[Ffmpeg-devel-irc] ffmpeg-devel.log.20140108

burek burek021 at gmail.com
Thu Jan 9 02:05:02 CET 2014

[01:19] <llogan> j-b: can I get the "Read FFmpeg git howto" removed from http://git.videolan.org/ ?
[01:19] <llogan> ...or can one of us do it?
[01:27] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:4b0cecb457c7: avcodec/hevc_ps: check that VPS referenced from SPS exists
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:908b951b4e94: avcodec/hevc: Fix modulo operations
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:5e21989de4c8: Revert "Merge remote-tracking branch 'qatar/master'" (43dec5ef9a360c9ffac3278f464832bd99af0cb0)
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:c094aec76e29: avcodec/ac3dec: check bap before use.
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:413065aff4ef: avcodec/g2meet: check available space before copying palette
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:dcecca07585f: avutil/log: check that len is within the buffer before reading it
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:69aa3d5b88ec: avcodec/hevc: clear HEVClc when its deallocated in hevc_decode_free()
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:898ab02557b0: avformat/avidec: Dont assert the existence of an index for video streams.
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:d1a919586317: avcodec/mjpegdec: check len in mjpeg_decode_app() more completely
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:2864c987d8ce: avformat/ipmovie: check OPCODE_INIT_VIDEO_BUFFERS size more completely
[01:28] <cone-113> ffmpeg.git 03Michael Niedermayer 07release/2.1:580d3ef6fea3: avformat/mov: Check that we have a stream before accessing it in mov_read_ares()
[01:43] <cone-113> ffmpeg.git 03Carl Eugen Hoyos 07master:94cf4f8bac12: Use the h264 parser when decoding VSSH in avi.
[02:03] <ramiro> I'm trying to capture from the webcam to a video and img2 files at the same time. When img2 fails, ffmpeg exits without flushing the video. I'm writing a patch to make FFmpeg only close the stream that failed on av_interleaved_write_frame(), but I can't get past this: http://pastebin.com/c2JaaTKa
[02:04] <ramiro> after img2 fails, ffmpeg stop reading new frames from the webcam, but keeps on writing the last frame to the video.
[02:44] <BBB> ubitux: fixed your comment about code duplication
[05:05] <cone-362> ffmpeg.git 03Michael Niedermayer 07master:1c010fd035c1: avformat/mxfdec: detect loops during header parsing
[05:10] <michaelni> ramiro, just spotted your question ... if you havnt solved it by tomorrow, ping me and ill take a look (need to go to bed now)
[05:11] <michaelni> btw nicolas also should know that code
[05:20] <cone-362> ffmpeg.git 03Martin Storsjö 07master:67bb3a4e285a: arm: cosmetics: Reindent the h264dsp neon init function
[05:21] <cone-362> ffmpeg.git 03Michael Niedermayer 07master:9ef7c0c5510f: Merge commit '67bb3a4e285a5871770cbaa2d78bf9024961dd0f'
[05:36] <cone-362> ffmpeg.git 03Martin Storsjö 07master:b7b932f5e360: arm: Remove a leftover define for the pld instruction
[05:36] <cone-362> ffmpeg.git 03Michael Niedermayer 07master:3c00d4c5f0ae: Merge commit 'b7b932f5e3602bd34c3cc634b71c8bbbc0fb8dc0'
[05:44] <cone-362> ffmpeg.git 03Martin Storsjö 07master:5dae48723576: arm: Allow overriding the alignment set in the function macro
[05:44] <cone-362> ffmpeg.git 03Michael Niedermayer 07master:edba54630b6f: Merge commit '5dae4872357613a0b51120b54a4c5221e0ec3f69'
[05:51] <cone-362> ffmpeg.git 03Martin Storsjö 07master:952d3187d8ed: arm: Add a missing # as prefix for an immediate constant
[05:51] <cone-362> ffmpeg.git 03Michael Niedermayer 07master:8be8dddd13f4: Merge remote-tracking branch 'qatar/master'
[08:22] <ubitux> BBB: the branch LGTM, thx :)
[09:53] <cone-780> ffmpeg.git 03Stefano Sabatini 07master:a008deb17bec: doc/muxers/matroska: reformat documentation
[09:53] <cone-780> ffmpeg.git 03Stefano Sabatini 07master:e9edf001d415: doc/ffmpeg: remove documentation for removed vlang/slang options
[09:53] <cone-780> ffmpeg.git 03Helmut K. C. Tessarek 07master:db70f2d36c2d: doc/muxers/matroska: add information to the matroska metadata language entry
[10:12] <cone-780> ffmpeg.git 03Ramiro Polla 07master:6a99ecb84e6d: doc/muxers/image2: document strftime option
[10:21] <cone-780> ffmpeg.git 03Stefano Sabatini 07master:2358458a7587: doc/muxers/image2: apply several formatting changes
[10:28] <cone-780> ffmpeg.git 03Michael Niedermayer 07release/1.2:f974c54909dc: avformat/mxfdec: detect loops during header parsing
[10:28] <cone-780> ffmpeg.git 03Carl Eugen Hoyos 07release/1.2:6f7fd2f58956: Use the h264 parser when decoding VSSH in avi.
