Announcement

Collapse
No announcement yet.

Experimental Code Published For Virtual CRTCs

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

  • #31
    Originally posted by dfx. View Post
    in F/OSS world users and devs are potentially the same thing therefore it's better to apply same standards for both. besides, there are also occasional hackers in between.

    i also thought about F/OSS software as of mostly neat and clean solutions in contrast to closed stuff. proprietarists "good enough"? saying is not good enough for me anymore.
    Everyone who works in GPUs calls them crtcs. HW documentation calls them crtcs. It would probably be more confusing to call them something else.

    Comment


    • #32
      How about "the block formerly known as CRTC" ? That should make everyone happy
      Test signature

      Comment


      • #33
        re: re: re: naming

        Originally posted by agd5f View Post
        Everyone who works in GPUs calls them crtcs. HW documentation calls them crtcs. It would probably be more confusing to call them something else.
        it's just creating/enforcing "professional slang" (if you believe in that kind of term) only the narrow group of people can understand. i can't see how is that a good thing to create/enforce a language barrier beyond actual language with newcomers.

        Originally posted by bridgman View Post
        How about "the block formerly known as CRTC" ? That should make everyone happy
        that actually makes sense. and any sense is better than none.

        Comment


        • #34
          Originally posted by dfx. View Post
          it's just creating/enforcing "professional slang" (if you believe in that kind of term) only the narrow group of people can understand. i can't see how is that a good thing to create/enforce a language barrier beyond actual language with newcomers.
          It's no more confusing than any other name. You have to have some understanding of the hardware to understand the code. It goes back to vga which is a good starting point and on some chips the registers even alias with the vga variants. What would you call them? I've yet to see a credible alternative. Also the exact hw bits covered by the object varies from vendor to vendor. The names would get too long to be useful:

          display_timing_and_pll_and_pitch_and_fb_offset_and _a_bunch_of_other_stuff_on_most_chips_object

          Would that be any better? If you don't know what a crtc is you probably don't know what a pll is or display timing or pitch, etc. By your logic we'd need alternative names for those as well.

          If you want to learn GPUs you need to learn about the relevant language; it's just like any other field or area. Networking, video, etc. have nomenclature that is specific to those areas. They aren't designed to make things hard for newcomers; it makes the lives of those working in those fields easier as they have a common set of terms to describe them. Would you suggest chemists drop the use of chemical formulas because it's potentially too hard for newcomers?

          Comment


          • #35
            Hum, this will make seamless hibrid laptopts (like the ones with optimus) work, without performance loss like in bumblebee and ironhide projects because of virtualGL, right?

            Comment


            • #36
              re: re: re: re: naming

              Originally posted by agd5f View Post
              It's no more confusing than any other name. You have to have some understanding of the hardware to understand the code.
              exactly, and it would be a good start to refer to hardware parts by some meaningful names. if arbitrary names is your choice then you might as well call it a "Strawberry".

              Originally posted by agd5f View Post
              It goes back to vga which is a good starting point and on some chips the registers even alias with the vga variants. What would you call them? I've yet to see a credible alternative. Also the exact hw bits covered by the object varies from vendor to vendor. The names would get too long to be useful:

              display_timing_and_pll_and_pitch_and_fb_offset_and _a_bunch_of_other_stuff_on_most_chips_object
              as i understand, "CRTC" is the first thing in output pipeline that takes finished data for output preparation. so, if it must be a 'controller' of some sorts, what's wrong with "[video] output controller", again ?

              i don't know about registers but imitation and emulation must have their limits. i find it dubious in general to learn new things by old ways and concepts that may not have much in common with each other anymore, besides those deprecated imitation and emulation layers.

              Originally posted by agd5f View Post
              Would that be any better? If you don't know what a crtc is you probably don't know what a pll is or display timing or pitch, etc. By your logic we'd need alternative names for those as well.
              by my logic, if it doesn't say something about properties or nature of the subject already, then yeah, make it so. or name it by the dude who came up with it in the name of science and common sense !

              Originally posted by agd5f View Post
              If you want to learn GPUs you need to learn about the relevant language; it's just like any other field or area. Networking, video, etc. have nomenclature that is specific to those areas. They aren't designed to make things hard for newcomers; it makes the lives of those working in those fields easier as they have a common set of terms to describe them.
              you know, most popular "nomenclature" for every piece of machinery on most Russian factories can be loosely translated as "that fucked-up dickery". but a buddy of mine, chemical installation designer, is against putting it into designs and documentation even though he himself says that "it makes the lives of those working in those fields easier". pretty much his exact words from recently.

              yes, i know, i know, that one is not in official texts (yet?) but the one from topic is. it doesn't make that name any less meaningless and arbitrary in scope of the present.

              Originally posted by agd5f View Post
              Would you suggest chemists drop the use of chemical formulas because it's potentially too hard for newcomers?
              and what the hell formulas have to do with it ? 0_0 formulas are awesome, they pretty clearly show substances structure, especially structural formulas (who could have though of such a name ?) which just are schematic pictures of molecules. everything is pretty clear there.

              but in case you tried to hint at arbitrary names for some chemical elements - that's understandable, there is only so much ways you can describe something as simple as an element.
              or maybe you were talking about confusion between different types of formulas ? well, structural formula is the most precise one and everything else is just simplifications for calculations where that precision is not needed. and unnecessary pictures tend to take up space unnecessarily.

              PS: can you, please, elaborate on "relevant language" of arbitrary and "historical" naming/representation in Networking ? just for all-round education, as they say. i know i hate damn schematics where Ethernet segments drawn in bus topology even though switches are not goddamn hubs and don't work like that even in scope of a single segment, arrgh.

              Comment


              • #37
                Originally posted by rohcQaH View Post
                I'm thinking about starting a new X server on a virtual CRTC and having the results simply composited into your main X server (goodbye Xephyr kludges!)
                I have just pushed a quick-and-dirty example that shows one way to implement a Xephyr/Xnest-like functionality in a extremely simple way. Take a look:

                Virtual CRTC documentation and examples. Contribute to ihadzic/vcrtcm-doc development by creating an account on GitHub.

                Comment


                • #38
                  Oh, that code is a very roundabout way to answer my question: video data is copied from GPU into CPU memory, then uploaded back to the GPU again.

                  While it's not a high priority to remove those copies, my question was whether the virtual CRTC architecture would eventually allow doing so. For example by creating a VCRTC-driver that'll copy the framebuffer onto a second (userspace-provided) buffer which can be used as a texture.

                  Comment


                  • #39
                    This look nice.
                    As I understand it I can redirect the output to Intel encoder via VAAPI and then the h264 stream over the network.
                    This could work for playing games onthego on netbook.
                    Is this possible?

                    Comment

                    Working...
                    X