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
ngaullier has quit [Remote host closed the connection]
ngaullier has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 248 seconds]
leo60228 has joined #ffmpeg-devel
vriska has quit [Read error: Connection reset by peer]
___nick___ has quit [Ping timeout: 276 seconds]
bbbccc has quit [Ping timeout: 260 seconds]
Traneptora_ has quit [Quit: Quit]
Traneptora has joined #ffmpeg-devel
Traneptora has quit [Remote host closed the connection]
Traneptora has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
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]
zsoltiv__ has quit [Ping timeout: 244 seconds]
zsoltiv has quit [Ping timeout: 276 seconds]
cone-926 has joined #ffmpeg-devel
<cone-926>
ffmpeg Andreas Rheinhardt master:289cb3beff8d: avcodec/cri,tdsc,tiff: Use ff_mjpeg_decoder directly
<cone-926>
ffmpeg Andreas Rheinhardt master:0d9b0015ba71: tools/target_dec_fuzzer: Remove mjpeg hack
<cone-926>
ffmpeg Andreas Rheinhardt master:0279eb9c9315: avcodec/imm5: Reference H.264/HEVC decoders directly
<cone-926>
ffmpeg Andreas Rheinhardt master:40544b485435: avcodec/ftr: Replace AVERROR_BUG that can be triggered
<cone-926>
ffmpeg Andreas Rheinhardt master:e73c59c6c334: tools/target_dec_fuzzer: Assert on AVERROR_BUG
zsoltiv_ has joined #ffmpeg-devel
zsoltiv has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
minimal has joined #ffmpeg-devel
bbbccc has joined #ffmpeg-devel
Thulinma has quit [Quit: Laters]
Thulinma has joined #ffmpeg-devel
witchymary has quit [Ping timeout: 244 seconds]
minimal has quit [Quit: Leaving]
cone-926 has quit [Quit: transmission timeout]
<toots5446>
Lynne, michaelni: any opinion on those ogg demuxer patches?
<Lynne>
if michaelni is happy I'll push them
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
<toots5446>
Great!
<haasn>
ramiro: made some sweeping refactors that should make your backend a lot happier
<haasn>
now compile() just returns a void (const SwsOpExec *exec, const void *priv, int x, int y, int x_end, int y_end); with arbitrary void *priv
<haasn>
you can do whatever you want and handle looping yourself in asm
<BradleyS>
re: the the twitter/x posts on tech companies cpu encoding using ffmpeg, my first thought was the benefits of software *de*coding which are numerous compared to hardware
<BradleyS>
anyway, i don't post on social media anymore
<BtbN>
jamrial: the HEVC-MV stuff is supposedly fixed in the 575 driver series
<BtbN>
need to check if the alpha layer stuff is fixed as well
ngaullier has quit [Remote host closed the connection]
MisterMinister has joined #ffmpeg-devel
<BtbN>
Lynne: did you test if that yuv420p10 thing actually works with nvenc, and produces correct colors?
<BtbN>
Cause to my knowledge, there is no format in nvenc that maps to FFmpeg AV_PIX_FMT_YUV420P10
<BtbN>
For 8 bit yuv, there's pretty much only NV12, YV12 and IYUV, which I think didn't exist back when nvenc was first added, cause otherwise that weird hack that is YUV420P in CUDA wouldn't be there.
microlappy has joined #ffmpeg-devel
<BtbN>
Though it was already there in the very first commit on the nv-codec-headers.
<BtbN>
Odd, then I don't remember why that hack is there
<BtbN>
Wonder if it can just be undone... Is that technically an ABI break?
<BtbN>
But I'm really not sure if this can be just changed...
