[Ffmpeg-devel-irc] ffmpeg-devel.log.20151220

burek burek021 at gmail.com
Mon Dec 21 02:05:03 CET 2015


[00:00:00 CET] <fritsch> and other drivers fail heavily since a long time and only get slightly better
[00:00:20 CET] <fritsch> i don't have anything to complain with intel's gpu people
[00:00:30 CET] <fritsch> I'd have more to complain with vaapi management people
[00:00:50 CET] <iive> well, i hope that disabling dpms doesn't turn off backlight anymore.
[00:01:11 CET] <fritsch> i bear with you
[00:01:36 CET] <fritsch> BtbN: have built your wrong branch ...
[00:01:58 CET] <BtbN> hm? It's just on my master
[00:02:07 CET] <fritsch> yes
[00:02:15 CET] <fritsch> i used the branch from 20 november
[00:02:31 CET] <fritsch> cause it had vaapi in name
[00:02:34 CET] <BtbN> ah, that's ffmpeg.c vaapi accel tmp
[00:02:38 CET] <fritsch> hehe
[00:02:43 CET] <fritsch> which does exactly nothing, yeah
[00:07:03 CET] <iive> i do have intel cpu, but I don't have vga/dvi/hdmi output for the built-in video card.
[00:07:27 CET] <iive> is it possible to use it anyway, for encoding/decoding and opencl stuff?
[00:08:59 CET] <BtbN> Sure, but not for rendering of course
[00:09:22 CET] <BtbN> if you have the device in /dev/dri, you can use it
[00:09:32 CET] <nevcairiel> if the board doesnt even have an output, the bios may not activate the gpu device
[00:15:52 CET] <kierank> michaelni: does this wavelet look like a standard wavelet?  https://usercontent.irccloud-cdn.com/file/4wbBUA82/
[00:17:21 CET] <fritsch> 1440p sucks on my 3150
[00:17:51 CET] <BtbN> It's just plain not fast enough for it
[00:18:03 CET] <fritsch> 1080p60 skips in renderer a bit
[00:18:04 CET] <BtbN> It seems to be made to run 1080p60, and nothing more
[00:18:12 CET] <fritsch> jep
[00:18:14 CET] <fritsch> i think so too
[00:18:29 CET] <fritsch> we could try 1440p24 or something
[00:18:45 CET] <BtbN> It's visualy lagging on 1440p, so less than 24fps
[00:19:45 CET] <fritsch> oki - code wise I will push the kodi patch to kodi the moment your ffmpeg patch is reviewed and in ffmpeg git
[00:19:48 CET] <fritsch> thanks very much!
[00:23:04 CET] <BtbN> damn, a NUC6i5SYH is way too expensive
[00:23:36 CET] <BtbN> Same for the i3 version
[00:23:36 CET] <nevcairiel> thats essentially a desktop skylake cpu, those things are pricey
[00:24:01 CET] <BtbN> Yes, that NUC6i5SYH propably outperforms my desktop
[00:29:41 CET] <michaelni> kierank, i dont know what is considered a standard wavelet.  in compression and in academia the lists likely are different, i just know the types i worked with or read about and iam not good at remembering names of these things anyway
[00:30:07 CET] <kierank> ok, just wondering if it is similar to anything we have in the code
[00:30:12 CET] <kierank> looks like a Haar but it isnt'
[00:31:34 CET] <fritsch> BtbN: my haswell mobile also display VAProfileVP9Profile0
[00:31:42 CET] <fritsch> BtbN: I thought they had patched that out again?
[00:31:49 CET] <BtbN> Well, you can try if it works now :D
[00:31:56 CET] <fritsch> on the way
[00:32:03 CET] <BtbN> I don't think I own a HSW CPU
[00:32:54 CET] <BtbN> My Laptop is an i5-5200U
[00:33:24 CET] <BtbN> Not going to get any SKL Hardware, Broxton seems to be close
[00:33:54 CET] <fritsch> jep
[00:33:58 CET] <fritsch> will also go for broxton
[00:34:00 CET] <fritsch> hevc 10 bit
[00:34:02 CET] <fritsch> :-)
[00:34:30 CET] <BtbN> I could also already implement that, but it would be 100% blind, so there's not realy any point.
