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
lemourin has quit [Ping timeout: 265 seconds]
lemourin has joined #ffmpeg-devel
thilo_ has quit [Ping timeout: 260 seconds]
thilo_ has joined #ffmpeg-devel
thilo_ has quit [Changing host]
thilo_ has joined #ffmpeg-devel
HarshK23 has quit [Quit: Connection closed for inactivity]
vipyne has joined #ffmpeg-devel
vipyne has quit [Client Quit]
<Marth64>
chapter seeking seem to be working better too at least in mpv
thilo_ has quit [Ping timeout: 252 seconds]
thilo_ has joined #ffmpeg-devel
wyatt8750 has quit [Remote host closed the connection]
wyatt8740 has joined #ffmpeg-devel
wyatt8750 has joined #ffmpeg-devel
wyatt8740 has quit [Ping timeout: 252 seconds]
arch1t3cht7 has joined #ffmpeg-devel
arch1t3cht has quit [Ping timeout: 265 seconds]
arch1t3cht7 is now known as arch1t3cht
jamrial has quit []
Sean_McG has quit [Quit: leaving]
Martchus_ has joined #ffmpeg-devel
Martchus has quit [Ping timeout: 244 seconds]
pross has quit [Ping timeout: 252 seconds]
cone-694 has joined #ffmpeg-devel
<cone-694>
ffmpeg Martin Storsjö master:b9145fcab27d: swscale: Fix aarch64 and i386 compilation failures
ngaullier has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
HarshK23 has joined #ffmpeg-devel
<elenril>
Marth64: av_malloc() is not magic, it's just malloc with a funny name
<Lynne>
and with alignment
<elenril>
right
<elenril>
but it does not do any tracking or auto-free or anything
ccawley2011 has joined #ffmpeg-devel
<Traneptora>
what does av_free do that free doesn't do?
feiw has joined #ffmpeg-devel
<Lynne>
on some platforms aligned pointers may not be passed to regular free, I think
Krowl has quit [Read error: Connection reset by peer]
<Marth64>
all good, i freed the allocation
<wbs>
Traneptora: also we used to support manually aligning allocations, i.e. we'd malloc(size+align) and return an aligned pointer, and then av_free would dig up the original pointer and free that
<Traneptora>
that's what mpv talloc does
<wbs>
yeah
<wbs>
we got rid of it in 4fb311c804098d78e5ce5f527f9a9c37536d3a08
<elenril>
we could be using talloc I guess
<elenril>
but then again memleaks are rarely a big problem for us
___nick___ has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
<Marth64>
i figured av_malloc wasn't magic but more imagined about magic in FF_INFMT_FLAG_INIT_CLEANUP
<Marth64>
also haven't slept in about 20 hours so braindead
Kei_N_ has joined #ffmpeg-devel
Kei_N has quit [Read error: Connection reset by peer]
cone-694 has quit [Quit: transmission timeout]
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 260 seconds]
j45_ is now known as j45
j45 has joined #ffmpeg-devel
j45 has quit [Changing host]
Krowl has quit [Read error: Connection reset by peer]
Kei_N_ has quit [Read error: Connection reset by peer]
IndecisiveTurtle has quit [Ping timeout: 248 seconds]
IndecisiveTurtle has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 276 seconds]
Krowl has quit [Read error: Connection reset by peer]
IndecisiveTurtle has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 265 seconds]
___nick___ has quit [Ping timeout: 248 seconds]
microlappy has joined #ffmpeg-devel
iive has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 248 seconds]
ngaullier has quit [Ping timeout: 252 seconds]
rvalue- is now known as rvalue
<thardin>
did some encoding experimentation earlier today. my idea of rotating macroblocks before dct is sadly not as good as I hoped. but in some cases it is indeed beneficial
<thardin>
it did bring to mind one thought: why do so few codecs allow switching transforms on a per-MB basis? I think VVC has something like it at least
<thardin>
some older papers combine horizontal dct with vertical dpcm
<iive>
thardin, what do you mean by rotating MB? iirc some of the early mpeg2/4 codecs combined the zig-zag with reordering of the coefficients, in a way to simplify the first pass of DCT.
<thardin>
consider an MB with a diagonal line in it
<thardin>
a better word may be shearing. shear the rows (or columns) of it such that diagonal shapes become vertical ones. then apply vertical DCT
<thardin>
and optionally a horizontal DCT after that, depending on whether it's beneficial or not
<iive>
oh, so rotating the texture before encoding.
<thardin>
yeah
<thardin>
then you get an asymmetry that you could potentially exploit. for example, you can have heavier compression in the direction of the diagonal line
vipyne has quit [Quit: Leaving.]
vipyne has joined #ffmpeg-devel
beastd has joined #ffmpeg-devel
microlappy has quit [Quit: Konversation terminated!]
beastd has quit [Remote host closed the connection]
beastd has joined #ffmpeg-devel
ngaullier has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
MikhailAMD has joined #ffmpeg-devel
cone-647 has quit [Quit: transmission timeout]
beastd has quit [Ping timeout: 252 seconds]
vipyne has quit [Quit: Leaving.]
<frankplow>
thardin: How would you perform the transformation reversibly?
<frankplow>
If you are doing this test on a codec with angular intra prediction modes, I think that deals well with a lot of this
feiw has quit [Ping timeout: 252 seconds]
feiw has joined #ffmpeg-devel
vipyne has joined #ffmpeg-devel
vipyne has quit [Client Quit]
philipl_1 is now known as philipl
Traneptora has quit [Quit: Quit]
System_Error has joined #ffmpeg-devel
beastd has joined #ffmpeg-devel
<thardin>
intra prediction has the problem of dependency between block
<thardin>
if you wrap the edges then the transform becomes reversible
galad has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
IndecisiveTurtle has joined #ffmpeg-devel
Raz- has quit [Quit: Leaving]
Raz- has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]