[Ffmpeg-devel-irc] ffmpeg-devel.log.20190220

burek burek021 at gmail.com
Thu Feb 21 03:05:03 EET 2019


[15:43:54 CET] Action: durandal_1707 wrote >3k lines of non-working code
[15:53:13 CET] <atomnuker> send it to an online'd public printer
[16:02:26 CET] <JEEB> :D
[16:05:34 CET] <atomnuker> wow, the bitstream seems uber-corporate-y
[16:05:53 CET] <atomnuker> all these descriptions no one is ever going to use
[16:06:23 CET] <JEEB> yea, it's got a crapload of stuff
[16:06:55 CET] <atomnuker> quantization looks pretty much like atrac9, but is segmented like aac (sce, cpe elements)
[16:07:19 CET] <atomnuker> I was expecting something to measure up to opus, not something which looks like it would sound worse than ac3
[16:21:34 CET] <durandal_1707> atomnuker: you did blind test on atrac9 ?
[16:27:50 CET] <atomnuker> no, just looking at the code
[16:36:33 CET] <philipl> BtbN: So, I was provoked to do more digging yesterday about the implications of compiling cuda kernels with the sdk vs licencing.
[16:37:32 CET] <philipl> The first thing I realised is that when we do ptx compilations, it's not compiling to a binary form, but a virtual assembly format. And then that is compiled to binary at runtime to match the physical hardware.
[16:37:47 CET] <philipl> So there's nothing linked in during the ffmpeg build.
[16:39:17 CET] <philipl> So if the filter code is written against the dynamic loader, then the ffmpeg binary code is still lgpl compatible, and the ptx files, which are aggregated and not linked, don't include anything from the sdk. The compiler was used and that's it. It's the same as compiling with vc++.
[16:39:23 CET] <philipl> Do you buy that?
[17:58:41 CET] <cone-842> ffmpeg 03Jan Sebechlebsky 07master:87d5686151bc: avcodec/bsf: Add ff_bsf_get_packet_ref() function
[17:58:42 CET] <cone-842> ffmpeg 03Nikolas Bowe 07master:aaf7fd168050: avcodec/extract_extradata_bsf: Fix leak discovered via fuzzing
[17:58:43 CET] <cone-842> ffmpeg 03James Almer 07master:be65995d23c9: avcodec/extract_extradata: Zero-initialize the padding bytes in all allocated buffers
[17:58:44 CET] <cone-842> ffmpeg 03James Almer 07master:70f11ee9981f: avcodec/extract_extradata: Do not allocate more space than needed when removing NALUs in h264/hevc
[17:58:45 CET] <cone-842> ffmpeg 03James Almer 07master:6442d8bad680: avcodec/extract_extradata: Move the reference in the bsf internal buffer
[17:58:46 CET] <cone-842> ffmpeg 03James Almer 07master:3365e91b1ed8: avcodec/extract_extradata: don't uninitialize the H2645Packet on every processed packet
[17:58:47 CET] <cone-842> ffmpeg 03James Almer 07master:97c9a5084479: avcodec/libaomenc: remove AVOption related to frame partitions
[17:58:48 CET] <cone-842> ffmpeg 03James Almer 07master:04e8b8b0530e: avcodec/libaomenc: export the Sequence Header OBU as extradata
[17:58:49 CET] <cone-842> ffmpeg 03Martin Storsjö 07master:5584abf69d83: arm: Emit .thumb_func directives
[17:58:50 CET] <cone-842> ffmpeg 03Zhong Li 07master:4c5e77e0bf14: lavc/qsvenc_jpeg: add async_depth support
[17:58:51 CET] <cone-842> ffmpeg 03Zhong Li 07master:1ff6cb2ca665: lavc/qsvenc_jpeg: set a default quality
[17:58:52 CET] <cone-842> ffmpeg 03James Almer 07master:493240a522fc: Merge commit '97c9a5084479eeb66f4beb100cc7589a2c8bfe81'
[17:58:53 CET] <cone-842> ffmpeg 03James Almer 07master:a024c3ce9a50: Merge commit '04e8b8b0530e2aa33010faba3d0b6b6c9c5b704e'
[17:58:54 CET] <cone-842> ffmpeg 03James Almer 07master:22b6e1011caa: doc/libav-merge: add a line about the previous skipped merge
[17:58:55 CET] <cone-842> ffmpeg 03James Almer 07master:b2f32d60eeaf: Merge commit '5584abf69d83169a010aca404cd1cf95c23ad9ef'
[17:58:56 CET] <cone-842> ffmpeg 03James Almer 07master:bec3b2041dcc: Merge commit '1ff6cb2ca6652e7d2a929afd33d8ed6268c45568'
[18:08:14 CET] <cone-842> ffmpeg 03James Almer 07master:cacb62f9cbf3: Revert "decode: copy the output parameters from the last bsf in the chain back to the AVCodecContext"
[18:08:15 CET] <cone-842> ffmpeg 03James Almer 07master:8d80046a0ff7: libaom: remove references to yuva444p pixfmt
[18:08:16 CET] <cone-842> ffmpeg 03Martin Storsjö 07master:a3a501df240f: libavutil: Undeprecate the AVFrame reordered_opaque field
[18:08:18 CET] <cone-842> ffmpeg 03Martin Storsjö 07master:80f85a95da29: libx264: Pass the reordered_opaque field through the encoder
[18:08:18 CET] <cone-842> ffmpeg 03Carl Eugen Hoyos 07master:ee3f62a90caa: pixfmt: Add GRAY10
[18:08:19 CET] <cone-842> ffmpeg 03Carl Eugen Hoyos 07master:f149a4a5fc5a: swscale: Add GRAY10
[18:08:20 CET] <cone-842> ffmpeg 03James Almer 07master:9bf9358b615d: avcodec: libdav1d AV1 decoder wrapper.
