| 12 Apr 2024 |
 xytovl | But for quality that works better than upscaling on the whole image | 19:13:03 |
 imrsvjon_damaestro | but we still have to decode them on the HMD gpu? or does that happen before decoding? | 19:14:03 |
| xytovl | What we decode is foveated | 19:16:36 |
| xytovl | Then during reprojection it gets un-distorted to full resolution (1.4x full resolution actually) | 19:17:10 |
| imrsvjon_damaestro | okay, makes sense | 19:19:17 |
| xytovl | I think that a fancy upscaler would need the undistorted image, before reprojection, so that's an additional step in the pipeline, we already waste some times if we split defoveation and reprojection. I don't think there is a performance gain, unless the goal is to render at lower resolution on PC side | 19:22:09 |
| xytovl | And this assumes that you don't need additional data such as depth or motion vectors | 19:22:25 |
 zappar_simon | Frame extrapolation is perhaps more useful on the client - sounds like that's what Virtual Desktop uses:
https://www.qualcomm.com/developer/blog/2022/09/virtual-boost-vr-rendering-performance-synchronous-space-warp | 21:22:25 |
| zappar_simon | * Frame extrapolation is perhaps more useful on the client - sounds like that's what Virtual Desktop uses after some input from Qualcomm:
https://www.qualcomm.com/developer/blog/2022/09/virtual-boost-vr-rendering-performance-synchronous-space-warp | 21:22:52 |
|  godiews | 22:16:49 |
|  theassclapper joined the room. | 22:51:20 |
| 13 Apr 2024 |
 marioeatsdirt | Frame extrapolation is also more resilient to frame drops. | 03:36:30 |
|  stu4055 joined the room. | 16:15:32 |
| imrsvjon_damaestro | yeah, that technique looks great | 18:37:09 |
|  evks79375 joined the room. | 21:10:12 |
| imrsvjon_damaestro | sweet, the lobby's new location/position (in the VR space) is alot more useful for being able to use hand controls for navigating the menues and connecting to a server. i had kept poking my monitor before making it to the button on the screen. | 23:51:19 |
| imrsvjon_damaestro | sweet, the casting when in the lobby is working now too 🙂 | 23:55:59 |
| imrsvjon_damaestro | i confirmed that we need to move the vulkan layer location. i'll update the bug report soon. | 23:57:20 |
| 14 Apr 2024 |
| imrsvjon_damaestro | i also went ahead and changed building with x264 to be optional (default off) in the fedora package, that seems to leave only the downloaded monado source to deal with and i'll work on dealing with that soon | 03:37:54 |
| imrsvjon_damaestro | * i also went ahead and changed building with x264 to be optional (default off) in the fedora package, that seems to leave only the downloaded monado source to deal with and i'll that soon | 03:38:26 |
| imrsvjon_damaestro | k, quickly looked at this. xytovl we need to map exact upstream releases of boostpfr and monado that i can bring in vs the server-build-deps source | 03:49:37 |
| imrsvjon_damaestro | i'll dig through the cmake stuff now to see if it's clear or not | 03:50:30 |
| imrsvjon_damaestro | * i'll dig through the cmake stuff now to see if it's clear or not, but we can't use server-build-deps | 03:51:19 |
| imrsvjon_damaestro | looks like it's all there | 03:56:00 |
| imrsvjon_damaestro | i'll have to bundle or package everything, but now have a worklist for this | 04:01:45 |
| imrsvjon_damaestro | * i also went ahead and changed building with x264 to be optional (default off) in the fedora package, that seems to leave only the downloaded monado source to deal with and i'll work on that soon | 04:09:11 |
| imrsvjon_damaestro | * looks like it's all there (noted that there is a patch for monado, i'll pull that in also) | 04:10:12 |
| imrsvjon_damaestro | * looks like it's all there (noted that there is a patch for monado, i'll pull that in also; do you have an upstream issue filed for this specific patch?) | 04:10:38 |
| imrsvjon_damaestro | * k, quickly looked at this. xytovl we need to map exact upstream releases of boostpfr and monado so that i can bring in vs the server-build-deps source | 04:11:00 |
| imrsvjon_damaestro | * k, quickly looked at this. xytovl we need to map exact upstream releases of boostpfr and monado so that i can bring them invs the server-build-deps source | 04:11:23 |
