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 7.0.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
feiw has quit [Ping timeout: 265 seconds]
cone-890 has quit [Quit: transmission timeout]
mkver has quit [Ping timeout: 245 seconds]
thilo has quit [Ping timeout: 264 seconds]
thilo has joined #ffmpeg-devel
lexano has quit [Ping timeout: 258 seconds]
IndecisiveTurtle has quit [Ping timeout: 265 seconds]
<JEEB>
there's also the `fate-rsync` target which syncs things up
<lain__>
Ok. Interesting
<lain__>
Thank you.
<JEEB>
so for example if you have built FFmpeg already, in the build root you can do something like `make fate-rsync SAMPLES=../../fate-suite/`
<JEEB>
and that should utilize rsync to sync all of the FATE suite so you can run the tests
<JEEB>
(running the tests would then be `make fate SAMPLES=../../fate-suite/` or so
<JEEB>
the SAMPLES path of course just being a reference to somewhere that you want that stuff to end up at, and not get overwritten by build roots being created or destroyed
<shinjite>
avformat_open_input seems to struggle to work with a win32 pipe. Both code is C++ externed correctly. It reads the magic AVI RIFF key
<shinjite>
and 1920x1080 but dies at reading pixel fmt(which in mingw build for windows picks up yuv or int of 0) any idea why mvsc toolchain sux
<shinjite>
I read pipe in first and dumped result to file for pipe error and got same exact avi out and it ran with not artificats
<shinjite>
im suspecting a warning is to blame somewhere using a int instead of a size_t and something is being cut off. Just asking if anyone knows whats the deal.
jamrial_ has joined #ffmpeg-devel
jamrial has quit [Ping timeout: 252 seconds]
cubicibo has quit [Ping timeout: 256 seconds]
nto has quit [Server closed connection]
nto has joined #ffmpeg-devel
shinjite has quit [Quit: Client closed]
Krowl has quit [Read error: Connection reset by peer]
ngaullie has quit [Read error: Connection reset by peer]
jamrial_ has quit []
jamrial has joined #ffmpeg-devel
MisterMinister has joined #ffmpeg-devel
<courmisch>
what did he do this time
Sean_McG has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
wyatt8740 has quit [Server closed connection]
wyatt8740 has joined #ffmpeg-devel
<cone-941>
ffmpeg Rémi Denis-Courmont master:7744c0824080: lavc/h264dsp: R-V V add_pixels4 and 8-bit add_pixels8
<cone-941>
ffmpeg Rémi Denis-Courmont master:3002310b7070: lavc/h264dsp: R-V V high-depth add_pixels8
* elenril
facepalms
<elenril>
ffv1 "frame threading" is worse than I thought
SystemError has quit [Remote host closed the connection]
<courmisch>
I demand a C version of ff_h264_loopfilter_strength, or else!!!11!one!
<cone-941>
ffmpeg Zhao Zhili master:cb9c98da16e8: avformat/file: guard fd_dup by FD_PROTOCOL or PIPE_PROTOCOL
<cone-941>
ffmpeg Zhao Zhili master:8ca60e14d200: configure: Fix Apple framework dependencies in .pc file
<jdarnley>
Damn! Anton was sitting on this commit for 3 years 11 months
<elenril>
what commit?
<jdarnley>
1e9615c5d4
<elenril>
isn't that in master?
<jdarnley>
It is
<jdarnley>
*was*
<jdarnley>
*was* sitting
<elenril>
it's not *that* long
<jdarnley>
I need to make git blame show commit dates not author dates
<elenril>
see e.g. 086a8048061bf9fb4c63943f6962db48175f655c
<jdarnley>
Much more!
<jdarnley>
Although that one says it got subsequent work
fmlatghor has quit [Server closed connection]
fmlatghor has joined #ffmpeg-devel
<jamrial>
jdarnley: yeah, but the general api didn't change much, just helpers and such
Krowl has quit [Read error: Connection reset by peer]
<Sean_McG>
courmisch: I agree with you that the PPC arch selection stuff in configure is not great, but I have to be honest I'm a bit anxious to change it given how long it has been around and how easy it might be to break a working config that we don't have hardware to test on
<Sean_McG>
also, I'm putting the VSX acceleration stuff aside for now, I've been struggling with the math too much.
