Xpra: Ticket #1316: stereo display support
For VR, etc..
This is likely to be tied to how each graphics drivers implement stereoscopic support.
I don't see how this can be implemented or emulated in the dummy driver in a meaningful way using the software opengl renderer. We could use two virtual dummy screens, but forcing applications to use separate screens for each eye is likely to be very difficult, even with a virtualgl-like interposer.
This may require us to run in desktop mode with multiple X11 screens and forward each one to the client.
Then the client code probably just use a PFD_STEREO
context and render each eye with glDrawBuffer(BACK_LEFT)
/ glDrawBuffer(BACK_RIGHT)
. See the opengl example in NVIDIA 3D VISION PRO AND STEREOSCOPIC 3D. (we don't need to worry about fulcrum and parallax)
See nvidia driver xconfigoptions.
Other links:
Tue, 20 Sep 2016 11:53:26 GMT - Antoine Martin: status changed
- status
changed from new to assigned
Referring to the nvidia stereo options values:
- 3 "onboard stereo" - not possible to forward (relies on a physical proprietary DIN plug)
- 4 "One-eye-per-display passive stereo": can be captured and forwarded
- 5 "Vertical interlaced stereo mode, for use with SeeReal? Stereo Digital Flat Panels.": could be made to work but probably just a variant of 4, and 4 is easier
- 6 "Color interleaved stereo mode, for use with Sharp3D Stereo Digital Flat Panels." - same as 5
- 7 "Horizontal interlaced stereo mode, for use with Arisawa, Hyundai, Zalman, Pavione, and Miracube Digital Flat Panels." - same as 5
- 8 "Checkerboard pattern stereo mode, for use with 3D DLP Display Devices." same as 5
- 9 "Inverse checkerboard pattern stereo mode, for use with 3D DLP Display Devices." - same as 5
- 10 "NVIDIA 3D Vision mode for use with NVIDIA 3D Vision glasses." - probably not possible (USB device)
- 11 "NVIDIA 3D VisionPro? mode for use with NVIDIA 3D VisionPro? glasses" - worse than 10
- 12 "HDMI 3D": unlikely to work
- 13 "Tridelity SL stereo mode, for use with Tridelity SL display devices": unlikely to work
So the only mode that we're likely to be able to forward is (4). Even if some of the other ones could be captured somehow (5, 6, 7, 8, 9), handling the data would be complicated and require extra work for no benefit.
What is used client-side for actually displaying the two views we have captured should not be a concern and we should even be able to use all 3D capable GPUs via the opengl client.
Tue, 20 Sep 2016 13:10:16 GMT - Antoine Martin: description changed
- description
modified (diff)
Tue, 20 Sep 2016 13:15:07 GMT - Antoine Martin: description changed
- description
modified (diff)
Mon, 06 Mar 2017 09:37:06 GMT - urzds: cc set
- cc
dennis.schridde@… added
Wed, 20 Mar 2019 05:06:15 GMT - Antoine Martin: milestone changed
- milestone
changed from 3.0 to 4.0
Milestone renamed
Sat, 21 Sep 2019 12:08:35 GMT - Antoine Martin: milestone changed
- milestone
changed from 4.0 to 4.1
Sun, 04 Oct 2020 12:58:45 GMT - Antoine Martin: milestone changed
- milestone
changed from 4.1 to future
Re-scheduling due to lack of interest.
Sat, 23 Jan 2021 05:20:56 GMT - migration script:
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/1316