[18:08:21 CET] <cone-842> ffmpeg 03Linjie Fu 07master:e716323fa852: qsvenc: Add VDENC support for H264 and HEVC
[18:08:22 CET] <cone-842> ffmpeg 03James Almer 07master:56f50183f380: libdav1d: fix build after a recent API break
[18:08:23 CET] <cone-842> ffmpeg 03James Almer 07master:70ab2778be9c: libdav1d: update API usage to the first stable release
[18:08:24 CET] <cone-842> ffmpeg 03James Almer 07master:6fc762b4fd2c: Merge commit '70ab2778be9c83dab84340af7e3ba83fa0f98576'
[18:08:55 CET] <JEEB> gg
[18:12:27 CET] <jamrial> bunch of noops for now. the "fun" comes later
[18:49:57 CET] <cone-842> ffmpeg 03Luca Barbato 07master:90adbf4abf33: cook: Use the correct table for 6-bit stereo coupling
[18:49:58 CET] <cone-842> ffmpeg 03James Almer 07master:ca44fa5d7fda: avcodec/libdav1d: properly free all output picture references
[18:49:59 CET] <cone-842> ffmpeg 03Martin Storsjö 07master:41cf3e3b1ca3: arm: Create proper .rdata sections for COFF
[18:50:00 CET] <cone-842> ffmpeg 03James Almer 07master:0c126431f9b2: Merge commit '90adbf4abf336f8042aecdf1e18fdf76a96304b1'
[18:50:01 CET] <cone-842> ffmpeg 03James Almer 07master:cd8087444bb4: Merge commit 'ca44fa5d7fda7e954f3ebfeb5b0d6d1be55fcaa3'
[18:50:02 CET] <cone-842> ffmpeg 03James Almer 07master:ebdc5c419aef: Merge commit '41cf3e3b1ca375962951fde1b90a03b16197d205'
[19:15:08 CET] <fjant> hi
[19:16:39 CET] <fjant> any plans of updating doc/examples folder? Many files use deprecated apis. E.g. demuxing_decoding.c and avcodec_decode_video2(). What is the proper way?
[19:17:11 CET] <JEEB> I think there recently was a patch to one of the examples, but I didn't have time to look at it
[19:17:31 CET] <JEEB> the newer decoding/encoding APIs are rather alright documented @ https://www.ffmpeg.org/doxygen/trunk/group__lavc__encdec.html
[19:18:16 CET] <JEEB> IIRC the transcoding example (which does demux & decode as well) was already updated by someone if I recall correctly (or possibly it's new enough to have been made like that)
[19:20:16 CET] <fjant> That document looks rather promising! Thank you!