[10:28] <cone-780> ffmpeg.git 03Michael Niedermayer 07release/2.1:de14fe7e297f: avformat/mxfdec: detect loops during header parsing
[10:28] <cone-780> ffmpeg.git 03Carl Eugen Hoyos 07release/2.1:a48440d85737: Use the h264 parser when decoding VSSH in avi.
[11:41] <j-b> llogan: sure, why?
[11:43] <ubitux> because the file doesn't exist i guess
[11:44] <ubitux> (it's now a texi file @ http://git.videolan.org/?p=ffmpeg.git;a=blob;f=doc/git-howto.texi;hb=HEAD)
[11:45] <ubitux> it should link to http://ffmpeg.org/git-howto.html if possible
[11:45] <ubitux> i doubt anyone is actually following that link anyway
[12:59] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:5d4541007f1d: avcodec/huffyuvenc: fix -context 1 encoding
[12:59] <cone-780> ffmpeg.git 03Anton Khirnov 07master:02dd178d9de9: avisynth: More av_new_packet changes
[12:59] <cone-780> ffmpeg.git 03Stephen Hutchinson 07master:16ae337bd8d4: avisynth: Use av_packet_from_data instead of av_new_packet
[13:01] <BBB> michaelni: merge request?
[13:02] <BBB> michaelni: github/rbultje/ffmpeg/vp9-simd
[13:03] <BBB> ubitux: comment for you (if you didn't fix it already), cglobal name, args_regs, 15 should be 16
[13:03] <BBB> since you're using m15
[13:03] <ubitux> ah, yeah
[13:04] <ubitux> i'm trying to figure out the transpose trick with the "butterflies"
[13:09] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:e84d14df10d0: vp9/x86: idct_32x32_add_ssse3.
[13:09] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:37b001d14d34: vp9/x86: idct_32x32_add_ssse3 sub-16x16-idct.
[13:09] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:04a187fb2ae2: vp9/x86: idct_32x32_add_ssse3 sub-8x8-idct.
[13:09] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:5b0fc0781911: vp9: reset contextual caches on frame size change with mt enabled.
[13:09] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:024fac5cd4f7: vp9: make mv bounds 32bit.
[13:09] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:dcc9009e141c: Merge remote-tracking branch 'rbultje/vp9-simd'
[13:14] <ubitux> :)
[13:15] <ubitux> BBB: i missed that: libavcodec/x86/vp9itxfm.asm:1208: warning: (VP9_IDCT32_1D:309) redefining multi-line macro `STORE_2X2'
[13:16] <ubitux> it should be taken out that scope
[13:17] <BBB> hm right, or made local
[13:20] <BBB> this better? I made the macro local
[13:20] <BBB> (%% prefix)
[13:21] <BBB> oh crap it was merged
[13:21] <BBB> hm
[13:21] <BBB> let me fix that
[13:21] <ubitux> :D
[13:23] <BBB> michaelni: patch on ml (an thanks for merge)
[13:23] <ubitux> what's this %%?
[13:23] <ubitux> i assume it's to make it local, but how does it work?
[13:23] <ubitux> does it defines a unique macro name each time?
