[Spice-devel] Information about video detection

[Matilde Yanez]

Hi, I go on the website with another browser (chrome) but i don't think
that it can change the result and the qxl driver is installed on the VM.
I don't know how to get the size of the stream created, can you explain me?
The test that i made is: open the excel file, in full screen, and going up
an down quickly (same test on website). I put log on mjpeg_encoder and
stream functions, to observe this stream creation. It takes about 10
seconds to launch a stream detection.

Thanks,
Matilde


2013/7/12 Yonit Halperin <yhalperi at redhat.com>

> Hi,
>
> On 07/02/2013 10:47 AM, Matilde Yanez wrote:
>
>>
>>
>>
>> 2013/7/2 Yonit Halperin <yhalperi at redhat.com <mailto:yhalperi at redhat.com
>> >>
>>
>>     On 07/02/2013 10:22 AM, Matilde Yanez wrote:
>>
>>
>>         Hi,
>>         thanks for answers.
>>
>>         2013/7/2 Yonit Halperin <yhalperi at redhat.com
>>         <mailto:yhalperi at redhat.com> <mailto:yhalperi at redhat.com
>>
>>         <mailto:yhalperi at redhat.com>>>
>>
>>
>>              Hi,
>>
>>              On 07/01/2013 04:27 AM, Matilde Yanez wrote:
>>
>>                  Hello,
>>
>>                  I need some information about video detection, on VM.
>>                  When I am on web pages, or documents, with images the
>>         VM seems
>>                  to detect
>>                  video and start the mjpeg_encoder process.
>>                  Thus, the display is slows down.
>>
>>              What vm are you using? We encountered such behaviour with
>>              gnome-shell, due to high frequency updates of regions over
>> the
>>              primary surface.
>>
>>
>>         On CentOs and Windows VMs, indeed with high frequency updates of
>>         regions
>>         it considers that is video. Such as large Excel documents or web
>>         page
>>         with many images. Adding log I can see that spice server starts
>>         stream,
>>         and does the same treatment as video.
>>
>>     Have you used the default setting of streaming-video
>>     (streaming-video=filter), or did you use streaming-video=on? Can you
>>     send me an example document and the Microsoft Office version you are
>>     using?
>>
>>
>> Using the default setting of streaming-video, scroling on
>> "http://www.lemonde.fr/" for exemple.
>> The Microsoft version is 14.0.4730.1010 (Office Professional Plus 2010).
>>
> I couldn't reproduce your experience. I browsed the website (with IE), and
> the excel file you gave me, and didn't observe video streams that are being
> created. Is it possible that the qxl driver is not installed?
> What are the sizes of the video streams that are being created?
>
> Cheers,
> Yonit.
>
>>
>>
>>
>>                  With argument streaming-video=off, the CPU consomation
>>         goes up.
>>
>>              This is unexpected. On which guest scenarios?
>>
>>
>>         When I play a movie, with default "streaming-video" arguments, on
>>         windows VM, I am around 60% of CPU consomation of VM's qemu
>> process
>>         ("top" command), and with "streaming-video=off" I am around 120%.
>>         Due to all images that are send to refresh client display?
>>
>>     Ok, it is expected for movie playback. I thought you meant that it
>>     happens for the document browsing.
>>
>>                  Increasing the value of
>>         "RED_STREAM_FRAMES_START_____**CONDITION",
>>
>>
>>                  does not
>>                  launch the video detection on these documents and keep
>>         fluidity on
>>                  videos and documents
>>
>>                  How can i make difference between real video, and
>>         documents with
>>                  images?
>>
>>              I'm afraid there is no simple answer for this
>>              (1) Why does the driver send frequent updates? Does the
>>         rectangles
>>              that are being updated really change? Can it be prevented
>>         on the
>>              driver side?
>>              (2) If the area that is being refreshed doesn't change
>>         much, we can
>>              apply a quick comparison of sampled regions of 2 successive
>>              "frames", and if most of the pixels stay the same, we won't
>>         consider
>>              those "frames" as part of a video stream. (for such
>>         scenarios we
>>              would also like to improve and employ the GLZ encoding, or
>>         change
>>              our caching mechanism, to be applied on slices of bitmaps)
>>
>>
>>         I think I will investigate on the 2nd point.
>>
>>                     On what part of the spice-server code can i make
>>         changes?
>>
>>              Code that is related to video streams is located in
>>         red_worker.c, in
>>              routines with "stream|upgrade" in their names.
>>
>>
>>                  Thanks,
>>                  Matilde.
>>
>>              Regards,
>>              Yonit.
>>
>>
>>         Matilde.
>>
>>                  ______________________________**_____________________
>>                  Spice-devel mailing list
>>                  Spice-devel at lists.freedesktop.**____org
>>                  <mailto:Spice-devel at lists.__fr**eedesktop.org<http://freedesktop.org>
>>         <mailto:Spice-devel at lists.**freedesktop.org<Spice-devel at lists.freedesktop.org>
>> >>
>>         http://lists.freedesktop.org/_**___mailman/listinfo/spice-**devel<http://lists.freedesktop.org/____mailman/listinfo/spice-devel>
>>         <http://lists.freedesktop.org/**__mailman/listinfo/spice-devel<http://lists.freedesktop.org/__mailman/listinfo/spice-devel>
>> **>
>>
>>         <http://lists.freedesktop.org/**__mailman/listinfo/spice-devel<http://lists.freedesktop.org/__mailman/listinfo/spice-devel>
>>         <http://lists.freedesktop.org/**mailman/listinfo/spice-devel<http://lists.freedesktop.org/mailman/listinfo/spice-devel>
>> >>
>>
>>
>>
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/spice-devel/attachments/20130715/b8392c5d/attachment.html>