<ramiro>
does nobody have a real ppc fate box? I used to have one back in the day.
<Sean_McG>
ramiro: there is a broken G4 Mini on my desk that I have been attempting to get booting with not much luck
<Sean_McG>
but I may put out a "wanted to buy" posting locally to see if someone has other G4s or maybe even a G5
<Sean_McG>
(although I am told the fans on a G5 are obnoxious)
<ramiro>
iirc mine was from 2005. do we still care about such old hardware?
<Sean_McG>
the AltiVec code is still used on modern POWER machines.
<ramiro>
then maybe we should get a modern power machine :)
<Sean_McG>
a RaptorCS Blackbird is $5K USD
<Sean_McG>
I've been flip-flopping on pulling the trigger on one of those for months now
<Sean_McG>
(that said, the Blackbird and the Talos are phenomenal machines)
<Sean_McG>
I think we'll be hearing about POWER11 within the next few months though -- and if I understand what I am reading the new machines will have larger vector width (AltiVec and VSX are only 128-bit)
<Sean_McG>
it might be worth to remove the "big-endian only" AltiVec stuff as POWER has largely transitioned to little-endian by default (at least on Linux), even though it can still be used in BE-mode
<Sean_McG>
if that is desired I can go ahead with it
<Lynne>
I wouldn't object to using the project's funds for a machine
<Lynne>
checkasm for altivec was broken for 4 years or so
kasper93 has quit [Ping timeout: 258 seconds]
<Sean_McG>
Andreas did a great job fixing it up
<Sean_McG>
s/Andreas/mkver/
<Lynne>
hey, I contributed a few patches too after a friend got one
<Sean_McG>
and you did :)
<Lynne>
I don't think power9 is that relevant though, after raptor multiplied all prices by 10 or so
<Sean_McG>
IBM decided not to share POWER10, so I think the eco-system is busted again
<Sean_McG>
and I suspect it will be the same for POWER11
<Sean_McG>
and yes, they haven't made new POWER9 chips since 2018, so there is that... but I still think the 8c4t is sexy as hell
kasper93 has joined #ffmpeg-devel
<Lynne>
sexy? have you see Zen 5's technical documents?
<Lynne>
dual decode pipeline! 6 ALUs! 6!
<Lynne>
you can chew through simple loops with tens of instructions in one clock cycle thanks to speculation
<Sean_McG>
but not SMT4 yet, is it?
<Lynne>
I find SMT4 offensive
* Sean_McG
giggles.
<Sean_McG>
anyways, no harm done.
<Lynne>
Zen 5 lets you use ALL of a CPU when running a single thread
<Lynne>
how badly must code be written to not use ALL of a CPU?
<Sean_McG>
someone who's never heard of threads?
<Sean_McG>
or an algorithm that is more memory-bound than CPU-bound
<Sean_McG>
which, if I recall correctly AMD has struggled with
<Sean_McG>
(full disclosure: my desktop is a Threadripper 2950X)
<elenril>
Lynne: true fact: beause of a lizardpeople conspiracy only 1% of transistors on a chip are actually computing anything
<Sean_McG>
hahah
<elenril>
the rest are drawing hail-hydra patterns on silicon
<Lynne>
its bad news if 100% of the transistors are switching, you'll fry your CPU in one clock cycle
<elenril>
I'd share an authoritative reference, but this line is too short
<elenril>
that's what they want you to think
marcj has quit [Ping timeout: 246 seconds]
<elenril>
the so-called second law of thermodynamics is more like a guidance
<haasn>
it's possible to reverse the arrow of time, but it requires sacrificing magical girls
<Sean_McG>
"the more you put things together, the more they fall apart"
<Sean_McG>
haasn gets it
<elenril>
／人◕ ‿‿ ◕人＼
<Sean_McG>
now I want to re-watch Madoka Magica
<JEEB>
I should grab the movies at some point, I don't think I watched anything after the original TV series
Krowl has joined #ffmpeg-devel
<elenril>
I was pleasantly surprised by the last one
<elenril>
first two are just a summary fo the series
