michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 6.1.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
rvalue has quit [Ping timeout: 256 seconds]
rvalue has joined #ffmpeg-devel
mkver has quit [Ping timeout: 240 seconds]
lexano has quit [Ping timeout: 246 seconds]
thilo has quit [Ping timeout: 255 seconds]
thilo has joined #ffmpeg-devel
qeed has joined #ffmpeg-devel
<j-b>
good mornign
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
wyatt8750 is now known as wyatt8740
Traneptora has joined #ffmpeg-devel
AbleBacon has quit [Read error: Connection reset by peer]
tmm1_ has quit [Ping timeout: 256 seconds]
tmm1 has joined #ffmpeg-devel
cone-963 has quit [Quit: transmission timeout]
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
qeed has quit [Remote host closed the connection]
MisterMinister has joined #ffmpeg-devel
qeed has joined #ffmpeg-devel
qeed_ has joined #ffmpeg-devel
qeed has quit [Read error: Connection reset by peer]
qeed_ has quit [Remote host closed the connection]
qeed_ has joined #ffmpeg-devel
jamrial_ has quit []
Wenbin_Chen_ has joined #ffmpeg-devel
Wenbin_Chen has quit [Ping timeout: 256 seconds]
APic has quit [Ping timeout: 264 seconds]
qeed_ has quit [Quit: Leaving]
Martchus has joined #ffmpeg-devel
qeed has joined #ffmpeg-devel
Martchus_ has quit [Ping timeout: 272 seconds]
qeed has quit [Remote host closed the connection]
Wenbin_Chen_ has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
Wenbin_Chen has joined #ffmpeg-devel
Wenbin_Chen has quit [Remote host closed the connection]
APic has joined #ffmpeg-devel
Kei_N has joined #ffmpeg-devel
Kei_N_ has quit [Ping timeout: 264 seconds]
qeed has quit [Quit: Leaving]
Kei_N has quit [Ping timeout: 256 seconds]
Kei_N has joined #ffmpeg-devel
cone-612 has joined #ffmpeg-devel
<cone-612>
ffmpeg Peter Ross master:db975ff00d21: avcodec/lead: support unaligned blocks
<cone-612>
ffmpeg Peter Ross master:37702e20663d: avcodec/lead: support format 0x0
Krowl has joined #ffmpeg-devel
<courmisch>
Lynne: what jkqxz said, I think
<courmisch>
and it wouldn't be so easy to define and implement on e.g. armv8
<courmisch>
you'd have to account for the address space size to avoid overflowing into PAC or tag bits
<another|>
urg. just accidently build ffmpeg single threaded. again. can we move to a better build system?
<nevcairiel>
just alias make to make -j
<another|>
hmm.. I mean that could work
<elenril>
there are no better build systems
<elenril>
there are build systems that suck in different ways
<elenril>
I'm yet to see one that is strictly better
<another|>
generally I find meson to suck less
<elenril>
I don't
<another|>
YMMV
<elenril>
I tried using it for some of my science stuff and regretted it
<elenril>
exactly
<nevcairiel>
there is plenty that sucks about meson, the number of unfixed bugs .. despite having PRs for years..
<nevcairiel>
at least our system lets us fix our own bugs
<another|>
sigh. alright
<nevcairiel>
if you want an integrated build system, it might be the least terrible
<nevcairiel>
what options are there, cmake, autotools? both also terrible :D
<wbs>
but ninja as build tool instead of make is pretty nice. I've played with the idea of generating ninja build files in ffmpeg just as an exercise, but I guess it wouldn't be maintainable unless we'd switch (which won't happen)
<wbs>
I've quite gotten used to cmake. it's not pretty, that's true, but it's at least mature and flexible enough to work for a wide variety of real world cases, contrary to meson :P
<elenril>
ugh, that sure is an opinion
<elenril>
I've always found cmake entirely opaque
<wbs>
it's not entirely obvious at the start, no, I agree with that
<elenril>
everything is some arcane invocation you have to dig out from google
<nevcairiel>
that is something you can say about any system you are not experienced with
<wbs>
yep. sure, it doesn't have the same level of transparency as a custom shell script though
<elenril>
./configure --help is pretty obvious
<elenril>
you can do most basic things with just that
<wbs>
what I really do appreciate, in particular with cmake+ninja, is how it keeps track of really rebuilding everything that does need to be rebuilt (say you reconfigure, and half of the files need to be rebuilt due to some cflags being changed, but the other half won't)
<wbs>
or if you check out a version that is older/newer so the build system files have changed, it will rerun as necessary. but if nothing has changed there, there's no unnecessary reconfigure
<wbs>
like, if you're bisecting ffmpeg, you'll need to rerun configure on every single build since you don't know if it's necessary or not
<another|>
hmm.. is ffv1 meaningfully SIMDable ?
