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 7.0.2 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
yigithanyigit has joined #ffmpeg-devel
arch1t3cht7 has joined #ffmpeg-devel
arch1t3cht has quit [Ping timeout: 265 seconds]
arch1t3cht7 is now known as arch1t3cht
yigithanyigit has quit [Ping timeout: 244 seconds]
thilo has quit [Ping timeout: 272 seconds]
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
thilo has joined #ffmpeg-devel
MrZeus has quit [Ping timeout: 244 seconds]
blb has quit [Ping timeout: 264 seconds]
blb has joined #ffmpeg-devel
yigithanyigit has joined #ffmpeg-devel
yigithanyigit has quit [Ping timeout: 248 seconds]
IndecisiveTurtle has quit [Ping timeout: 265 seconds]
rvalue has quit [Ping timeout: 265 seconds]
rvalue has joined #ffmpeg-devel
yigithanyigit has joined #ffmpeg-devel
yigithanyigit has quit [Ping timeout: 272 seconds]
rvalue has quit [Ping timeout: 248 seconds]
vipyne has joined #ffmpeg-devel
BiteriaFedoral has joined #ffmpeg-devel
rvalue has joined #ffmpeg-devel
mkver has quit [Ping timeout: 252 seconds]
BiteriaFedoral has quit [Changing host]
BiteriaFedoral has joined #ffmpeg-devel
jamrial has quit []
jarthur has quit [Quit: jarthur]
vipyne has quit [Quit: Leaving.]
Martchus_ has joined #ffmpeg-devel
Martchus has quit [Ping timeout: 265 seconds]
SystemError has quit [Remote host closed the connection]
yigithanyigit has joined #ffmpeg-devel
BiteriaFedoral has quit [Quit: Client closed]
Kei_N_ has quit [Read error: Connection reset by peer]
yigithanyigit has quit [Ping timeout: 244 seconds]
Kei_N has joined #ffmpeg-devel
<elenril>
BBB: don't really understand the question
<llyyr>
Lynne: it might be a good idea to remove the av_log call since there's only one that causes the problem
<Lynne>
llyyr: which one
<llyyr>
since it seems like setting AVClass might be more involved
<llyyr>
the one in ff_vk_get_pooled_buffer
<Lynne>
I'm still waiting for elenril to take a look at it, but he seems busy
<haasn>
Oh yeah you want to adjust the gamma to the actual peak
IndecisiveTurtle has joined #ffmpeg-devel
<haasn>
There's also some extra x_in * (1 - b) + b going on in my code
<haasn>
don't remember why
<haasn>
where b is some coefficient that depends on the contrast
<haasn>
I think that might be done in your code too though, just at a later stage
<haasn>
wheres my code prebakes the constant and does it at the beginning :shrug:
<rajivharlalka>
For the gamma, I took it as 1.2 as an approximate value for now, and would finetune to exact when I get similar results as I believe the extra terms would change the value from second decimal
<rajivharlalka>
I cross verified the formula from the ITU-BT Rec 2100 standard , didn’t find anything explicit different.
cone-785 has joined #ffmpeg-devel
<cone-785>
ffmpeg J. Dekker master:03f26549cd8e: checkasm: print only results to stdout
<cone-785>
ffmpeg J. Dekker master:42528ff83535: checkasm: add csv/tsv bench output
<cone-785>
ffmpeg J. Dekker master:d0986709a8b3: checkasm: improve print format
<cone-785>
ffmpeg J. Dekker master:e758b24396a4: checkasm: add wildcompares for test & functions
<elenril>
Lynne: I see FFVulkanContext does have an AVClass at the beginning
<elenril>
is it not set?