[19:40:32 CET] <cone-842> ffmpeg 03Janne Grunau 07master:bb515e3a735f: h264/aarch64: sign extend int stride in loop filter asm
[19:40:33 CET] <cone-842> ffmpeg 03Janne Grunau 07master:d7f4f5c4a18a: checkasm/h264: add loop filter tests
[19:40:34 CET] <cone-842> ffmpeg 03James Almer 07master:5ca7eb36b735: Merge commit 'bb515e3a735f526ccb1068031e289eb5aeb69e22'
[19:40:35 CET] <cone-842> ffmpeg 03James Almer 07master:d5d699ab6e6f: avcodec/h264dsp: change loop filter stride argument to ptrdiff_t
[19:40:36 CET] <cone-842> ffmpeg 03James Almer 07master:f32d29395557: Merge commit 'd7f4f5c4a18a0c9e62635cfa6fe8a9302b413c01'
[19:47:03 CET] <cone-842> ffmpeg 03Janne Grunau 07master:846c3d6aca54: h264/aarch64: optimize neon loop filter
[19:47:04 CET] <cone-842> ffmpeg 03Janne Grunau 07master:28a8b5413b64: h264/aarch64: add intra loop filter neon asm
[19:47:05 CET] <cone-842> ffmpeg 03Martin Storsjö 07master:eec93e57096a: libopenh264dec: Use a newer decoding entry point function
[19:47:06 CET] <cone-842> ffmpeg 03James Almer 07master:4dc1f06f0c84: Merge commit '846c3d6aca5484904e60946c4fe8b8833bc07f92'
[19:47:07 CET] <cone-842> ffmpeg 03James Almer 07master:e4e04dce1fab: Merge commit '28a8b5413b64b831dfb8650208bccd8b78360484'
[19:47:08 CET] <cone-842> ffmpeg 03James Almer 07master:f05f210526a3: Merge commit 'eec93e57096aa4804862d62760442380c70d489b'
[19:52:14 CET] <cone-842> ffmpeg 03Janne Grunau 07master:156ea66c91b1: h264/x86: sign extend int stride in deblock functions
[19:52:16 CET] <cone-842> ffmpeg 03Sven Dueking 07master:90b15f60bf49: srt: Set srto_sender flag to sender srt socket
[19:52:17 CET] <cone-842> ffmpeg 03James Almer 07master:0a319bcce571: Merge commit '156ea66c91b1986a87916f187216978d686725f6'
[19:52:18 CET] <cone-842> ffmpeg 03James Almer 07master:1811b7d1f533: Merge commit '90b15f60bf4919deaed66d868e18107eba371da7'
[21:15:59 CET] <cone-842> ffmpeg 03Adam Sampson 07master:bfa33f548db8: lavf/jacosubdec: compute subtitle duration correctly
[21:16:00 CET] <cone-842> ffmpeg 03Ruiling Song 07master:dfde416ea8d0: MAINTAINERS: add myself for tonemap_opencl
[21:53:21 CET] <cone-842> ffmpeg 03Lou Logan 07master:6bfda984b3d0: doc/mailing-list-faq: add Mail Delivery info
[21:53:29 CET] <durandal_1707> atomnuker: could you tell me how would you derive max_quant_idx[g][sfb] array in AC-4 by looking in specification? I'm thinking my approach might not be correct
[22:04:31 CET] <atomnuker> are the specs up somewhere?
[22:07:33 CET] <durandal_1707> atomnuker: https://www.etsi.org/deliver/etsi_ts/103100_103199/10319001/01.03.01_60/ts_10319001v010301p.pdf
[22:08:47 CET] <durandal_1707> atomnuker: see also https://github.com/richardpl/FFmpeg/tree/ac4
[22:17:35 CET] <atomnuker> yeah, max_quant_idx[g][i] looks correct in case there are only 4 codebook dims
[22:52:35 CET] <durandal_1707> atomnuker: this codec is big pile of stinking baloons with elephants inside them - the binary code for asf_spectral_data is clearly different - it does not use divisions and multiplications but only & and >> operations
[23:13:59 CET] <cone-842> ffmpeg 03Michael Niedermayer 07master:5ccf296e7472: avcodec/h264_direct: Fix overflow in POC comparission
[23:42:00 CET] <cone-842> ffmpeg 03James Almer 07master:aa844dc46f93: aarch64/h264dsp: change loop filter stride argument to ptrdiff_t
[23:42:01 CET] <cone-842> ffmpeg 03James Almer 07master:7b9ca44cbca4: arm/h264dsp: change loop filter stride argument to ptrdiff_t
[23:42:02 CET] <cone-842> ffmpeg 03James Almer 07master:72458b3f0828: ppc/h264dsp: change loop filter stride argument to ptrdiff_t
[23:57:29 CET] <cone-842> ffmpeg 03Carl Eugen Hoyos 07master:7e4d3dbe18ce: lavc/aarch64/h264dsp_init: Only use neon horizontal intra loopfilter for 4:2:0.
[00:00:00 CET] --- Thu Feb 21 2019


More information about the Ffmpeg-devel-irc mailing list