<BtbN>
The only things it'd affect would be things also only accessing data[0] and expecting the entire data in there, like nvenc
<BtbN>
And people mixing library versions I guess
cosimone has quit [Remote host closed the connection]
<BtbN>
hm, it does actually not work properly when using hwupload
kasper93 has quit [Read error: Connection reset by peer]
<BtbN>
that's wild. This works when supplying nvenc a software frame, but not when hwuploading the same frame
minimal has joined #ffmpeg-devel
<Lynne>
BtbN: nope, I didn't
<BtbN>
Yeah, pretty much concluded it can't work by now sadly
<BtbN>
nvenc does not support that format
kasper93 has joined #ffmpeg-devel
kasper93 has quit [Ping timeout: 276 seconds]
kasper93 has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
kasper93 has joined #ffmpeg-devel
kasper93 is now known as Guest2031
Guest2031 has quit [Killed (tantalum.libera.chat (Nickname regained by services))]
<fflogger>
[editedticket] aarubui: Ticket #10843 ([ffmpeg] A -shortest-video option that acts just like -shortest but ignores audio) updated https://trac.ffmpeg.org/ticket/10843#comment:1
<jamrial>
BtbN: what was the problem again? missing paremeter sets for the higher layer?
<BtbN>
Their "fetch global headers" function not returning a lot of headers
<jamrial>
right
<jamrial>
i remember they said the SEI would only ever be output with the first packet and not with that function, too
<BtbN>
"There is a fix in the 575 driver which now returns the VPS, SPS (Layer 0), PPS (Layer 0), SPS (Layer 1), PPS (Layer1)."
mkver has quit [Ping timeout: 260 seconds]
witchymary has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
LainIwakura has joined #ffmpeg-devel
LainIwakura has quit [Ping timeout: 240 seconds]
LainIwakura has joined #ffmpeg-devel
witchymary has quit [Remote host closed the connection]
<cone-851>
ffmpeg Andreas Rheinhardt master:3e21df2353ac: avcodec/huffman: Switch to ff_vlc_init_from_lengths()
<cone-851>
ffmpeg Andreas Rheinhardt master:e946ba64bf71: avcodec/vp6: Don't reload unnecessarily often in get_vlc2()
<cone-851>
ffmpeg Andreas Rheinhardt master:be7495c32ce6: avcodec/vp6: Use fewer number of bits in run VLCs
<cone-851>
ffmpeg Andreas Rheinhardt master:a2acb46a6e38: avcodec/vp6: Forward error codes
<cone-851>
ffmpeg Andreas Rheinhardt master:6676038b23bd: avcodec/cbs: Use put_bits63()
<cone-851>
ffmpeg Andreas Rheinhardt master:d60445258ce5: avcodec/webp: Check more directly for invalid codes
<cone-851>
ffmpeg Andreas Rheinhardt master:e0df21b8c379: avcodec/webp: Switch to ff_vlc_init_from_lengths()
<cone-851>
ffmpeg Andreas Rheinhardt master:1fc0ee484f09: avcodec/webp: Check before allocations
<cone-851>
ffmpeg Andreas Rheinhardt master:e0a05456cd62: avcodec/webp: Combine allocations
<cone-851>
ffmpeg Andreas Rheinhardt master:c33f16db1b0e: avcodec/webp: Avoid loop
<cone-851>
ffmpeg Andreas Rheinhardt master:7c7e161437f2: avcodec/magicyuv: Set properties via AVPixFmtDescriptor
<cone-851>
ffmpeg Andreas Rheinhardt master:122f86d85987: avcodec/magicyuv: Simplify check for RGB
<cone-851>
ffmpeg Andreas Rheinhardt master:6dd83fab44ca: avformat: Use ffio_read_size() where appropriate
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
Luna_Rabbit has quit [Quit: Do you believe in magic?]
Moon_Rabbit has joined #ffmpeg-devel
slpslp has joined #ffmpeg-devel
IndecisiveTurtle has joined #ffmpeg-devel
<haasn>
ramiro: as a compromise, I settled on making the signature (OpExec *, void *priv, int pixels)
<haasn>
handling basically a single line
<haasn>
we can still do the fusing into a big line in the wrapper, and this frees asm from having to deal with strides and subsampled formats and whatnot
<haasn>
most of the benefit comes from the per-line loop anyway
<haasn>
and with the fusing optimization that's the only loop anyway
<haasn>
OpExec continues holding the x/y coordinate as always, for dither; but the entry point can decide whether to keep track of the x component or not
<haasn>
with avx2 we actually don't care about the x coordinate for dithering at all because the 16x16 dither matrix is smaller than the block size