<elenril>
...apparently not in most cases
<elenril>
if the idea is that some external user of that context sets it to some random value, then that's broken
<elenril>
I think you should make it dynamically allocated, and set its class in the constructor
<elenril>
then the caller can pass a parent logging context, that would be used via parent_log_context_offset
<elenril>
IMO that is The Right Thing To Do(tm)
kasper93_ is now known as kasper93
rvalue has quit [Read error: Connection reset by peer]
rvalue has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<wbs>
ramiro: nice, I think this looks fine now. (many of the things I suggested are things that stand out to myself when I read it, thinking "oh, that looks odd", but then I remember the purpose of it ;-)
<BBB>
elenril: IIUC with mv, a "frame" (in the h264 sense) can contain multiple views? and your implementation means we select one and output that one (only), I think?
<BBB>
elenril: so I asked whether it was meaningful/possible to output multiple views of the same frame together in one AVFrame
<wbs>
ramiro: so if some of the suggestions, like using two separate pointers for launching two ld3 at the same time, don't really give a benefit, you can just scrap it too. e.g. for the second line, it could be possible to just do regular ld3, [xN], #24, but interleaved with other things, avoiding the need for the separate registers
<elenril>
BBB: the HEVC spec never treats different views bundled together as "one frame"
<elenril>
not sure about h264
<elenril>
in my implementation, AVFrame = one view
vipyne has quit [Quit: Leaving.]
<BBB>
elenril: ok, I understand. so is it possible to "retrieve" more than one view at the same time? or do we have to choose a single view and the decoder discards the others?
<BBB>
elenril: (tnx for explaining)
av500 has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
<jamrial>
Lynne: can you look at ticket #11145?
<elenril>
BBB: of course, that's the whole point
<elenril>
you get all the views you want
<elenril>
(as long as it's not more than 2)
<elenril>
you just get them in separate frames
<BBB>
ok, then I misunderstood the patch, that's the best-case scenario
<BBB>
awesome
<BBB>
\o/
<BBB>
nice work
<elenril>
what did you misundertand?
<BBB>
I thought you had to select one view using the api
<BBB>
you=the user/application
<elenril>
you select a set of views you want
<elenril>
(hence the patches for array option handling)
<BBB>
ok, sounds cool. sorry for the noise
<darkdrgn2k>
@BBB thanks for the help yesterday, turned out i was using the wrong binary after all, your suggestion worked great. Thanks!
bbbccc has quit [Ping timeout: 252 seconds]
cone-785 has quit [Quit: transmission timeout]
sm2n has quit [Remote host closed the connection]
thardin_ is now known as thardin
sm2n has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<jamrial>
<@elenril> (as long as it's not more than 2) <-- something something black cars
kasper93 has quit [Ping timeout: 264 seconds]
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg-devel
<elenril>
black = best color
<elenril>
also, the 2 view limitation is only in the implementation, not the API
Krowl has quit [Read error: Connection reset by peer]
<ramiro>
wbs: btw there are a few places where we can reorder smlal (and uadalp) for better performance. I still have to benchmark all of it, but in swscale I got a 1.3x speedup.
<Daemon404>
elenril, lies.
<Daemon404>
it is green.
<BBB>
darkdrgn2k: cool, no prob
HarshK23 has joined #ffmpeg-devel
<Lynne>
elenril: I thought users weren't allowed to allocate one themselves?
<elenril>
one what?
Krowl has joined #ffmpeg-devel
<Lynne>
avclass
<elenril>
I don't mean allocating an AVClass
<elenril>
just the context
<Lynne>
how do I set its class in the constructor?
<elenril>
define an AVClass for FFVulkanContext in lavu/vulkan.c, then have the constructor set FFVulkanContext.class = &vulkan_context_class
<Marth64[m]>
good day
<elenril>
then add a void *parent_log to FFVulkanContext, and set vulkan_context_class.parent_log_context_offset to offsetof(FFVulkanContext, parent_log)
<elenril>
and pass a parent logging context to this new constructor, which would assign it to parent_log
<elenril>
that way logging gets propagated from a FFVulkanContext to its parent
<elenril>
or maybe FFVulkanContext does not actually need to be dynamically allocated, if the functions that work on it are not cross-library