[00:35:06 CET] <fritsch> they intel vaapi guys seem to use an emulator
[00:35:20 CET] <fritsch> in one of the last commits I have read, that they had to do some real life changes
[00:35:21 CET] <BtbN> I wonder if VP8 hwaccel is worth it
[00:35:25 CET] <fritsch> cause the emulator did it differently
[00:35:43 CET] <fritsch> VP8 is gen9 only i think
[00:35:50 CET] <fritsch> that profile I don't have on my hsw box
[00:36:10 CET] <BtbN> More like a general question, there are no vp8 hwaccel hooks in ffmpeg
[00:36:25 CET] <BtbN> I'd guess the CPU load for it is light enough so it's not worth the effort.
[00:36:54 CET] <fritsch> i don't think VP8 will be a future 4k uhd codec
[00:37:17 CET] <BtbN> Well, it's still supported in VAAPI and at least DXVA
[00:38:48 CET] <BtbN> So might as well add it
[00:58:32 CET] <fritsch> BtbN: http://sprunge.us/fGIf
[00:59:29 CET] <BtbN> hm
[01:00:27 CET] <fritsch> http://sprunge.us/CVFa
[01:00:51 CET] <fritsch> crashing heavily in their hybrid driver
[01:00:52 CET] <fritsch> hehe
[01:01:29 CET] <BtbN> What did you do to make it explode?
[01:01:39 CET] <fritsch> just start it
[01:01:45 CET] <fritsch> that's hsw
[01:01:47 CET] <fritsch> bsw is fine
[01:01:48 CET] <BtbN> ah
[01:02:03 CET] <fritsch> which sucks - as we will need a setting or something
[01:02:39 CET] <BtbN> Well, I wouldn't bother implementing workaround for that hybrid-driver. It doesn't come anywhere.
[01:02:49 CET] <fritsch> i will check tomorrow
[01:03:09 CET] <fritsch> and see where it crashes exactly
[01:03:42 CET] <BtbN> It just shouldn't expose VP9 if it's not supported.
[01:03:49 CET] <fritsch> yes
[01:07:01 CET] <fritsch> http://lists.freedesktop.org/archives/libva/2015-October/003585.html
[01:07:06 CET] <fritsch> seems it should not be there at all
[01:15:21 CET] <fritsch> BtbN: https://github.com/01org/intel-hybrid-driver/issues/21
[01:21:39 CET] <BtbN> Oh, I just noticed a Windows 10 feature I realy like. I propably have been using it forever now, but just not realized it.
[01:21:48 CET] <BtbN> You don't need to focus a Window anymore in order to scroll in it.
[01:21:55 CET] <BtbN> It behaves like it does on Linux now.
[01:25:43 CET] <fritsch> hehe
[01:35:29 CET] <BtbN> Ok, decoding with mpv also works.
[01:40:13 CET] <BtbN> VP9, 1080p60: 93% GPU load.
[01:40:23 CET] <BtbN> HEVC, 4K60: 22% GPU load.
[01:46:00 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.8:3a3be0220834: avcodec/h264_refs: Fix long_idx check
[01:46:01 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.8:bdc6ba460429: avcodec/mpeg4videodec: also for empty partitioned slices
[03:06:54 CET] <cone-872> ffmpeg 03Andreas Cadhalpun 07release/2.8:63ecbb82fc1d: mlvdec: check that index_entries exist
[03:06:55 CET] <cone-872> ffmpeg 03Andreas Cadhalpun 07release/2.8:9be3441c311c: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[03:06:56 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.8:644179e0d415: Update Changelog
[03:09:57 CET] <jamrial> ubitux: is 5083900 worth backporting to release branches?
