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 7.1.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
Kwiboo- has joined #ffmpeg-devel
mkver has quit [Ping timeout: 244 seconds]
Kwiboo has quit [Ping timeout: 268 seconds]
usagi_mimi has quit [Ping timeout: 252 seconds]
iive has quit [Quit: They came for me...]
kasper93_ has joined #ffmpeg-devel
kasper93 has quit [Killed (platinum.libera.chat (Nickname regained by services))]
kasper93_ is now known as kasper93
pshufb has joined #ffmpeg-devel
pshufb has quit [Read error: Connection reset by peer]
pshufb has joined #ffmpeg-devel
pshufb has joined #ffmpeg-devel
pshufb has quit [Changing host]
pshufb has quit [Ping timeout: 260 seconds]
Guest95 has joined #ffmpeg-devel
pshufb has joined #ffmpeg-devel
Guest95 has quit [Ping timeout: 272 seconds]
<aaabbb> oooh swscale finally gonna get full chroma and accurate_rnd by default soon? i like
jamrial has quit []
pshufb has quit [Ping timeout: 260 seconds]
Martchus_ has joined #ffmpeg-devel
Martchus has quit [Ping timeout: 252 seconds]
Traneptora has quit [Quit: Quit]
Traneptora has joined #ffmpeg-devel
Traneptora has quit [Quit: Quit]
mkver has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
pshufb has joined #ffmpeg-devel
pshufb has quit [Client Quit]
System_Error has quit [Remote host closed the connection]
pelotron has quit [Quit: ~pelotron()]
System_Error has joined #ffmpeg-devel
pelotron has joined #ffmpeg-devel
devinheitmuell-1 has joined #ffmpeg-devel
devinheitmueller has quit [Ping timeout: 248 seconds]
TheVibeCoder has joined #ffmpeg-devel
TheVibeCoder has quit [Changing host]
TheVibeCoder has joined #ffmpeg-devel
mkver has quit [Ping timeout: 260 seconds]
lexano has quit [Ping timeout: 268 seconds]
jluthra has quit [*.net *.split]
Nightrose has quit [*.net *.split]
Dmitri_Ovch has quit [*.net *.split]
bossjones has quit [*.net *.split]
tmatth has quit [*.net *.split]
jluthra has joined #ffmpeg-devel
Dmitri_Ovch has joined #ffmpeg-devel
bossjones has joined #ffmpeg-devel
Nightrose has joined #ffmpeg-devel
tmatth has joined #ffmpeg-devel
lexano has joined #ffmpeg-devel
Traneptora_ has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 260 seconds]
pelotron has quit [Ping timeout: 260 seconds]
Traneptora has quit [Read error: Connection reset by peer]
MisterMinister has joined #ffmpeg-devel
kasper93 has quit [Remote host closed the connection]
pelotron has joined #ffmpeg-devel
kasper93 has joined #ffmpeg-devel
ngaullier has joined #ffmpeg-devel
<fflogger> [editedticket] pross: Ticket #8349 ([avcodec] Dolby AC-4 Support) updated https://trac.ffmpeg.org/ticket/8349#comment:95
<Lynne> pross: 1.) the decoder is incomplete and still has issues 2.) please don't
<Lynne> this is exactly how phoronix operates and makes mistakes in every single article
meego has joined #ffmpeg-devel
<kode54> but, but, isn't that great? you can just go use the fork that has less quality control when it comes to accepting patches. and then advertise it on the original project's bug tracker when they don't accept your request to include the incomplete code
<kode54> "I'll go make my own ffmpeg! with blackjack and hookers!"
<TheVibeCoder> "less quality control"
<TheVibeCoder> "with blackjack and hookers!" ---> whatever that means to you - it means shit to me.
