Re: Vino performance question for an IBM T41 (Radeon M10)

[Mark McLoughlin <markmc redhat com>]

On Wed, 2005-03-30 at 17:10 -0500, Martin Weinberg wrote:
> Folks,
> 
> I've been experimenting with vino and have found the following odd
> behavior with my dual-head laptop (IBM T41p running Debian testing). 
> 
> I have my second card/screen configured at 1024x768 for video
> presentations since the main LCD is 1400x1050 which most projectors
> can't handle.  So I start acroread or whatever on the second screen.
> 
> To see the projected screen on the laptop monitor, one can use VNC in
> one of the following three ways:
> 
> 1) Run a vncserver on the second screen (:0.1) and attach a vncviewer
>    to this.
> 
> 2) Run x11vnc on the second screen and attach a vncviewer
>    to this.
> 
> 3) Install vino and run a viewer on localhost:1.
> 
> All three of these work.  But I find that with vino, the cpu usage is
> pegged at 100% with the xserver taking up the lion's share of the
> cycles.  Whereas with the other two the cpu usage is very low.  On the
> other hand, the vino response seems better than the other two options.
> 
> I will also point out that it does not have to do with the second
> screen directly: if I attach remotely to the laptop on the main
> screen the same thing happens.
> 
> This does not happen with other desktops (with different xserver
> drivers) so I guess this is somehow XFree86 drv_ati related.
> Does anyone know what is going (and what I can do to investigate)?
> No additional log message anywhere during the vino server usage.

	Sounds like a bug ..

	A debug log might help to figure out what's going on. I think the best
way to get that is:

  1) gconftool-2 -s /desktop/gnome/remote_access/enabled -t bool false

  2) killall vino-server

  3) VINO_SERVER_DEBUG=1 /usr/libexec/vino-server

  4) gconftool-2 -s /desktop/gnome/remote_access/enabled -t bool true

	and then connect.

Cheers,
Mark.