I'm pretty sure KVM can already do that, at least with a secondary card. But you need a motherboard equipped with an IOMMU.

                Comment


                • D3D 9

                  Is there any possibility that Direct3D 9 is also implemented by the same way. Because there are thousands of existing Direct3D 9 games that can be played by WINE.

                  Comment


                  • Originally posted by imamdxl View Post
                    Is there any possibility that Direct3D 9 is also implemented by the same way. Because there are thousands of existing Direct3D 9 games that can be played by WINE.
                    VMWare is implementing a closed source D3D9 tracker, but I wouldn't count on anything being added into Mesa for WINE. The entire Gallium architecture has basically been built around the D3D10 specs, which is what made it fairly simple to do this. D3D9 would be much more complicated to implement (10 dropped a lot of nasty legacy stuff).

                    Comment


                    • Originally posted by Svartalf View Post
                      Considering that people change their browser agent string in Linux to sidestep bad javascript, etc. on pooly designed websites, coupled with the reality that most of the browser stats sites are covering stuff that is less likely to have a Linux browser show up...

                      Your choice isn't ANY better than the ones I use...
                      I doubt that many people change their browser strings. I don't personally know anybody that has done that and I also seriously doubt there are hordes of Ubuntu users changing their browser strings. Wouldn't we have more Windows users of Firefox having changed their browser strings back in the day than total Linux users anyway? Nobody ever brings that up.

                      Comment

                      Working...
                      X