<elenril>
huh, the STF submission deadline was like a week ago
<elenril>
was anything actuall submitted?
<another|>
seems like most time is spent in decode_plane.isra
Krowl has quit [Read error: Connection reset by peer]
Krowl has joined #ffmpeg-devel
ngaullie has joined #ffmpeg-devel
omegatron has joined #ffmpeg-devel
rooisnoek has joined #ffmpeg-devel
rooisnoek has quit [Client Quit]
derpydoo has joined #ffmpeg-devel
rajivharlalka has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
cone-612 has quit [Quit: transmission timeout]
Krowl has quit [Read error: Connection reset by peer]
TD-Linux has quit [Ping timeout: 252 seconds]
TD-Linux has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
ePirat has quit [Remote host closed the connection]
Krowl has quit [Read error: Connection reset by peer]
Krowl has joined #ffmpeg-devel
rajivharlalka has quit [Quit: Connection closed for inactivity]
derpydoo has quit [Ping timeout: 260 seconds]
<kurosu>
another|: lossless codecs are often bound by entropy enc/dec. You have to look at the bitstream reader itself or how it is used there
<kurosu>
(or near lossless)
<another|>
:(
<kurosu>
and... it's a rangecoder. So yeah...
<Lynne>
for ffv1's base profile it's just golomb
<Lynne>
you can speed that up with a similar way to the dirac/vc2 golomb decoder
elastic_dog has quit [Ping timeout: 264 seconds]
<JEEB>
hmm, should we somehow attempt to pass input stream metadata etc through filter_complex so that you don't need to map_metadata *all* streams by hand if you are using filter_complex?
elastic_dog has joined #ffmpeg-devel
cone-406 has joined #ffmpeg-devel
<cone-406>
ffmpeg James Almer master:ce7b519ab7df: avformat/mov: factorize out setting the output packet properties
<cone-406>
ffmpeg James Almer master:dbfa8381f6a4: avformat/movenc: add support for Immersive Audio Model and Formats in ISOBMFF
<cone-406>
ffmpeg James Almer master:9ba327e70fb6: avformat/mov: make MOVStreamContext refcounted
<cone-406>
ffmpeg James Almer master:b140b8332c61: avformat/demux: allow demuxers to output more than one packet per read_packet() call
<cone-406>
ffmpeg James Almer master:fe637161dbe6: avformat/mov: add support for Immersive Audio Model and Formats in ISOBMFF
<cone-406>
ffmpeg James Almer master:8c00fe64082d: fate: add IAMF in mp4 tests
Flat_ has joined #ffmpeg-devel
Flat has quit [Ping timeout: 260 seconds]
<cone-406>
ffmpeg Martin Storsjö master:58ffe0db4d50: flvdec: Honor the "flv_metadata" option for the "datastream" metadata field
<BradleyS>
🎉
darkapex has joined #ffmpeg-devel
AbleBacon has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
Krowl has joined #ffmpeg-devel
qeed has joined #ffmpeg-devel
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
jarthur has joined #ffmpeg-devel
qeed has quit [Read error: Connection reset by peer]
Marth64 has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
MetaNova has quit [Ping timeout: 246 seconds]
MetaNova has joined #ffmpeg-devel
Marth64 has quit [Ping timeout: 264 seconds]
Marth64 has joined #ffmpeg-devel
Marth64 has quit [Ping timeout: 246 seconds]
devinheitmuell-1 has joined #ffmpeg-devel
devinheitmuell-1 has quit [Client Quit]
devinheitmueller has quit [Ping timeout: 264 seconds]
devinheitmueller has joined #ffmpeg-devel
<llyyr>
what is required to get VK_KHR_video_decode_av1 support on ffmpeg? I compiled mesa from av1 branch but ffmpeg still doesn't think I have it
<llyyr>
is it enough to just replace the string VK_MESA string literals with VK_KHR?