[03:11:14 CET] <ubitux> probably not that important, as you wish
[03:18:39 CET] <cone-872> ffmpeg 03Michael Niedermayer 07n2.8.4:HEAD: avcodec/mpeg4videodec: also for empty partitioned slices
[03:19:47 CET] <jamrial> mmh, nevermind. not even 2.8 needs it
[03:20:11 CET] <jamrial> the commit that introduced the bug was applied after that release was made
[03:45:57 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:1586fda2fc52: avfilter/vf_mpdecimate: Add missing emms_c()
[03:45:58 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:e274976c9208: Revert "avcodec/aarch64/neon.S: Update neon.s for transpose_4x4H"
[03:45:59 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:a628b7083100: avcodec/h264_slice: Simplify ref2frm indexing
[03:46:00 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:50ce06cc7c5a: avcodec/h264_mc_template: prefetch list1 only if it is used in the MB
[03:46:01 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:5124bf8ea6c3: avcodec/h264_refs: Fix long_idx check
[03:46:02 CET] <cone-872> ffmpeg 03Michael Niedermayer 07release/2.7:4bfaa040fc47: avcodec/mpeg4videodec: also for empty partitioned slices
[08:26:47 CET] <philipl> BtbN: Where are you seeing a nuc6i5 for sale? Can't find it anywhere in the US.
[08:27:02 CET] <philipl> And you should add the vp8 support anyway. It'll be fun :-)
[08:54:59 CET] <nevcairiel> intel promised to add vp8 dxva support, once that happens i'll add that, likely
[08:58:06 CET] <fritsch> philipl: http://geizhals.de/intel-nuc-kit-nuc6i5syh-boxnuc6i5syh-a1356132.html all 3 weeks+
[09:01:20 CET] <TD-Linux> BtbN, presumably all of that GPU load is just mpv's display pipeline in the HEVC case
[10:03:20 CET] <fritsch> TD-Linux: not really it uses egl directly with vaapi, so zero copy
[10:07:16 CET] <nevcairiel> copying doesnt show up as gpu load that heavily, it just eats up bandwidth
[10:17:52 CET] <fritsch> http://hevcadvance.com/pdf/RoyaltyRatesSummary.pdf <- perhaps interesting
[11:49:08 CET] <BtbN> philipl, they are listed here in germany, i could just order one.
[11:49:17 CET] <BtbN> Various online stores have them in stock
[11:50:06 CET] <fritsch> BtbN: check again
[11:50:22 CET] <BtbN> http://www.heise.de/preisvergleich/eu/intel-nuc-kit-nuc6i5syh-boxnuc6i5syh-a1356132.html
[11:50:24 CET] <fritsch> but who would seriously consider a skylake nuc
[11:50:37 CET] <fritsch> none in stock
[11:50:41 CET] <BtbN> Yeah, way too expensive, waiting for broxton
[12:18:03 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07master:9fcfe4a3cdf9: mlvdec: check that index_entries exist
[12:18:03 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07master:699e68371ec7: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[12:29:07 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07master:f6830cf5ba03: nuv: sanitize negative fps rate
[12:43:22 CET] <philipl> BtbN: I think prices will be much more reasonable in the US.
[13:42:17 CET] <hard_iron> hello everyone
[13:42:39 CET] <hard_iron> i was going through gsoc ffmpeg ideas of projects
[13:42:53 CET] <hard_iron> there i found true hd encoder
[13:43:06 CET] <hard_iron> i want to work on it now
[13:43:20 CET] <hard_iron> can someone help m out with that project
[13:43:29 CET] <hard_iron> how i should start
[13:44:33 CET] <J_Darnley> Start by making sure you can compile ffmpeg.
[13:45:05 CET] <hard_iron> compile ffmpeg?
[13:45:49 CET] <J_Darnley> Grab the source and then run configure and make without encountering any errors.
[13:46:29 CET] <J_Darnley> If you can't build the project I don't know how you could ever hope to add any feature.