Everything has joined #ffmpeg-devel
<kode54> your name says all I need to know about you
HarshK23 has quit [Quit: Connection closed for inactivity]
<aaabbb> maybe, hopefully, it's ironic
<kode54> I would hope
kurosu has joined #ffmpeg-devel
<fflogger> [editedticket] zenden2k: Ticket #10385 ([undetermined] ddagrab fails on Windows if adapter ID is specified) updated https://trac.ffmpeg.org/ticket/10385#comment:1
meego has quit [Remote host closed the connection]
Everything has quit [Quit: leaving]
Kwiboo- is now known as Kwiboo
MetaNova has quit [Ping timeout: 276 seconds]
meego has joined #ffmpeg-devel
MetaNova has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 260 seconds]
j45 has quit [Quit: ZNC 1.9.1 - https://znc.in]
j45 has joined #ffmpeg-devel
j45 has joined #ffmpeg-devel
kasper93 has quit [Ping timeout: 252 seconds]
<CounterPillow> also that Jim Scarborough comment is LLM generated output btw
Kei_N_ has joined #ffmpeg-devel
Kei_N has quit [Ping timeout: 272 seconds]
paulk has quit [Ping timeout: 248 seconds]
paulk has joined #ffmpeg-devel
cone-130 has joined #ffmpeg-devel
<cone-130> ffmpeg Kacper Michajłow master:43dc443446e4: avutil/intmath: use AV_HAS_BUILTIN to detect builtin availability
<cone-130> ffmpeg Kacper Michajłow master:26807592e125: avformat/internal: add missing __clang__ check
<haasn> Lynne: why do you allocate 1024 layouts upfront? there is nowhere close to that many layouts defined by the spec
<haasn> and given that you already realloc on failure, why not just query the correct number from the driver
kasper93 has joined #ffmpeg-devel
<Lynne> haasn: crazy drivers, all of that code is gone with my unified layouts patch
Anthony_ZO has joined #ffmpeg-devel
<meego> I noticed Paul Mahol made some nice improvements to ebur128 filter in his fork. I'd like to try and port them back here but I'm not sure if that's ok licence-wise (or otherwise). According to its LICENCE.md, changes made in librempeg since the fork are GPL licensed.
rvalue has quit [Read error: Connection reset by peer]
<meego> My understanding is that inserting GPL-licensed code into f_ebur128.c would require relicencing it GPL-licensed, which sounds like a deadend. Am I understanding this right?
rvalue has joined #ffmpeg-devel
<TheVibeCoder> meego: did you know that transposed direct form II for IIR biquads filters exist?
<TheVibeCoder> meego: did speed goes faster than yours DI SIMD code?
<TheVibeCoder> with those 2 biquads from ebur128
<meego> TheVibeCoder: no i didnt'
<meego> he got better results than us (20% vs 17%)
<TheVibeCoder> for biquads, that are those filters that ebur128 uses
<TheVibeCoder> you have DI, DII, TDI, and TDII
<TheVibeCoder> and others beast, but that is off-topic
<TheVibeCoder> the simples of them all, DI is good for fixed point only code
<TheVibeCoder> for floating point TDII is best
<TheVibeCoder> thought for double sample formats that ebur128 filter usws it is less important
<TheVibeCoder> i have not yet tried to add full single-precision float path to ebur128 because i'm scared it might cause precision issues in output
<TheVibeCoder> also you could merge those 2 biquadst to 4 order IIR filter, i guess will be faster but might be worse precision wise/unstable for float (single-precision)
<meego> ok thanks. This is really useful.
<meego> I'm still very new to both C & signal processing, and I'm not sure I'll have the time to improve my skills in the near term. I'll post a bug bounty on the mailing list.
<meego> I wish i could have made better use of your advice, sorry
jamrial has joined #ffmpeg-devel
Traneptora_ has quit [Quit: Quit]
<cone-130> ffmpeg Marvin Scholz master:3d7668dae12f: avcodec/vvc/dec: fix possible null-pointer dereference
<cone-130> ffmpeg Marvin Scholz master:9ad6dca28ccb: avcodec/vvc: parse all SEI messages
kode548 has joined #ffmpeg-devel
kode54 has quit [Ping timeout: 252 seconds]
kode548 is now known as kode54
lexano has quit [Ping timeout: 252 seconds]
HarshK23 has joined #ffmpeg-devel
lexano has joined #ffmpeg-devel
Anthony_ZO has quit [Ping timeout: 260 seconds]
mkver has joined #ffmpeg-devel
minimal has joined #ffmpeg-devel
<thardin> the more i think about it, the more cursed handling edit lists in mov.c seems
<cone-130> ffmpeg Zhao Zhili master:f3bf5356966b: avcodec/libx264: Remove unnecessary include
<haasn> meego: TheVibeCoder: I implemented the same change in ffmpeg's f_ebur128.c and it's also faster
<haasn> I will add SIMD on top of that
<haasn> regarding the bug bounty I'm happy with only being awarded the speedup caused by SIMD on *top* of Paul's change, rather than versus current master
<haasn> these are just IIR biquads, shouldn't we just steal some uber high performance biquad kernel from pipewire or whatever?
<thardin> grm my email client is acting up
<haasn> that seems to be a better use of time than the "process 4 channels in parallel" simd approach from meego's original patch
<haasn> af_biquads also exists
<haasn> really what we should be doing is lifting the biquad kernel out of both of these files into some common place and adding SIMD there
<OctopusET> Hello, it's Moon. I was wondering if the VDD25 date has been set and where it's being discussed (mailing list or IRC channel).
