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 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
vipyne has joined #ffmpeg-devel
vipyne has quit [Ping timeout: 245 seconds]
thilo_ has quit [Ping timeout: 244 seconds]
thilo_ has joined #ffmpeg-devel
thilo_ has quit [Changing host]
thilo_ has joined #ffmpeg-devel
iive has quit [Quit: They came for me...]
vipyne has joined #ffmpeg-devel
vipyne has quit [Ping timeout: 245 seconds]
vipyne has joined #ffmpeg-devel
vipyne has quit [Ping timeout: 272 seconds]
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
arch1t3cht1 has joined #ffmpeg-devel
arch1t3cht has quit [Ping timeout: 265 seconds]
arch1t3cht1 is now known as arch1t3cht
mkver has quit [Ping timeout: 276 seconds]
Marth64 has quit [Quit: Leaving]
jamrial has quit []
Martchus has joined #ffmpeg-devel
Martchus_ has quit [Ping timeout: 252 seconds]
Marth64 has joined #ffmpeg-devel
vipyne has joined #ffmpeg-devel
vipyne_ has joined #ffmpeg-devel
MisterMinister has quit [Remote host closed the connection]
vipyne has quit [Quit: Leaving.]
vipyne_ has quit [Ping timeout: 248 seconds]
vipyne has joined #ffmpeg-devel
vipyne has quit [Ping timeout: 276 seconds]
ngaullier has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
ngaullier has quit [Read error: Connection reset by peer]
ngaullier has joined #ffmpeg-devel
feiw1 has quit [Ping timeout: 272 seconds]
feiw1 has joined #ffmpeg-devel
j45 has quit [Ping timeout: 248 seconds]
j45 has joined #ffmpeg-devel
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
ccawley2011 has joined #ffmpeg-devel
ccawley2011_ has joined #ffmpeg-devel
ccawley2011 has quit [Ping timeout: 260 seconds]
ccawley2011 has joined #ffmpeg-devel
ccawley2011_ has quit [Ping timeout: 245 seconds]
ccawley2011_ has joined #ffmpeg-devel
ccawley2011 has quit [Ping timeout: 245 seconds]
cone-717 has joined #ffmpeg-devel
<cone-717>
ffmpeg Anton Khirnov master:ad50254ee1bc: fftools/ffmpeg_demux: use proper logging contexts everywhere
<cone-717>
ffmpeg Anton Khirnov master:e79ac9312f02: fftools/ffmpeg_filter: use proper logging contexts
<cone-717>
ffmpeg Anton Khirnov master:6d6bd86f2b97: fftools/opt_common: stop accessing a private field
<cone-717>
ffmpeg Anton Khirnov master:242ee7b0a23f: fftools/ffmpeg_sched: be smarter about flushing the pre-muxing queues
<elenril>
it's impressive how broken that x265 patch is, given its size
<wbs>
... wow
<elenril>
and we really should have marked x265 as experimental
<cone-717>
ffmpeg Anton Khirnov master:a96244df77fe: lavfi/vf_setparams: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:965e9b646ed0: lavfi/vf_showpalette: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:e2f4c69779dd: lavfi/vf_shuffleplanes: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:bdaaddb65c7e: lavfi/vf_stack: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:6ed357ce77c2: lavfi/vf_stereo3d: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:05aad90c69bd: lavfi/vf_subtitles: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:06292d34dee0: lavfi/vf_swaprect: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:a18418f39d2e: lavfi/vf_swapuv: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:14ece0bf589e: lavfi/vf_telecine: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:ce5aa740451c: lavfi/vf_tile: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:ef4913e30d0c: lavfi/vf_tpad: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:5d0fbd714543: lavfi/vf_transpose: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:223327d51a36: lavfi/vf_untile: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:73c863fad888: lavfi/vf_v360: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:e5b530f13751: lavfi/vf_vmafmotion: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:ab01f0522bb7: lavfi/vf_vpp_qsv: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:6bff95be0332: lavfi/vf_weave: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:d08feb900348: lavfi/vf_zscale: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:2094a7256cc0: lavfi/vsrc_life: avoid modifying the context in query_formats()
<cone-717>
ffmpeg Anton Khirnov master:e0aa644978a1: lavfi/vsrc_life: convert to query_func2()
<cone-717>
ffmpeg Anton Khirnov master:c03421308337: lavfi/vsrc_perlin: convert to query_func2()
<galad>
fun fact: x265 has got some neon dotprod and i8mm optimizations now, unfortunately no one thought of adding some checks to see if those instructions are available at run time (actually someone tried to add a partial one, but it was reverted for some unknown reasons)
<JEEB>
ouch
<JEEB>
elenril: it's experimental in a manner that's mostly invisible to end users who kinda get a more or less working bit stream :D
<elenril>
except when it randomly corrupts your memory or SIGILLs
<wbs>
galad: oh geez
<JEEB>
elenril: yea random memory touching and incorrectness exists indeed. which only showed up under valgrind or msan or whatever, which is why my FATE test first got merged :D
<elenril>
"it's a benign UB"
<Lynne>
is x265 used much these days?
<elenril>
I think so
<JEEB>
not really, just noting that it very much gets unnoticed by ye normie and then hits us devs
<JEEB>
yea pretty much all of those HEVC encodes around are being done with it methinks
<Lynne>
I have the impression most that need hevc use hardware encoders, and those that don't are stuck on h264 and av1
<Lynne>
ah... pirates
<elenril>
how horrible is svt?
