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
<BtbN>
I got a 7950X if you want to test something on Zen4, kierank
jamrial has quit [Read error: Connection reset by peer]
microlappy has quit [Quit: Konversation terminated!]
microlappy has joined #ffmpeg-devel
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 252 seconds]
tufei has joined #ffmpeg-devel
tufei has quit [Remote host closed the connection]
rvalue- is now known as rvalue
<BBB>
Traneptora: if sws actually implements the things sws does, that's reasonable, yes
<BBB>
same for colormatrix
<BBB>
but when it was written, sws didn't do these things :)
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
microlappy has quit [Quit: Konversation terminated!]
tufei has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
tufei has quit [Quit: Leaving]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
tufei has joined #ffmpeg-devel
tufei has quit [Ping timeout: 264 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
zsoltiv_ has joined #ffmpeg-devel
zsoltiv has joined #ffmpeg-devel
tufei has joined #ffmpeg-devel
<haasn>
okay
<haasn>
I rewrote the color space conversion core as a generalized 5x5 linear operation pipeline
<haasn>
fusing every single operation together while constructing the pipeline
<haasn>
and then at the end, checking to see which coefficients are still nonzero and then dispatching a dedicated function for that pattern of coefficients
<haasn>
so we have e.g. a fast path for a purely diagonal matrix (compiling down to scalar-vector multiplication), or a 3x3 subset of the 5x5 affine transform
<haasn>
this approach works very well at letting us fuse as many operations together effortlessly
<haasn>
while still allowing us to dispatch to however many dedicated fast paths we end up deciding is worth it on a given platform
<haasn>
and satisfying my autismal quest to make the minimal viable set of operations as small as possible
<haasn>
currently the requirement is only 7 operations :)
<haasn>
read, write, swap_bytes, swizzle, shift, convert, and the new 7th generalized mega-op: linear
Marth64 has joined #ffmpeg-devel
<haasn>
expect to see 1-2 more for filtering (including scaling), some for LUTs, at least one for handling of nonlinear scaling (constant luminance colorspaces)
Marth64 has quit [Remote host closed the connection]
Marth64 has joined #ffmpeg-devel
Thulinma has quit [Quit: Laters]
Thulinma has joined #ffmpeg-devel
tufei has quit [Quit: Leaving]
tufei has joined #ffmpeg-devel
<fflogger>
[newticket] zcybercomputing: Ticket #11461 ([ffmpeg] Corruption of 608 captions data on extraction) created https://trac.ffmpeg.org/ticket/11461
tufei has quit [Remote host closed the connection]
Mirarora has quit [Quit: Mirarora encountered a fatal error and needs to close]
<BtbN>
Isn't that the third or 4th ticket for that exact thing?
<Marth64>
this one is different
<Marth64>
but CC issues have been hot lately
<Marth64>
past 2 issues were on selecting the right stream (auto detection is not great), this is i believe an issue with the SCC muxer i've seen before
<Marth64>
I've seen SubtitleEdit will choke on this output
<Marth64>
and there was 1 issue about the FIFO that ended up being user error
<IndecisiveTurtle>
Lynne: I think I'm done with all review comments I can recall, storage images are working nicely for me too with both haar and legall wavelets. Could you take a quick look at the changes, or comment on github about anything else that comes to mind https://github.com/raphaelthegreat/FFmpeg/commit/1785147e
<IndecisiveTurtle>
Took a week longer cause I started playing persona 5 sorry
<IndecisiveTurtle>
After everything is done I will split it appropriately and wait for your GRAY32 commit to get merged separately