<haasn> what af_biquads is doing is putting each channel on a separate thread
<TheVibeCoder> haasn: i already tried using threads within ebur128 and it was slower
<haasn> what was your threading strategy?
<TheVibeCoder> guess main reason for slowdown is additional memory moving around
<haasn> did you try to transpose the loop to do for (ch) { for (sample) { }} instead of the current for (sample) { for (ch) { }} ?
Kei_N_ has quit [Read error: Connection reset by peer]
<TheVibeCoder> obviously i did
<TheVibeCoder> that is main poit of it
<TheVibeCoder> *point
<thardin> what are some codecs with large frame size?
<jamrial> anything lossless for sure
Kei_N has joined #ffmpeg-devel
<fflogger> [newticket] Assasans: Ticket #11634 ([avfilter] rubberband causes SIGFPE crash on sample rates < 6024 Hz and pitch > 1) created https://trac.ffmpeg.org/ticket/11634
<thardin> the mov spec mandates being able to cut on samples
<thardin> aac can have big ones too I think
<haasn> TheVibeCoder: did you try merging both biquads into a 4th order filter?
<TheVibeCoder> nope
krez has joined #ffmpeg-devel
MisterMinister has joined #ffmpeg-devel
krez has quit [Quit: leaving]
<haasn> fwiw, changing from double to float is about 20% speedup
<TheVibeCoder> does it hurts fate?
ngaullier has quit [Remote host closed the connection]
<TheVibeCoder> is G.722.1 Annex C using IMDCT at all?
<TheVibeCoder> i kind of get warbling when using TX IMDCT code
<TheVibeCoder> the transform size is 640
<haasn> did not test fate with that change
<haasn> currently still busy with other refactors :)
<haasn> speaking of which, why does `make fate` on git master fail for me? missing filter/tremolo.pcm among others
<haasn> I of course did fate-rsync
<haasn> oh, it somehow no longer expands ~ in the fate sample dir path. I blame my shell for this
cone-130 has quit [Quit: transmission timeout]
rjw has joined #ffmpeg-devel
<tmatth> is it known that `make checkasm` fails on ARM 64?
<tmatth> (specifically on a pi 4)
rjw is now known as rjw_
rjw has joined #ffmpeg-devel
<welder> tmatth: can you share the compilation error?
rjw_ has quit [Ping timeout: 272 seconds]
<welder> Why is is trying to compile for x86? Did you pass --arch=amd64 to ./configure?
<tmatth> welder: I did not
<kierank> shouldn't need to
<tmatth> but let me double check that the repo is clean
<tmatth> it's possible I copied it from another machine and didn't clean it in the interim
<welder> It won't hurt to "grep ARCH_ config.h", is the appropriate arch (ARCH_AARCH64) set to 1?
iive has joined #ffmpeg-devel
rjw94 has joined #ffmpeg-devel
rjw has quit [Ping timeout: 272 seconds]
<tmatth> it is set to 1
rjw has joined #ffmpeg-devel
rjw94 has quit [Ping timeout: 272 seconds]
<tmatth> welder: that was it (I should've noticed the x86 there), sorry for the noise
<TheVibeCoder> looks like IMDCT does some strange left/right shifting with coefficients during inverse MDCT
drv_ is now known as drv
rjw has quit [Quit: Client closed]
<Lynne> make sure its not like siren where the "imaginaries" are inverted
<TheVibeCoder> nope. not that.
<TheVibeCoder> one part is more/less solved
<TheVibeCoder> its just final output shited left/right
<TheVibeCoder> aka multiplied or divided by 1 << -x or 1 / (1 << x)
<TheVibeCoder> the unfinished part is sub_shift and it happens before final MDCT output step
<TheVibeCoder> probably its there just because its fixed-point decoder
<TheVibeCoder> and float one can simplity it
<TheVibeCoder> and actually its successor of siren7/siren14
kurosu has quit [Quit: Connection closed for inactivity]
TheVibeCoder has quit [Ping timeout: 260 seconds]
BtbN has quit [Ping timeout: 276 seconds]
BtbN has joined #ffmpeg-devel
mkver has quit [Ping timeout: 268 seconds]
mkver has joined #ffmpeg-devel
lemourin1 has joined #ffmpeg-devel
lemourin has quit [Killed (iridium.libera.chat (Nickname regained by services))]
lemourin1 is now known as lemourin
<fflogger> [editedticket] mg3242: Ticket #11629 ([ffmpeg] Memory Leak in FFmpeg During Extended Processing) updated https://trac.ffmpeg.org/ticket/11629#comment:11
acryo has quit [Ping timeout: 252 seconds]
acryo has joined #ffmpeg-devel
acryo has joined #ffmpeg-devel