[13:46:52 CET] <hard_iron> actually i am new to all this.....can you guide me from basics
[13:47:38 CET] <cbsrobot_> hard_iron: see https://trac.ffmpeg.org/wiki/CompilationGuide
[13:48:04 CET] <JEEB> of course, those guides have the whole kitchen sink, which you don't need
[13:48:36 CET] <JEEB> so you can just have the initial build dependencies (yasm+compiler suite+shell and friends) and run ./configure
[13:49:25 CET] <JEEB> also I have no idea how much information there is available for truehd (specs, multimedia.cx wiki etc)
[13:51:29 CET] <fritsch> JEEB: most specs are in ffmpeg's source code
[13:51:35 CET] <fritsch> quite well documented
[13:51:36 CET] <JEEB> true
[13:51:50 CET] <JEEB> well, not specs but in some things there's quite a bit of documentation
[13:52:15 CET] <fritsch> specs are not free I think - but available "somewhere"
[13:52:15 CET] <JEEB> multimedia.cx wiki can be a treasure trove, as well as any official specs
[13:52:22 CET] <JEEB> yeah, that's the usual way :P
[13:52:32 CET] <JEEB> whether or not they are available "somewhere" is the major question
[13:52:42 CET] <fritsch> they are
[13:52:51 CET] <fritsch> but that makes "somewhere" to the biggest part of the question
[13:53:03 CET] <JEEB> was TrueHD Meridian Lossless before it became TrueHD?
[13:53:09 CET] <JEEB> or do I remember incorrectly
[13:53:55 CET] <nevcairiel> it was
[13:54:02 CET] <nevcairiel> and i dont think anyone found the specs yet
[13:54:06 CET] <JEEB> yeah
[13:54:18 CET] <JEEB> so this isn't exactly an easy mission if someone embarks on it :P
[13:54:30 CET] <JEEB> even with the possible docs in the code or multimedia.cx
[13:55:12 CET] <nevcairiel> lossless encoders might be easier there, even if its somewhat inefficient, you can just implement bitstream features you know, not like decoders that need to implement all the features
[13:55:29 CET] <JEEB> yeah
[13:55:32 CET] <JEEB> that's true
[13:59:31 CET] <rcombs> isn't TrueHD just MLP with a different syncword
[14:00:13 CET] <nevcairiel> some extensions too
[14:00:16 CET] <nevcairiel> mlp was only 5.1
[14:06:44 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:292842a0ed80: avcodec/vp8: Do not use num_coeff_partitions in thread/buffer setup
[14:06:45 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:dd4c2fe74afe: avcodec/h264_slice: Limit max_contexts when slice_context_count is initialized
[14:06:46 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:2da8c533869a: avcodec/cabac_functions: Fix "left shift of negative value -31767"
[14:06:47 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:1f6aea2cc4e3: avcodec/cabac: Check initial cabac decoder state
[14:06:48 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:6157938408a0: avcodec/hevc: Check entry_point_offsets
[14:06:49 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:20a96b9d8c71: avcodec/jpeg2000dwt: Check ndeclevels before calling dwt_decode*()
[14:06:50 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:40d5f570d8aa: avcodec/jpeg2000dwt: Check ndeclevels before calling dwt_encode*()
[14:06:51 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:4916aa238315: avcodec/hevc_cabac: Fix multiple integer overflows
[14:06:52 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:c84ec8d5c938: avcodec/hevc: allocate entries unconditionally
[14:06:53 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:67b29888f4dc: avcodec/vp3: Clear context on reinitialization failure
[14:06:54 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:80a24e88e3db: avcodec/utils: Use 64bit for aspect ratio calculation in avcodec_string()
[14:06:55 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:a0a2bfbfd876: avcodec/utils: Clear dimensions in ff_get_buffer() on failure
[14:06:56 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:93e168705dbe: avformat/dump: Fix integer overflow in av_dump_format()
[14:06:57 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:d8a66a150ae1: avutil/integer: Fix av_mod_i() with negative dividend
[14:06:58 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:100dbe1bf007: avutil/mathematics: Do not treat INT64_MIN as positive in av_rescale_rnd
[14:06:59 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:4d8d2f1991cd: avcodec/vp3: always set pix_fmt in theora_decode_header()
[14:07:00 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:0350f373f7f0: avcodec/apedec: Check length in long_filter_high_3800()
[14:07:01 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:00fde782a109: avutil/mathematics: return INT64_MIN (=AV_NOPTS_VALUE) from av_rescale_rnd() for overflows
[14:07:02 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:2e77ab8f1009: avutil/timecode: Fix fps check
[14:07:03 CET] <cone-972> ffmpeg 03Timo Teräs 07release/2.6:e8fa0e3ec633: mpegencts: Fix overflow in cbr mode period calculations
[14:07:04 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:ad09bebe0cad: avcodec/vp3: Fix "runtime error: left shift of negative value"
[14:07:05 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:354fa47018e3: avformat/smacker: fix integer overflow with pts_inc
