On 2016/8/17 2:33, Mark Thompson wrote: > On 16/08/16 09:51, Jun Zhao wrote: >> >> barry@barry:~/Source/video/yami/ffmpeg_libyami$ ./ffmpeg -y -vaapi_device >> /dev/dri/card0 -hwaccel vaapi -hwaccel_output_format vaapi -i >> ../ffmpeg_yami_testcase/skyfall2-trailer.mp4 -an -vf >> 'format=nv12|vaapi,hwupload' -c:v h264_vaapi -profile 77 -level:v 40 -b >> 4000k output_vaapi_transcode.mp4 >> ... >> barry@barry:~/Source/video/yami/ffmpeg_libyami$ mediainfo >> output_vaapi_transcode.mp4 >> File size : 69.8 MiB >> ... >> barry@barry:~/Source/video/yami/ffmpeg_libyami$ ./ffmpeg -y -c:v >> libyami_h264 -i ../ffmpeg_yami_testcase/skyfall2-trailer.mp4 -c:v >> libyami_h264 output_yami_transcode.mp4 >> ... >> barry@barry:~/Source/video/yami/ffmpeg_libyami$ mediainfo >> output_yami_transcode.mp4 >> File size : 74.2 MiB > > I'm assuming you are trying to show them with identical options? Since the > hardware is the same, you really should be able to get those two encodes to > produce pretty much identical results. > > Here I think the significant difference is probably that h264_vaapi is using > 2 B-frames by default, but there might be more subtle differences to remove > as well. > > - Mark
Hi, Mark: I just used this show how to run ffmpeg/vaapi and ffmpeg/libyami :) For the performance gap, I think the root cause is that ffmpeg/vaapi transcode use VPP in the pipeline, but ffmpeg/libyami transcode without VPP. I will double-check this case. > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
