Hi Philipp, Thanks for answering. 2014-11-26 17:55 GMT+01:00 Philipp Zabel <p.zabel@xxxxxxxxxxxxxx>: > Hi Jean-Michel, > > Am Mittwoch, den 26.11.2014, 14:33 +0100 schrieb Jean-Michel Hautbois: >> Hi, >> >> We are writing a gstreamer plugin to support CODA960 encoder on i.MX6, >> and it is not working so now trying to use v4l2-ctl for the moment. >> As I am asking about encoder, is there a way to make it support YUYV >> as input or is the firmware not able to do it ? I could not find a >> reference manual about that... > > The H.264 and MPEG-4 encoders support planar 4:2:0 subsampled formats > only: YU12, YV12, and chroma-interleaved NV12. > The JPEG encoder can also handle planar 4:2:2 subsampled frames, but > none of the interleaved (YUYV, UYVY, ...) variants. OK, just read this in TRM. This means that when the sensor is YUV 4:2:2 (say, a ADV76xx :)) it will have to be converted to NV12 in order to have it as input of the encoder... >> So back to the issue. >> $> cat /sys/class/video4linux/video0/name >> coda-encoder >> $> v4l2-ctl -d0 --list-formats >> ioctl: VIDIOC_ENUM_FMT >> Index : 0 >> Type : Video Capture >> Pixel Format: 'H264' (compressed) >> Name : H264 Encoded Stream >> >> Index : 1 >> Type : Video Capture >> Pixel Format: 'MPG4' (compressed) >> Name : MPEG4 Encoded Stream >> >> $> v4l2-ctl -d0 --list-formats-out >> ioctl: VIDIOC_ENUM_FMT >> Index : 0 >> Type : Video Output >> Pixel Format: 'YU12' >> Name : YUV 4:2:0 Planar, YCbCr >> >> Index : 1 >> Type : Video Output >> Pixel Format: 'YV12' >> Name : YUV 4:2:0 Planar, YCrCb > > Please apply all coda patches in the media-tree master branch first. > The output format list should include NV12 then. OK, applying right now. >> ==> First question, vid-cap should be related to the capture format, >> so YUV format in the encoder case, no ? >> >> $> v4l2-ctl -d0 --set-fmt-video-out=width=1280,height=720,pixelformat=YU12 >> $> v4l2-ctl -d0 --stream-mmap --stream-out-mmap --stream-to x.raw >> unsupported pixelformat >> VIDIOC_STREAMON: failed: Invalid argument > > On v3.18-rc6 with the coda patches currently in the pipeline applied, I > get this: > > $ v4l2-ctl -d0 --set-fmt-video-out=width=1280,height=720,pixelformat=YU12 > $ v4l2-ctl -d0 --stream-mmap --stream-out-mmap --stream-to x.raw > K>P>P>P>P>P>P>P>P>P>P>P>P>P>P>P>K>P>P>P>P>P>P>P>P>P>P>P>P>P>P>P>K>P>P>P>P>P>P 38 fps >> 38 fps This seems really better, yes :). Well, I come with another question : it seems to be limited in framerate throughput ? I get the same thing with the video capture, I can go up to 33 fps but no more. When I use --stream-mmap=8 I have only two buffers used. Is this the case for you in the encoder too ? JM -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