[14:07:06 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:5164067861ed: avcodec/wmaprodec: Fix overflow of cutoff
[14:07:07 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:e12f07b4751c: avcodec/wmaprodec: Check bits per sample to be within the range not causing integer overflows
[14:07:08 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:cc88d7a64088: avcodec/dirac_parser: Fix potential overflows in pointer checks
[14:07:09 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:09305dca59fd: avcodec/dirac_parser: Add basic validity checks for next_pu_offset and prev_pu_offset
[14:07:10 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:929fd6160271: avcodec/hevc: Fix integer overflow of entry_point_offset
[14:07:11 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:0fa22a42d6b2: swscale/utils: Fix for runtime error: left shift of negative value -1
[14:07:12 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:8681296bb97f: avcodec/pgssubdec: Fix left shift of 255 by 24 places cannot be represented in type int
[14:07:13 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:7294d95689f9: avcodec/jpeg2000dec: Check bpno in decode_cblk()
[14:07:14 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:63c50350b6e5: avcodec/vp3: ensure header is parsed successfully before tables
[14:07:15 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:3e8d9eade655: avcodec/hevc: Check max ctb addresses for WPP
[14:07:16 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:f821441795c4: avutil/mathematics: Fix division by 0
[14:07:17 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:8e053f61f562: avformat/matroskaenc: Check codecdelay before use
[14:07:18 CET] <cone-972> ffmpeg 03Rainer Hochecker 07release/2.6:bc264004e081: avformat/utils: estimate_timings_from_pts - increase retry counter, fixes invalid duration for ts files with hevc codec
[14:07:19 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:10395b4738df: avformat/mxfenc: Do not crash if there is no packet in the first stream
[14:07:20 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:3f7bbb0ded3f: avfilter/vf_mpdecimate: Add missing emms_c()
[14:07:21 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:d0d9fad2ba5a: avcodec/h264_refs: Fix long_idx check
[14:07:22 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.6:c1b7e9fe52d3: avcodec/mpeg4videodec: also for empty partitioned slices
[14:09:32 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:6d7b4dbcb410: on2avc: limit number of bits to 30 in get_egolomb
[14:09:33 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:945ae04fab45: exr: fix out of bounds read in get_code
[14:09:34 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:38f8c8090103: nutdec: only copy the header if it exists
[14:09:35 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:778c8de40f2c: xwddec: prevent overflow of lsize * avctx->height
[14:09:36 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:174ec7d744d3: nutdec: reject negative value_len in read_sm_data
[14:09:37 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.8:79f407b79a82: nuv: sanitize negative fps rate
[14:31:32 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:3e187a9a2dfa: on2avc: limit number of bits to 30 in get_egolomb
[14:31:33 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:5b88d24f241f: exr: fix out of bounds read in get_code
[14:31:34 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:7f320ed15d7b: nutdec: only copy the header if it exists
[14:31:35 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:f040f3e5d5ab: xwddec: prevent overflow of lsize * avctx->height
[14:31:36 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:1dc135e03d83: nutdec: reject negative value_len in read_sm_data
[14:31:37 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:93eda8dcbadf: mlvdec: check that index_entries exist
[14:31:38 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:3219de21f4e9: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[14:31:39 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.7:7ad058168665: nuv: sanitize negative fps rate
[15:24:17 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:564e63a0334f: on2avc: limit number of bits to 30 in get_egolomb
[15:24:18 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:350f1f4919a1: exr: fix out of bounds read in get_code
[15:24:19 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:ea5ac146cd94: nutdec: only copy the header if it exists
[15:24:20 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:c4b2985d9ab0: xwddec: prevent overflow of lsize * avctx->height
[15:24:21 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:4d2f93680343: nutdec: reject negative value_len in read_sm_data
[15:24:22 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:c51800d324a9: mlvdec: check that index_entries exist
[15:24:23 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:462afe5d00c8: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[15:24:24 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.6:b2cb7db14f82: nuv: sanitize negative fps rate
[15:30:30 CET] <BtbN> vp8 is way more annoying than vp9...