<Lynne>
there's a patch on the ML
<Lynne>
it's not enough lol
<JEEB>
yea there seem to have been large'ish chagnes
<JEEB>
if you are able to test the patch set on the ML, that would be *very* welcome
<llyyr>
alright I'll give it a try
<llyyr>
uh, what am I looking for? searching av1 or vulkan doesn't return anything that seems like the patch for it
<llyyr>
thanks, need to update vulkan-headers first since my distro only updates them when a new vulkan-sdk is tagged
Mister_D has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 260 seconds]
rvalue has quit [Ping timeout: 255 seconds]
<llyyr>
that patchset needs rebasing after `{avcodec, tests}: rename the bundled Mesa AV1 vulkan video headers`
<llyyr>
but I tested it on an earlier commit and it works just fine for me
MisterMinister has joined #ffmpeg-devel
Mister_D has quit [Ping timeout: 272 seconds]
rvalue has joined #ffmpeg-devel
<JEEB>
nice
<JEEB>
at least it seems like that patch set got reviews
<llyyr>
I had a YCgCo AV1 sample that was decoded incorrectly by vaapi and d3d11va, so I wanted to try vulkan as well which also decodes it incorrectly
<nevcairiel>
i'm surprised that hasnt cause more weird issues, or do primaries and matrix typical common values happen to share the same numbers =P
<JEEB>
undefined I think does?
<JEEB>
but I think mostly this is due to how little AV1 hwdec there has been so far
<JEEB>
and thus how little usage ffAV1dec got so far
<llyyr>
I can send a patch unless it'd be faster for someone else to do it
<JEEB>
I can whip it and run FATE
<llyyr>
thanks
<jamrial>
mkver: you can remove the extra parenthesis, but lgtm either way
<cone-406>
ffmpeg Andreas Rheinhardt master:e2afcb82429a: avfilter/af_pan: Uninitialize channel layout
<cone-406>
ffmpeg Andreas Rheinhardt master:cffadfe99a1c: avformat/movenc: Make check actually check what is intended
<llyyr>
I ran FATE and nothing is failing that could be related to it
<llyyr>
I got a `Test ffmpeg-fix_sub_duration_heartbeat failed. Look at tests/data/fate/ffmpeg-fix_sub_duration_heartbeat.err for details.` though
<JEEB>
lol
<JEEB>
that hasn't failed once for me personally but yea, I've heard its stories
<JEEB>
also funny how it wouldn't change any tests
<JEEB>
I guess it then didn't get tested... at all :D
<llyyr>
I guess all the test samples have the same colorspace and primaries
<llyyr>
so it wouldn't matter
<mkver>
Or maybe the test output (that actually gets tested) just doesn't contain colorspace and primaries at all.
<JEEB>
yes
<JEEB>
in any case, the code effectively doesn't get tested
<JEEB>
I guess time to check the test samples in FATE
<nevcairiel>
either way the fix is correct, checking fate would mostly be to see if it needs to be updated
<JEEB>
yes
<JEEB>
I just ran fate and it worked, I'm just trying to figure if some test can be easily added
<llyyr>
the samples do have different colorspace and primaries
ngaullie has quit [Ping timeout: 240 seconds]
<JEEB>
yea I just checked the file names before my VM decided that it wants to freeze my keyboard input
<JEEB>
at least I have SSH open to my dev vm
<JEEB>
will have to I guess disable dav1d in my build just in case
<JEEB>
although I guess -c:v av1 should do it
<llyyr>
why does this only affect hwaccel though?
