12:30 <memeka> mszyprow^: saw your v2 dmabuf MFC patch, will try when I come back from weekend holiday :)

12:30 <mszyprow^> memeka: it won't work with current gst

12:31 <mszyprow^> memeka: I've checked gst queues different buffers each time

12:32 <mszyprow^> memeka: MFC requires not to change buffer addresses during its operation

12:32 <mszyprow^> memeka: so a fix in gst is needed

12:32 <mszyprow^> memeka: it will also improve performance, as there is really no point to setup buffers every frame

12:33 <mszyprow^> it should be easy to fix this, but I didn't look into the gst code yet

12:34 <mszyprow^> the other thing that can be improved in v4l2 support in gst is to ensure that the same number of buffers are used between v4l2 elements

12:36 <mszyprow^> so when mfc use 8 capture (destination) buffers, the next v4l2 element (video convert) should also use 8 output (source) buffers

12:37 <mszyprow^> this way it is easy to have constant assignments of the buffers

12:37 <mszyprow^> the true zero-copy once everything is initialized

12:37 <memeka> In ffmpeg you can set them

12:38 <memeka> But ffmpeg doesn’t have zero copy support yet

12:38 <memeka> I’ll try to follow up on gst

12:39 <memeka> Work was quite busy last days tho, didn’t get to test much

12:39 <mszyprow^> memeka: no problem

12:39 <memeka> But I’m excited to see 4.14 with LTS, and a few extra patches to make everything work :)

12:41 <mszyprow^> well, too bad that I didn't manage to take a look into all this one kernel release earlier - then v4.14 would already have those fixes...

12:42 <memeka> Late is better than never