[Ffmpeg-devel-irc] ffmpeg-devel.log.20171020

burek burek021 at gmail.com
Sat Oct 21 03:05:04 EEST 2017


[05:57:57 CEST] <rcombs> can someone add http://www.mpgedit.org/mpgedit/testdata/mpeg1/layer1/fl4.mp1 to fate samples
[05:58:08 CEST] <rcombs> (we don't have an mp1 encoder)
[05:59:36 CEST] <atomnuker> was that a dare?
[06:00:42 CEST] <rcombs> &it was not intended as one
[06:01:01 CEST] <rcombs> I'm not complaining; I can't think of a use for an MP1 encoder
[06:01:18 CEST] <rcombs> I just need fate to have a test sample for the demuxer and muxer
[06:03:12 CEST] <atomnuker> hey rcombs 
[06:03:44 CEST] <atomnuker> I WANT THE FLAC COVERART PATCH
[06:04:31 CEST] <atomnuker> I've been waiting months for the bloody thing, I want to make a cyanrip release but I can't because it wouldn't be feature complete without this
[20:21:25 CEST] <wbs> JEEB: jamrial: on the topic of hevc arm; there are some hevc idct optimizations further ahead in the merge log that you might want to merge before having others contribute more
[20:21:50 CEST] <JEEB> k, I just happened to notice the post and gave some generic hints
[20:22:43 CEST] <wbs> JEEB: jamrial: the stuff from libav unfortunately overlaps something that you already have, but I think I've benchmarked that the libav ones are equal or faster for all cases. some of the functions I have a feeling that I'd like to rewrite from scratch to see if I could get them better structured etc, but they should be pretty ok performance wise I think
[20:23:36 CEST] <jamrial> wbs: there's a lot of overlap in hevc asm, including x86
[20:23:57 CEST] <wbs> ah, yeah, those as well. I don't deal that much with x86 asm :P
[20:23:58 CEST] <jamrial> MC was written from scratch for libav instead of using the OpenHEVC patches, for example
[20:25:34 CEST] <jamrial> I took the SAO work OpenHEVC left unfinished, extended and optimized it a bit before merging it into ffmpeg
[20:25:58 CEST] <JEEB> -12
[20:26:05 CEST] <jamrial> tried to do the same for libav, but there was just too much previous stuff that hadn't been committed there
[20:26:10 CEST] <BBB> at some point Im hoping all_this will get people excited about merging
[20:26:20 CEST] <jamrial> like, hevc_filter.c looked like a different file :p
[20:26:20 CEST] <BBB> but I guess NotYetEnoughPain
[20:26:29 CEST] <BBB> -ENEEDMOREPAIN
[22:07:13 CEST] <atomnuker> durandal_1707: are you up for writing a filter which expands a video to a given aspect ratio and fills in the border with a zoomed in blurred part of the video?
[22:08:56 CEST] <J_Darnley> Like every news network nes it?
[22:09:10 CEST] <durandal_1707> atomnuker: use already available filters
[22:09:21 CEST] <durandal_1707> its doable
[22:13:58 CEST] <Compn> with geq ?
[22:14:05 CEST] <Compn> geqqqqqqqqq
[22:14:27 CEST] <ubitux> scale, blur, overlay
[22:16:28 CEST] <jamrial> michaelni: did you have time to look at the merge i mentioned wednesday?
[22:17:57 CEST] <ubitux> atomnuker: ffplay -f lavfi -i "movie=$HOME/samples/GoneNutty.avi,split[a][b]; [a]scale=iw+50:ih+50,boxblur=lr=5[blur]; [blur][b] overlay=x=25:y=25 [out]"
[22:18:11 CEST] <ubitux> you can probably use eval variables in overlay but you get the idea
[22:20:08 CEST] <atomnuker> I was thinking of using zoompan
[22:20:20 CEST] <atomnuker> but zoompan seems incorrectly designed
[22:20:27 CEST] <atomnuker> why does it modify the duration and the fps as well?
[22:20:33 CEST] <JEEB> huh? :D
[22:22:06 CEST] <atomnuker> well it does
[22:22:30 CEST] <atomnuker> and its broken too, output fps is definitely not 25 fps
[22:22:39 CEST] <atomnuker> (the default)
[22:22:54 CEST] <atomnuker> I think it should be removed