[13:23] <ubitux> or it's just syntaxic sugar?
[13:29] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:9e7b477c8c75: avcodec/ffv1enc: choose 1.2 by default for higher resolutions to ensure multi-threaded decodability
[13:34] <BBB> ubitux: I don't know ;]
[13:34] <BBB> ubitux: it's a trick I learned from pengvado, but that's all I know
[13:34] <BBB> it's "local within the scope of its own containing macro"
[13:34] <BBB> it's documented in the yasm manual also
[13:35] <BBB> and nasm, I believe
[13:35] <ubitux> ok :)
[13:35] <BBB> oh I also have some ongoing work for a 16x16 iadst various functions
[13:35] <BBB> I can show it off if you want, but it's not terribly useful yet
[13:36] <ubitux> no hurry :p
[13:36] <ubitux> i'm still busy with the transpose :)
[13:37] <BBB> yeah they are funny
[13:38] <BBB> """""funny""""""
[13:39] <ubitux> for now i'm just trying to guess the sequence based on the different pattern
[13:39] <ubitux> and looking at the resulting matrix
[13:39] <ubitux> :p
[13:41] <ubitux> BBB: loopfilter was definitely fun ;)
[13:54] <ubitux> ok it seems it get it :)
[14:02] <BBB> what get what?
[14:04] <BBB> ubitux: probably just look at the 8x8w transpose for x86-32
[14:04] <BBB> a
[14:04] <BBB> and try to reproduce what it does to solve the register shortage for x86-64 16x16b
[14:04] <BBB> is what I would do, but up to you ofc
[14:04] <ubitux> yes that's exactly what i did
[14:05] <ubitux> and it works, just needs to do the swap
[14:05] <ubitux> and actually move it to asm (i wrote a separated PoC)
[14:06] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:9d098ea32fde: avcodec/huffyuvenc: Factorize store_huffman_tables() out
[14:19] <BBB> lol
[14:19] <BBB> ok
[14:19] <BBB> michaelni: can you merge that patch I posted on the ml? removes a compiler warning
[14:19] <BBB> (assembler warning)
[14:20] <BBB> ubitux: don't forget there's still all the other lf functions you need to do - they're easy b/c you did the big one, but still, don't forget :)
[14:20] <ubitux> heh, i won't
[14:20] <BBB> \o/
[14:21] <ubitux> but i'm assuming v and h 16x16 will be the largest improvement
[14:21] <BBB> yes absolutely
[14:22] <BBB> 8/4 still beat iadst by a long shot
[14:22] <BBB> 5.7% and 4.9% (h 8 and v 8) vs. 1.0, 0.5 and 0.7% for the iadst/idct variants
[14:23] <BBB> and 1.6% and 1.5% for v4 and h4
[14:25] <BBB> bbl - work
[14:27] <ubitux> ok, got the "butterfly timeline" and swaps working
[14:28] <ubitux> now i need to deal with the temporary spills thing
[14:36] <cone-780> ffmpeg.git 03Ronald S. Bultje 07master:c6fe984f2ff0: vp9/x86: make STORE_2X2 macro local.
[15:32] <ubitux> aah finally a dvb sub fix that looks sane
[17:23] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:c8c7736c1025: swscale: disable ARM code until its build failure with clang/iphone is fixed
[17:35] <michaelni> j-b, disabled that clang/iphone breaking arm asm. Its author was busy till now but he will look into it when he has time
[19:09] <j-b> michaelni: thx, noted.
[19:09] <j-b> michaelni: I don't plan to do a iOS release soon anyway ;)
[19:49] <llogan> j-b: what ubitux said. linking to http://ffmpeg.org/git-howto.html is fine with me
[23:56] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:1a0370ad94de: avcodec/kgv1dec: remove unused function and variable
[23:57] <cone-780> ffmpeg.git 03Michael Niedermayer 07master:48ad1f392a0f: avcodec/libstagefright: drop dependancy on sizeof(AVFrame)
[23:57] <cone-780> ffmpeg.git 03rogerdpack 07master:2be1e1b223d9: avformat/udp: debug log udp socket size
[00:00] --- Thu Jan  9 2014

More information about the Ffmpeg-devel-irc mailing list