<JEEB>
SVT-AV1 at least has people caring about visual quality and they are upstreaming their changes
<elenril>
i mean svt-hvc
<JEEB>
SVT-HEVC is quite horrible since nobody cared about it
<JEEB>
yea, they announced they stopped maintaining it too
Krowl has joined #ffmpeg-devel
<JEEB>
for the past few years there were no real changes
microlappy has joined #ffmpeg-devel
<elenril>
writing good encoders seems to be lost technology
<kurosu>
M1s do not have i8mm. But the proper fix is obviously for customers to update to M3
<elenril>
we should merge x264 into lavc and have someone extend it to support hevc
<elenril>
should be easy, no?
<kurosu>
With your growing knowledge of hevc, that sounds like you'd be enjoying it, for sure
ccawley2011 has joined #ffmpeg-devel
ccawley2011_ has quit [Ping timeout: 244 seconds]
Workl has joined #ffmpeg-devel
<Lynne>
the psychovisual system should be portable
<Lynne>
the rest, well, its a bitstream writer
Krowl has quit [Ping timeout: 255 seconds]
microlappy has quit [Quit: Konversation terminated!]
Krowl has joined #ffmpeg-devel
Workl has quit [Ping timeout: 265 seconds]
ccawley2011_ has joined #ffmpeg-devel
ccawley2011 has quit [Ping timeout: 265 seconds]
Workl has joined #ffmpeg-devel
Krowl has quit [Ping timeout: 245 seconds]
<Daemon404>
hey courmisch (because youre likely to know): whats the least bad way to stay in contact with someone who is travellingnto mainland china and doesnt wanna use a vpn? wechat?
<Daemon404>
(in this case my wife is going and im not)
Krowl has joined #ffmpeg-devel
Workl has quit [Ping timeout: 264 seconds]
<ramiro>
haasn: sorry, I missed the follow up to your swscale cleanup patches. are you keeping your work on a public repo so that it's easier to track than using the ml and patchwork?
<haasn>
seems like wrong color matrix being appliad
<haasn>
but I looked at the color matrices in gdb and they're identical both before and after the patch
<haasn>
but the difference is so small it seems more like rounding error
<haasn>
maybe there's some issue in copying some floating point coefficient somewhere?
<haasn>
though it doesn't really make sense for this patch to have affected that
<haasn>
that also doesn't really make sense as the yuv coeffs are ints
<haasn>
maybe the flags get lost somehow ending up in it not being bitexact
<haasn>
oh indeed
<haasn>
that seems to be what's happening
<haasn>
removing the bitexact flag reproduces the same output without the patch
<haasn>
oh, I see the mistake
<elenril>
Daemon404: does china block all the signals and telegrams?
<Daemon404>
i assume so
<Daemon404>
anything they cant spy on
<ramiro>
in vsrc_testsrc.c:color_config_props(), the function parameter is named inlink, but it seems to configure an output. shouldn't it be named outlink, or am I just misunderstanding this?
<elenril>
ramiro: it should
<JEEB>
haasn: congrats on successful rubber ducky
<ramiro>
JEEB: I thought the same :D
<haasn>
so the issue is with how swscale handles flags
<haasn>
AV_OPT_TYPE_FLAGS is `int` only
<haasn>
but I tried making flags an int64_t
cubicibo has quit [Ping timeout: 256 seconds]
<haasn>
and according to docs it should be unsigned
jamrial has joined #ffmpeg-devel
<haasn>
do we not have a FLAGS64 type?
<JEEB>
I don't think so
<JEEB>
always the best state where you start hitting the limits of the system standard int bitwise vOv
<elenril>
haasn: should be easy enough to add
rvalue has quit [Read error: Connection reset by peer]
rvalue has joined #ffmpeg-devel
<haasn>
I guess I'll keep it as is for now
<haasn>
I'll soon move the scaler selection out of the flags
<haasn>
then we can reclaim all those bits
<elenril>
the grand bit reclamation project
<elenril>
"all the flags you can see around you were built on reclaimed bits"
<haasn>
JEEB: it's not that, the problem was about big endian
<haasn>
which is not happy when you dereference an int64 as an int
vipyne has joined #ffmpeg-devel
<JEEB>
ah
<haasn>
ramiro: just updated with the fix
<haasn>
I'd appreciate some reviews so we could maybe hopefully get this merged before the end of the week
mkver has joined #ffmpeg-devel
vipyne_ has joined #ffmpeg-devel
rvalue has left #ffmpeg-devel [#ffmpeg-devel]
<thardin>
looking a bit at jxl, I wonder how any encoder would ever be able to make full use of its features. and also, can't the predictors be made effectively arbitrarily complicated? sound potentially worse than j2k in terms of cycles per pixel
rvalue has joined #ffmpeg-devel
vipyne has quit [Quit: Leaving.]
vipyne_ has quit [Ping timeout: 276 seconds]
zsoltiv_ has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
<Lynne>
thardin: libjxl makes pretty full use
<Lynne>
yeah, its famous for having very flexible predictors
<Lynne>
which is why you can get pixel art in under a few kib
<Lynne>
the standard and reference implementation are float-only, which makes it even worse performance-wise
<Lynne>
I still have a decoder I need to finish
<Lynne>
fully fixed-point implementation, because floats have no place in anything video related, other than filtering
Workl3 has quit [Read error: Connection reset by peer]
ccawley2011__ has joined #ffmpeg-devel
ccawley2011_ has quit [Ping timeout: 264 seconds]
ccawley2011 has quit [Ping timeout: 252 seconds]
<elenril>
down with integer supremacism
<elenril>
vectorized quad precision floats for the people
<Lynne>
if they were guaranteed to be bitexact and cost 1 cycle per vector multiply...
Krowl has joined #ffmpeg-devel
<Daemon404>
Lynne, youre not gonna use the x87 fp emulator already in avcodec? ;)
<Lynne>
clearly we need x86inc.asm code which emulates SSE instructions with x87
lemourin has quit [Read error: Connection reset by peer]