[FFmpeg-devel] [RFC] Regression test speed

[Reimar Döffinger]

On Sat, Dec 05, 2009 at 09:53:54PM +0100, Stefano Sabatini wrote:
> On date Saturday 2009-12-05 20:04:26 +0100, Reimar D?ffinger encoded:
> > On Sat, Dec 05, 2009 at 07:26:46PM +0100, Stefano Sabatini wrote:
> > > Is OK to apply?
> > 
> > IMO yes, if I had know it works so easily I would have added -vframes right from the start.
> > Could you try if it is possible to add a regression test for RoQ like this?
> > Or is it too slow even when just encoding 1 or 2 frames?
> 
> With -vframes 5:
> 
> 23.58user 0.28system 0:24.19elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+40368outputs (0major+38065minor)pagefaults 0swaps

Seems acceptable to me.

> Index: ffmpeg/tests/codec-regression.sh
> ===================================================================
> --- ffmpeg.orig/tests/codec-regression.sh	2009-12-05 21:17:11.000000000 +0100
> +++ ffmpeg/tests/codec-regression.sh	2009-12-05 21:52:49.000000000 +0100
> @@ -241,6 +241,11 @@
>  do_video_decoding "" "-pix_fmt yuv420p -sws_flags area+accurate_rnd+bitexact"
>  fi
>  
> +if [ -n "$do_roqvideo" ] ; then
> +do_video_encoding roqvideo.roq "" "-vcodec roqvideo -vframes 5 -an"

I think you should leave the audio enabled, and -vcodec should not be necessary...