xpra icon
Bug tracker and wiki

This bug tracker and wiki are being discontinued
please use https://github.com/Xpra-org/xpra instead.


Changes between Version 2 and Version 3 of Ticket #2022, comment 43


Ignore:
Timestamp:
01/01/19 19:02:14 (2 years ago)
Author:
Antoine Martin
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #2022, comment 43

    v2 v3  
    1 >
    2 > I'm encountering two unrelated problems:
    3 >
    4 > 1. x264 and vpx are both laggy, but I think this might be an insurmountable computational intensity issue (on my the old Westmere chip I'm testing this on, at least).  I'm seeing other behaviors consistent with this theory: if I turn off the mesh, for instance, it gets less laggy (and the lag goes away sometimes).
    5 >
    6 > 2. When I set the video region by hand (after using it a little) XPRA ends up in nvenc, but when I set it right at start it is in x264 and stays in x264 forever, and it never switches to nvenc.
    7 >
    8 > My users are comparing xpra to VNC (and NX), and the only consistent complaint I have heard (from a small number of them) has to do with (at least I think it has to do with) the time it takes nvenc to start.
    9 >
    10 > This region that I am flagging with DBUS is either resting (a constant lossless frame) or animating.  Nothing like scrolling or UI happens in this region.  My experiments tell me that VPX and X264 cannot keep up, but nvenc is amazing and turbojpeg is the best when there is no nvenc (at least on old westmere chips).
    11 >
    12 > Based on my experience, therefore, and assuming I've made no other mistakes, I would like for my DBUS calls to make a *permanent* nvenc context in that region, as the shape of the region never changes without triggering my resize callback which will notify xpra.
    13 
    141I've done a little attempt at debugging and I've found that the score for nvenc never gets better than x264 when I set the region using dbus.  I think it is because the setup penalty always swamps everything else for some reason, but I'm not sure that I'm reading this right.
    152