[15:30:45 CET] <BtbN> The ffmpeg vp8 struct doesn't match the stuff VAAPI wants in the slightest
[15:32:45 CET] <nevcairiel> vp9 didnt start out like this either =p
[15:35:09 CET] <BtbN> You modified the data the vp9 decoder exposes in its struct?
[15:35:40 CET] <BtbN> I'd guess everything VP8 needs is there, but named diffrently and potentialy not in the right format.
[15:38:50 CET] <BtbN> VAAPI also wants an out_of_loop_frame reference frame, but there is no notion of that in ffmpeg and the DXVA2 spec at all. 
[15:39:34 CET] <nevcairiel> i didnt look at vp8 at all yet, there is no driver support yet, so meh
[15:40:03 CET] <BtbN> VAAPI seems to be the one that's special about it, the DXVA spec and ffmpeg seem to match somewhat better
[15:40:18 CET] <BtbN> Might be because VP8 support is in VAAPI for 3 years now
[15:42:26 CET] <fritsch> BtbN: afaik only braswell supports it via hybrid driver
[15:42:29 CET] <fritsch> my hsw does not
[15:42:33 CET] <fritsch> my ivb also not
[15:42:40 CET] <BtbN> it? vp8?
[15:42:43 CET] <fritsch> jep
[15:42:51 CET] <BtbN> It's listed in vainfo for me
[15:42:53 CET] <fritsch> http://sprunge.us/gbcY
[15:42:54 CET] <BtbN> From the primary driver
[15:42:57 CET] <fritsch> on your braswell
[15:43:03 CET] <fritsch> most likely
[15:43:16 CET] <BtbN> Only VP9 comes from the hybrid driver, everything else is native
[15:43:17 CET] <fritsch> but that's it - so SKL, BSW and later
[15:43:31 CET] <nevcairiel> full vp8 hw support should be slightly older
[15:43:41 CET] <BtbN> libva supports VP8 since 2012
[15:43:43 CET] <fritsch> i also expected haswell to have it
[15:43:49 CET] <fritsch> but yeah - not the case
[15:43:54 CET] <fritsch> broadwell might have it
[16:01:25 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:cc77012329f0: ffmdec: reject zero-sized chunks
[16:01:27 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:26b74a7213bf: aaccoder: prevent crash of anmr coder
[16:01:27 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:5478e7bbc106: ffm: reject invalid codec_id and codec_type
[16:01:29 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:31d690016122: opus_silk: fix typo causing overflow in silk_stabilize_lsf
[16:01:30 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:46e7a63b6e01: sonic: make sure num_taps * channels is not larger than frame_size
[16:01:31 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:105be66545ec: on2avc: limit number of bits to 30 in get_egolomb
[16:01:32 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:79d75f1554de: exr: fix out of bounds read in get_code
[16:01:33 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:0f1e39887439: nutdec: only copy the header if it exists
[16:01:34 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:882391f9758e: xwddec: prevent overflow of lsize * avctx->height
[16:01:35 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:2db78f83bad1: nutdec: reject negative value_len in read_sm_data
[16:01:36 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:94711d3719a1: mlvdec: check that index_entries exist
[16:01:37 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:06edade7ee8a: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[16:01:38 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.5:eca27eeea95f: nuv: sanitize negative fps rate
[16:01:39 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.5:8cd0e23be83d: avcodec/mpeg4videodec: also for empty partitioned slices
[16:11:55 CET] <cone-972> ffmpeg 03Paul B Mahol 07master:577f05735511: avcodec/s302menc: set supported channel layouts by codec
[16:11:56 CET] <cone-972> ffmpeg 03Paul B Mahol 07master:db6e337b41fc: avcodec/s302menc: check if buf_size can actually be put into 16bit size
[16:28:24 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:7f33fef2a28d: aaccoder: prevent crash of anmr coder