[22:28:12 CEST] <michaelni> jamrial, yes, i didnt find a case it breaks, iam undecided if i like the change
[22:28:45 CEST] <jamrial> anything wrong with it? usually, bytestream2 api is better than manually playing with pointers
[22:30:12 CEST] <michaelni> if it was written as a change to bytestream2 on top of our code it would be fine. Its the replacement of code that has no known issues by a different implementation that makes me feel a bit undecided
[22:32:43 CEST] <michaelni> thats no objection or anything like that
[22:32:52 CEST] <jamrial> i basically had to rewrite portions of it since it didn't really apply cleanly. i made sure it's functionally the same than pre commit, like keeping checks libav didn't have (or at least that was the attempt)
[22:33:13 CEST] <michaelni> ahh, then it sounds good, thanks
[22:33:42 CEST] <jamrial> worst case scenario we can revert it. we're too far from branching a new release after all
[22:34:33 CEST] <jamrial> and thanks again for testing
[22:38:08 CEST] <durandal_1707> atomnuker: if you dont want zoompan to change fps set right param to 1
[22:40:39 CEST] <durandal_1707> atomnuker: zoompan is designed primarly to zoom and pan slideshows, not movies
[22:42:33 CEST] <jamrial> michaelni: http://coverage.ffmpeg.org/libavcodec/h2645_parse.c.gcov.html if you happen to have samples that trigger the currently untested non-failure paths of the function to make fate tests then it would be great as well
[23:23:07 CEST] <durandal_1707> Compn: geq is michaelni gpl pet,  slow as hell
[23:25:52 CEST] <Compn> its not slow for people with quantom computers
[23:26:54 CEST] <durandal_1707> ah,  so you have one?
[23:43:26 CEST] <jamrial> durandal_1707: a lot of filters are slow as hell until you write some simd for them
[23:47:46 CEST] <durandal_1707> jamrial: really?, you should not buy potato next time
[23:53:51 CEST] <atomnuker> kierank: have you seen the new xps laptops? http://www.dell.com/en-uk/shop/laptop-and-2-in-1-pcs/xps-15/spd/xps-15-9560-laptop
[23:54:14 CEST] <JEEB> they upgraded them again? even the last time I checked recently they looked good
[23:54:21 CEST] <atomnuker> none have a 4k screen now, none but the most expensive one have 16gb of ram, 2 of them have a cheaper cpu
[23:54:26 CEST] <atomnuker> and they're more expensive
[23:54:33 CEST] <JEEB> wat
[23:54:33 CEST] <kierank> atomnuker: what does it do that the old ones doesn't
[23:54:41 CEST] <kierank> I have the xps15
[23:54:44 CEST] <kierank> With 4k
[23:54:49 CEST] <atomnuker> so a 9550?
[23:55:03 CEST] <durandal_1707> kierank: my is 8k
[23:55:04 CEST] <kierank> 9650 our 
[23:55:07 CEST] <kierank> Iirc
[23:55:22 CEST] <kierank> durandal_1707: meh, you live on a different that's well known
[23:55:25 CEST] <atomnuker> I guess they removed all 4k display options but incresed the price
[23:55:45 CEST] <JEEB> at least the 13" screen res seems same
[23:55:47 CEST] <JEEB> 3200x1800
[23:55:50 CEST] <kierank> 13 seems sane
[23:56:17 CEST] <JEEB> I really kind of want the xps 13 :/
[23:56:30 CEST] <atomnuker> too small for general use
[23:56:41 CEST] <JEEB> I've used 14 and 12 inchers and I've been OK
[23:56:56 CEST] <JEEB> unless they got the keyboard extremely small
[23:57:10 CEST] <JEEB> my coworker has one and it seemed alright
[23:58:46 CEST] <JEEB> oh, right 13 still hasn't got the 8th gen I series upgrade
[23:58:47 CEST] <JEEB> meh
[23:59:40 CEST] <jkqxz> Nice, you can pay 450GBP for the extra 16GB of the memory to get the top model.  Sounds like a good deal!
[00:00:00 CEST] --- Sat Oct 21 2017


More information about the Ffmpeg-devel-irc mailing list