[Ffmpeg-devel-irc] ffmpeg-devel.log.20190106

[burek]

[14:19:19 CET] <durandal_1707> atomnuker: i improved performance drastically even without SIMD
[19:14:09 CET] <cone-824> ffmpeg 03James Almer 07master:674b59c07515: avcodec/bsf_list: implement a AVBSFContext.flush callback
[19:14:09 CET] <cone-824> ffmpeg 03James Almer 07master:4ec2daf0de1a: configure: add dxva2 extralibs to avcodec
[19:34:43 CET] <cone-824> ffmpeg 03Michael Niedermayer 07master:96e495082aed: avcodec/huffyuvdec: Check that slice height is non negative.
[19:34:44 CET] <cone-824> ffmpeg 03Michael Niedermayer 07master:ab160efa2850: avcodec/huffyuvdec: Check that slices do not exceed frame height
[20:44:08 CET] <kurosu_> that vp4 loop filter rings a bell
[20:44:54 CET] <kurosu_> I think the vp6 decoder isn't bit exact (4CC VP6F iirc), and I suspected this and mc to be wrong
[20:55:24 CET] <atomnuker> BBB: why is the default trc for all=bt601 in vf_colorspace SMPTE170M instead of AVCOL_TRC_GAMMA28 which is "ITU-R BT470BG" according to the docs
[20:55:41 CET] <atomnuker> when all=bt601 sets everything else to bt470bg
[20:55:55 CET] <atomnuker> I just find it odd, the results come out looking correct
[20:58:28 CET] <BBB> uhm
[20:58:32 CET] <BBB> that's a long time ago
[20:58:34 CET] <BBB> I don't know?
[20:59:02 CET] <BBB> are they different?
[20:59:10 CET] <BBB> I recall some trcs having different names but the same values
[20:59:16 CET] <BBB> I don't know if that's true for these particular two
[20:59:47 CET] <atomnuker> AVCOL_PRI_SMPTE170M   = 6,  ///< also ITU-R BT601-6 525
[21:00:19 CET] <atomnuker> maybe what separates bt601 and bt470bg is just the trc, while everything else is identical
[21:03:35 CET] <atomnuker> I can believe vp8 was blind enough not to have seen bt709 as the future
[22:11:10 CET] <durandal_1707> atomnuker: do you want to comment anlmdn?
[22:15:07 CET] <jamrial> durandal_1707: why not split compute_distance off into its own dsp context now instead of in a following patch?
[22:15:29 CET] <atomnuker> durandal_1707: I'd like to test it first, do you have any noisy files?
[22:16:12 CET] <durandal_1707> atomnuker: what kind of audio engineer does not have noisy audio?
[22:17:31 CET] <durandal_1707> i have some librivox recordings, with different light to mid background noise
[22:18:05 CET] <atomnuker> durandal_1707: the kind who works on digital audio, where analog noise is rare?
[22:58:40 CET] <durandal_1707> atomnuker: get random librivox audio book and find file with enough of noise? other samples are more for voice extraction
[22:59:30 CET] <durandal_1707> i could upload single sample though
[23:03:43 CET] <durandal_1707> atomnuker: https://0x0.st/sRRb.mp3
[23:05:09 CET] <durandal_1707> defaul options values are for 8k sample rate, for 22050 rate multiply by 3 values r and p option default values
[23:06:00 CET] <durandal_1707> anlmdn=n=10:p=100:r=400:s=300 for example
[23:59:37 CET] <pross> kurosu_: vp56_mc looks very similar
[00:00:00 CET] --- Mon Jan  7 2019