[16:28:25 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:6fc3f6f43b24: ffm: reject invalid codec_id and codec_type
[16:28:26 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:7a26ea7a7e67: opus_silk: fix typo causing overflow in silk_stabilize_lsf
[16:28:27 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:465dd4bc941e: sonic: make sure num_taps * channels is not larger than frame_size
[16:28:28 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:e32095807b86: on2avc: limit number of bits to 30 in get_egolomb
[16:28:29 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:e7b09eaefa5d: exr: fix out of bounds read in get_code
[16:28:30 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:2306964b3a03: nutdec: only copy the header if it exists
[16:28:31 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:dcecc180a6ad: xwddec: prevent overflow of lsize * avctx->height
[16:28:32 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:bbe1c9839bb9: nutdec: reject negative value_len in read_sm_data
[16:28:33 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:f0eea9cc3a11: mlvdec: check that index_entries exist
[16:28:34 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:b0a8095f2bf0: rawdec: only exempt BIT0 with need_copy from buffer sanity check
[16:28:35 CET] <cone-972> ffmpeg 03Andreas Cadhalpun 07release/2.4:1317c63b4b1d: nuv: sanitize negative fps rate
[16:28:36 CET] <cone-972> ffmpeg 03Michael Niedermayer 07release/2.4:fbfb2814b28d: avcodec/mpeg4videodec: also for empty partitioned slices
[17:02:12 CET] <BtbN> Is ff_vp8_dct_cat_prob in vp8data.h just missing a static, or am I missing something else?
[17:02:51 CET] <BtbN> Including that header generates a multiple definition error for that symbol on linking
[17:51:29 CET] <cone-972> ffmpeg 03Paul B Mahol 07master:367ffa0c1517: avcodec/flacenc: use designated initializers for AVClass
[18:00:36 CET] <BtbN> Kinda interesting that the "header size" stuff VAAPI wants for VP8 matches exactly the parameters that GStreamer uses for parsing that stuff
[18:00:56 CET] <BtbN> But there's nothing similar in ffmpeg, it handles parsing the bitstream entirely diffrent
[18:01:18 CET] <BtbN> And it kinda expects you to know what it means by range, value and count.
[18:15:59 CET] <BtbN> Hm, I don't think it's possible to pass the header size, the FFMpeg Range-Coder does not seem to store that information
[19:54:35 CET] <cone-972> ffmpeg 03Paul B Mahol 07master:9e569abe995d: avfilter/avf_showfreqs: make it possible to split channels
[20:22:46 CET] <kurosu> jamrial, looked at your patchset for hevc sao, tested them, they look fine
[20:23:24 CET] <kurosu> I didn't go very in-depth, but that mostly looked like moving around code
[20:23:38 CET] <jamrial> kurosu: ok, thanks
[20:24:19 CET] <kurosu> so there might some change I missed besides this
[20:24:49 CET] <kurosu> sorry for holding you, as I said, the review was probably not worth it
[20:26:52 CET] <jamrial> it basically makes the 16 width avx2 functions use ymm registers instead of xmm, something i should have done from the beginning, which allowed me simplify all the macro madness
[20:33:02 CET] <kurosu> jamrial, I was indeed verifying whether the generated code was different and had just noticed that
[20:43:53 CET] <cone-972> ffmpeg 03James Almer 07master:34b2bd03cf94: x86/hevc_sao: simplify sao_band_filter 10/12bit
[20:47:50 CET] <cone-972> ffmpeg 03James Almer 07master:3ff2beff65af: x86/hevc_sao: simplify sao_edge_filter 10/12bit
[21:01:55 CET] <cone-972> ffmpeg 03James Almer 07master:d4c47333e148: x86/hevc_sao: add ff_hevc_sao_edge_filter_{8,16}_{10,12}
[21:11:29 CET] <BtbN> Anyone familiar with the bitstream format of VPx? I have no idea what values VAAPI wants for the bool coder context.
