[osg-users] FFmpegImageStream thread safety

[OpenSceneGraph Users]

Sorry, I don't know off hand.  I didn't work much on the code myself.

Rob

On 8/10/20 11:38 PM, Ben Meijering wrote:
> Hi Rob,
>
> Thanks for the link.
> In this version I also see that m_pImgStream->setImage is invoked 
> in FFmpegRenderThread::run(), another thread.
> 2015 is a long time ago, but do you perhaps remember why this is safe ?
>
> Thanks in advance.
>
> Cheers,
>
> Ben
>
> Op ma 10 aug. 2020 om 23:03 schreef Rob Spearman 
> <rob at digitaliseducation.com <mailto:rob at digitaliseducation.com>>:
>
>     We made changes back in 2015 to fix threading issues with the osg
>     ffmpeg plugin.  I don't think these changes ever made it into OSG.
>
>     Source repo: https://bitbucket.org/digitalis/osg-ffmpeg-plugin/
>
>     On Monday, August 10, 2020 at 3:46:32 AM UTC-7 bjs.me... at gmail.com
>     <mailto:bjs.me... at gmail.com> wrote:
>
>         Hi Robert,
>
>         Thanks for your response.
>
>         If FFmpegImageStream::publishNewFrame is invoked from a
>         separate thread.
>         Is it then somehow guaranteed that (for example)
>         Texture2D::apply is not using/reading from the image object,
>         while FFmpegImageStream::publishNewFrame is modifying it ?
>
>         Cheers,
>
>         Ben
>
>         Op vr 31 jul. 2020 om 21:24 schreef Robert Osfield
>         <robert.... at gmail.com>:
>
>             Hi Ben,
>
>             I'm away from my dev computer so just commenting off the
>             top of my. FFmpegImageStream should be double buffered so
>             that the ffmpeg thread writing to the image will be
>             writing to one part of the data, while the other half is
>             available to be read by the graphics thread.  This should
>             avoid threading issues.
>
>             The threading used by the ffmpeg plugin is separate from
>             the viewer threading so settings related to the viewer
>             won't be important, nor will general settings of the scene
>             graph.
>
>             You should be able to just add the ffmpeg genearted
>             imagestream to a texture, start the steam going and then
>             largely forget about it.
>
>             Robert.
>
>             On Friday, 31 July 2020 at 20:15:45 UTC+1
>             bjs.me... at gmail.com wrote:
>
>                 Hi,
>
>                 I was just looking at FFmpegImageStream, I am not very
>                 familiar with this code, but I have some questions.
>
>                 It is not immediately clear to me how
>                 FFmpegImageStream::publishNewFrame is thread safe.
>                 It seems like the image data is set (setImage) from
>                 the video decoder thread.
>                 The image then uses a pointer to one of the buffers of
>                 the video decoder (of which the contents might also
>                 change ?).
>
>                 FFmpegImageStream also doesn't seem to override
>                 requiresUpdateCall, which I believe will result in the
>                 texture not being dynamic (Texture2D::setImage).
>                 Which could be used, for example in
>                 StateSet::computeDataVariance(), to determine whether
>                 the StateSet should be dynamic (which is needed for
>                 multithreaded rendering).
>
>                 If anyone could shed more light on this subject, it
>                 would be very much appreciated.
>
>                 Thank you.
>
>                 Cheers,
>
>                 Ben
>

-- 
You received this message because you are subscribed to the Google Groups "OpenSceneGraph Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to osg-users+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/osg-users/6e750bea-6962-0f6c-b161-7e98cbd5c0ce%40digitaliseducation.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openscenegraph.org/pipermail/osg-users-openscenegraph.org/attachments/20200811/66205e01/attachment.html>