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
bwu25 has quit [Ping timeout: 260 seconds]
iive has quit [Quit: They came for me...]
minimal has quit [Quit: Leaving]
_whitelogger has joined #ffmpeg-devel
j45 has joined #ffmpeg-devel
IndecisiveTurtle has quit [Quit: IndecisiveTurtle]
averne has quit [Read error: Connection reset by peer]
TheVibeCoder has joined #ffmpeg-devel
Everything has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
HarshK23 has quit [Quit: Connection closed for inactivity]
jamrial has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
<TheVibeCoder>
Lynne: thanks for DC mask fix, now i have working 1 version of proresraw
<Lynne>
no, its still broken
<Lynne>
non-hq is broken, and hq is inaccurate
<TheVibeCoder>
broken in what sense?
<Lynne>
non-compliant
<Lynne>
the qmatrix is off by 1 for hq
<Lynne>
15 >> 1 == 7, not 8
<fflogger>
[newticket] qubodup: Ticket #11656 ([ffmpeg] -bitexact does not remove metadata encoder tag in OGG Vorbis files) created https://trac.ffmpeg.org/ticket/11656
<ePirat>
if I use just videotoolbox it just silently falls back to software…
<jamrial>
hwaccel is to choose the hardware type, not codec+hardware
<jamrial>
so -hwaccel videotoolbox is correct
<jamrial>
make sure you're using the decoder that has the hwaccel hooks too
<ePirat>
hwaccel videotoolbox just silently falls back to sw
<jamrial>
does it even try to start the decoder? maybe your gpu doesn't support that stream
<ePirat>
well thats the whole point of forcing the hw path so I can actually have it fail when it doesnt work
<ePirat>
so I can see whats going on
BradleyS has quit [Read error: Connection reset by peer]
BradleyS has joined #ffmpeg-devel
<ePirat>
the way the hwaccels for VT are done is really annoying as it doesnt fit the API at all as its not a typical hwaccel
<galad>
well, at least it works, much better than the standalone qsv decoders for example, and all their missing features
<ePirat>
jamrial, forgot the entry in the hwaccels list
<Lynne>
ePirat: make sure to add the hwaccel to hwaccels.c
<ePirat>
Lynne, hwaccels.h you mean?
<Lynne>
yes
<ePirat>
I get 'Selecting decoder 'prores_raw' because of requested hwaccel method videotoolbox' but it never calls the init function for it even so I guess more is needed
<ePirat>
Ah, got the ifdef for the pixfmt wrong
System_Error has quit [Ping timeout: 244 seconds]
<Lynne>
ePirat: updated my repo to add all hooks
System_Error has joined #ffmpeg-devel
Everything has quit [Read error: Connection reset by peer]
Everything has joined #ffmpeg-devel
<Lynne>
ePirat: pushed another fix, this time the vulkan hwaccel is working and able to draw something on screen
<ePirat>
thanks, will try, got it to the point where it properly opens the videotoolbox accell at least
<ePirat>
well for now I just used the non-hw output path to see if I even get anything so the result is a clipping mess of what it should look like. Will try now to output actual videotoolbox frames and see what native pixfmt it gives me there
mkver has joined #ffmpeg-devel
<Lynne>
non-hw output? as in the videotoolbox, but in software mode?
<Lynne>
the C decoder should output something correct-looking
MisterMinister has joined #ffmpeg-devel
<ePirat>
Lynne, kCVPixelFormatType_444YpCbCr8BiPlanarVideoRange = '444v', /* Bi-Planar Component Y'CbCr 8-bit 4:4:4, video-range (luma=[16,235] chroma=[16,240]) … */ is what I get from VT
<ePirat>
Lynne, the C decoder produces something close but colors are weird green-ish compared to how it looks in quicktime
<TheVibeCoder>
cant you force singel plane gray with >8 bits at all?
<ePirat>
actually the code was still forcing some format……
<ePirat>
without forcing anything I get kCVPixelFormatType_64RGBAHalf = 'RGhA', /* 64 bit RGBA IEEE half-precision float, 16-bit little-endian samples */
<ePirat>
trying to make it give me kCVPixelFormatType_16VersatileBayer fails for some reason…
System_Error has quit [Remote host closed the connection]
<ePirat>
Ah, if I disable VT hwdec and then force kCVPixelFormatType_16VersatileBayer it works
<ePirat>
Lynne, I assume thats what you wanted?
<Lynne>
yeah, could you upload them somewhere?
<Lynne>
NUT supports bayer, btw
<ePirat>
Lynne, still need to figure out how to map that to a ffmpeg pixfmt
<Lynne>
AV_PIX_FMT_BAYER_RGGB16
<ePirat>
yeah but I need to make sure its RGGB first
<ePirat>
Ah there is kCVPixelBufferVersatileBayerKey_BayerPattern which seems to contain that info…
bsFFFFFF has quit [Ping timeout: 260 seconds]
mkver has quit [Ping timeout: 244 seconds]
System_Error has joined #ffmpeg-devel
MisterMinister has quit [Read error: Connection reset by peer]
MisterMinister has joined #ffmpeg-devel
Everything has quit [Ping timeout: 260 seconds]
j45 has quit [Ping timeout: 260 seconds]
<Lynne>
ePirat: by spec it should be RGGB16
<Lynne>
I assume, the bitstream carries no bayer order info, so it must be constant
j45 has joined #ffmpeg-devel
<ePirat>
Lynne, is av_get_pix_fmt_name not implemented for the bayer formats?
<Lynne>
no, it is
<ePirat>
ok then something else gets upset…
<ePirat>
Pixel format '(null)' is not supported
<Lynne>
hwcontext_videotoolbox?
<ePirat>
yes
MisterMinister has quit [Remote host closed the connection]
MisterMinister has joined #ffmpeg-devel
<ePirat>
Ah I think I found the issue
<ePirat>
videotoolbox_best_pixel_format
<ePirat>
hmm no even that should always return a valid one…
<ePirat>
Lynne, not really sure how the sw_format ends up as AV_PIX_FMT_NONE…
<Lynne>
for now, you can just dump it via fwrite directly to a file
<ePirat>
for the version0, I get ExtendedPixelsRight = 8;
<ePirat>
not entirely sure what that means as it does not match the padding when looking at the stride…
<ePirat>
so maybe those are actually included in the visible rect or something…
iive has quit [Ping timeout: 244 seconds]
MisterMinister has quit [Ping timeout: 248 seconds]
MisterMinister has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 252 seconds]
<IndecisiveTurtle>
Lynne: Do you have a suggestion on how to best parallelize the run-coded AC coefficient estimation https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/proresenc_kostya.c#L666 Because the codebook depends on previous run I'm not too sure how multiple threads can work on it, but it would be nice as the loop has a lot of iterations
MisterMinister has joined #ffmpeg-devel
Gramner has quit [Ping timeout: 252 seconds]
Gramner has joined #ffmpeg-devel
MisterMinister has quit [Read error: Connection reset by peer]
MisterMinister has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 245 seconds]
MisterMinister has joined #ffmpeg-devel
System_Error has quit [Ping timeout: 244 seconds]
MisterMinister has quit [Read error: Connection reset by peer]