<JEEB>
because that's the only place av1dec is used
<JEEB>
otherwise you get dav1d parsing/exposing it
<llyyr>
ah
<JEEB>
(or aom)
<JEEB>
also meh, it seems like this information does not get exposed - at least not in the avformat context dump
<JEEB>
llyyr: at least the stuff under av1 and av1-test-vectors doesn't seem to output anything in the `-v verbose` output of ffprobe
<JEEB>
anyways, will send the patch since FATE passes
<JEEB>
I guess (enum AVColorSpace) is recommended
<JEEB>
granted, nothing else in that function is doing that
<JEEB>
so I will Keep Consistent In The File
devinheitmueller has quit [Quit: devinheitmueller]
devinheitmueller has joined #ffmpeg-devel
devinheitmueller has quit [Client Quit]
<JEEB>
sent
<JEEB>
llyyr: would be good if you could verify that I didn't somehow mess it up and/or if there was a secondary location to change
<llyyr>
nah, that's the same change I made
<JEEB>
gotcha. then would be nice if you could reply to the patch and say you verified it fixes the issue @ mpv issue
<JEEB>
will quickly check trac if there's an issue for that there
devinheitmueller has joined #ffmpeg-devel
<llyyr>
the mpv issue wasn't about this though :D
<llyyr>
they just had a broken file that rendered correctly because firefox and vlc ignored the colorspace
<llyyr>
mpv didn't so it looked "broken" to them
<JEEB>
lol
derpydoo has joined #ffmpeg-devel
<JEEB>
jamrial: thanks for the LGTM
<JEEB>
probably will have to be back-ported :D
<jamrial>
yes
<mkver>
jamrial: Did you look at implementing support for FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM in the av1dec.c?
<jamrial>
no
<cone-406>
ffmpeg Andreas Rheinhardt master:520c1b0a7c8f: avutil/hwcontext_vaapi: Allocate public and priv device hwctx together
<cone-406>
ffmpeg Andreas Rheinhardt master:870d9730d6f4: avutil/hwcontext_vaapi: Allocate pub and priv frames hwctx together
<jamrial>
mkver: problem is that skip_frame is currently checked when parsing a frame header, long after a sequence header obu would have been parsed, where a hwaccel pix_fmt will/should have been selected
<jamrial>
i guess setting that can be postponed
<mkver>
jamrial: Actually, I think the way the av1dec.c does things is nuts: If a pixel format gets chosen (which currently means: if a HW accel could be initialized successfully), it never calls get_pixel_format() again, even when a new sequence header with a different pixel format comes along.
<cone-406>
ffmpeg Jan Ekström master:37936b09ce31: avcodec/av1dec: fix matrix coefficients exposed by codec context
<jkqxz>
That does not sound good since it declares support for AV_CODEC_HW_CONFIG_METHOD_HW_FRAMES_CTX, which definitely requires it to do that.
<mkver>
jamrial: Definitely wrong. In this case the HW accel never sees the sequence header.
<jamrial>
how so?
<mkver>
If the first sequence header comes along, get_format has not been called yet and no hwaccel has been set up, so the "if (FF_HW_HAS_CB(avctx, decode_params))" is false and the decode_params call does not happen.
<jamrial>
ah
<jamrial>
afaik, no hwaccel has a decode_params() callback
<mkver>
You need to track whether decode_params has already been called with the current sequence header.
<mkver>
vulkan_av1 has
<mkver>
And vulkan_h264 and vulkan_hevc and videotoolbox_h264/hevc
<jamrial>
i was talking about av1 specifically, but yeah, forgot vulkan was added recently
<cone-406>
ffmpeg Jan Ekström release/6.1:a267d4ad4c92: avcodec/av1dec: fix matrix coefficients exposed by codec context
jarthur_ has joined #ffmpeg-devel
<JEEB>
woah, release/6.0 doesn't have a lot of backports :D
<JEEB>
/33/33
<JEEB>
whoops
Krowl has joined #ffmpeg-devel
jarthur has quit [Ping timeout: 246 seconds]
<cone-406>
ffmpeg Jan Ekström release/6.0:511b84442323: avcodec/av1dec: fix matrix coefficients exposed by codec context