[21:13:35 CET] <BtbN> http://cgit.freedesktop.org/vaapi/libva/tree/va/va_dec_vp8.h#n51 those values, to be precise. 
[21:13:47 CET] <BtbN> And also these ones, which seem to be somehow related: http://cgit.freedesktop.org/vaapi/libva/tree/va/va_dec_vp8.h#n187
[21:19:04 CET] <fritsch> BtbN: did you check libyami?
[21:19:41 CET] <BtbN> nothing usefull there, that stuff is entirely undocumented
[21:19:51 CET] <fritsch> https://github.com/01org/libyami/blob/master/decoder/vaapidecoder_vp8.cpp
[21:20:25 CET] <fritsch> fileSliceParam and fillPictureParam
[21:20:39 CET] <fritsch> looks quite informative
[21:22:09 CET] <BtbN> it just happens to have a bool decoder where these 3 values fall out of
[21:22:14 CET] <BtbN> what they are isn't mentioned at all
[21:22:22 CET] <BtbN> and the ffmpeg decoder doesn't have anything remotely similar
[21:26:10 CET] <nevcairiel> the dxva spec often refers to "variable of the same name in the reference decoder", which helps looking things up, i dont suppose vaapi does that?
[21:32:41 CET] <cone-972> ffmpeg 03Paul B Mahol 07master:484cc66f577b: avcodec/indeo2: use init_get_bits8
[21:37:30 CET] <BtbN> nevcairiel, not for these values
[21:37:37 CET] <BtbN> For most of the other ones it does
[21:37:58 CET] <BtbN> It has some cryptic descriptions, for some of them that don't make much sense
[21:38:27 CET] <BtbN> gst/yami just copied parts of the reference decoder where exactly these values fall out of, but ffmpeg doesn't.
[21:40:53 CET] <fritsch> hehe
[22:13:53 CET] <kierank> atomnuker: fosdem people asking you to add a description
[22:16:06 CET] <atomnuker> kierank: ok, any info on how long it should be?
[22:16:13 CET] <kierank> 50 words
[22:16:16 CET] <kierank> dunno
[22:46:54 CET] <atomnuker> kierank: wanna proof-read? http://paste.debian.net/hidden/c04974d9/
[22:48:40 CET] <kierank> lol
[22:48:45 CET] <kierank> it's like a speech...
[22:48:47 CET] <kierank> good :)
[22:50:33 CET] <atomnuker> I especially like the last sentence
[22:50:50 CET] <kierank> yes it'll troll the other speakers
[22:53:25 CET] <atomnuker> the thor folks wouldn't happen to be there, would they?
[22:53:38 CET] <atomnuker> I know some google folks would be there, they're everywhere
[22:53:52 CET] <kierank> dunno if thor people will be there
[22:53:55 CET] <kierank> thor is a small team afaik
[22:53:59 CET] <kierank> ask tim
[22:56:12 CET] <J_Darnley> atomnuker: "there would have to exists a support for such formats" isn't very good English
[23:07:15 CET] <ubitux> "to adapt [to?] new technologies"?
[23:09:15 CET] <atomnuker> J_Darnley: replaced with "support for such formats would have to exist within the browsers themselves"
[23:09:56 CET] <atomnuker> I can always make the excuse and get away by saying it's not my native language
[23:11:08 CET] <atomnuker> kierank: tell the fosdem people I've updated the abstract and descriptions as well as the event image
[23:11:22 CET] <J_Darnley> That's alwaysa good excuse and that change does read better.
[23:11:25 CET] <kierank> ok
[00:00:00 CET] --- Mon Dec 21 2015


More information about the Ffmpeg-devel-irc mailing list