On 7/10/19, Eoff, Ullysses A <ullysses.a.e...@intel.com> wrote: >> -----Original Message----- >> From: Eoff, Ullysses A >> Sent: Tuesday, July 09, 2019 6:45 PM >> To: ffmpeg-devel@ffmpeg.org >> Cc: Eoff, Ullysses A <ullysses.a.e...@intel.com> >> Subject: [PATCH] avfilter: add rawdump filter >> >> This filter enables raw frames to be dumped to a >> file before they are sent through the auto-inserted >> scaler filter and useful when you want unscaled >> raw frames in an output file. >> > > The primary goal that I'm trying to achieve here is > dump raw "unscaled" yuv to an output file while > decoding a multi-resolution video. Unfortunately, > the ffmpeg utility auto-inserts a scaler into the filter > graph, so -f rawvideo doesn't suffice here since the > decoded data passes through the scaler before it's > handed off to the rawvideo filter. > > I experimented with the alternative methods suggested > in earlier feedback. More specifically, "reinit_filter" > and "split" were suggested. > > For "split" filter, I couldn't find a way that it could bypass > the auto-inserted scaler. > > Although -reinit_filter 0 works in some cases, it is quite > verbose. And I have found many other cases where it > doesn't work when we need to specify additional filters > (e.g. hwupload, hwdownload when doing qsv or vaapi > accelerated decoding with hw output formats). > > So, afaict, this patch is the only reliable method I can > find to intercept the decoded data before it reaches > the scaler for various use-cases (although the patch still > needs a little bit of work as pointed out in previous > replies). > > For added context, my organizations team and I contribute > to both msdk/qsv and vaapi plugins for ffmpeg and > gstreamer development. The idea is we want a way to > validate that direct decoder output is an exact bitmatch > across every decoder solution that we contribute > to (to provide a better and more consistent end-user > experience regardless of end-user choice). But since > ffmpeg auto-inserts this "scaler" and not all scaling > algo's are equal, it poses a challenge to verify.
You can use this in your own branch. I see no valid usecase except in this testing scenario. > > Regards, > U. Artie > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