<AMM>
Rebellion is a must, I don't think the recaps changed much
<Lynne>
didn't like the TV series
<JEEB>
yea I guess it had the recap movies
<JEEB>
the new stuff is what I mostly was interested in
<cone-941>
ffmpeg Michael Niedermayer master:af9935835335: avcodec/vc2enc: Fix overflows with storing large values
<cone-941>
ffmpeg Michael Niedermayer master:7147c3c911f4: avcodec/ratecontrol: Handle wanted bits overflow
<cone-941>
ffmpeg James Almer master:b248dace929e: avformat/iamf_parse: keep substream count consistent
<cone-941>
ffmpeg Michael Niedermayer master:664fbfb9ac55: avcodec/mscc: move frame allocates to later
<cone-941>
ffmpeg Michael Niedermayer master:8a4e5e0e15fc: tools/target_dec_fuzzer: Adjust threshold for RV30
<cone-941>
ffmpeg Michael Niedermayer master:a84fbd747119: avcodec/j2kenc: Merge dwt_norm into lambda
<cone-941>
ffmpeg Michael Niedermayer master:69e90491f15d: avcodec/utvideoenc: Use unsigned shift to build flags
<cone-941>
ffmpeg Michael Niedermayer master:0993ef675f06: avcodec/mpeg12enc: Use av_rescale() in vbv_buffer_size computation
<Sean_McG>
michaelni: cheers
<Sean_McG>
(for the j2kenc one)
<jamrial>
elenril: the ml seems to not like your patchset much or something. i see a dozen in the archive, and only five in my mailbox
<elenril>
seems ML decided to stop talking to me
jamrial has quit [Read error: Connection reset by peer]
jamrial_ has joined #ffmpeg-devel
<courmisch>
elenril: gimme a reason to reject a patch of yours and I'll oblige
<courmisch>
and cultivate my smartass jerk image, two birds with one stone
<Sean_McG>
oh dear.
georgereynolds8 has quit [Quit: Ping timeout (120 seconds)]
georgereynolds8 has joined #ffmpeg-devel
<Lynne>
elenril: same
<Lynne>
I thought it was my mail server, which resulted in me breaking it and having to restore from a backup, not what I wanted to do today
<jamrial_>
looks like it's solved now. i see the entire set
BBB has quit [Server closed connection]
BBB has joined #ffmpeg-devel
<elenril>
commit message for 27/39 is dedicated to mkver
<mkver>
?
<mkver>
Why did you not ask me for my branch?
<elenril>
last I asked you did not answer at all
<mkver>
When?
<elenril>
a few weeks ago IIRC
<mkver>
I don't remember that at all.
<courmisch>
so why won't libreoffice open paths from the CLI anymore
<courmisch>
this has got to be a violation of my rights as a Unix shell user
<JEEB>
now to wait for the SCO people asking whether you have a license for that Unix(TM)
<Sean_McG>
derp
<courmisch>
JEEB: you mean Oracle?
<JEEB>
possibly. whichever corp now has Unix under its belt
<JEEB>
SCO I think was it at some point?
<courmisch>
well there have been two related SCO's.
<courmisch>
One that got bought by Sun and the eviler one
<JEEB>
right'o :D
Krowl has quit [Read error: Connection reset by peer]
<courmisch>
that eviler one, i.e. Caldera, was liquidated
<JEEB>
so now we just have solaris... wait, that is now forked, isn't it?
<Sean_McG>
it's effectively defunct
<courmisch>
I think Oracle only really cared about Java, and maybe ZFS?
<Sean_McG>
Oracle only cares about ZFS insofar as it benefits their database product
<JEEB>
not sure how much ZFS development there actually was in its "upstream"
<Sean_McG>
or whatever buys Larry a new yacht
marcj has joined #ffmpeg-devel
<JEEB>
yea, larry needs many more
<courmisch>
yachts are so last century
<courmisch>
any self-respecting tech mogul builds space rockets nowadays
<microchip_>
Novell holds the UNIX licenses
<Sean_McG>
ohhhh right
<Sean_McG>
well, they are gone since 2014
<Sean_McG>
the Internet says it basically belongs to The Open Group now
<Sean_McG>
since August 2021
<Lynne>
great, my mail server is bouncing mails from the ML
<Lynne>
why is email so horrible?
<Sean_McG>
and yet it is the killer app of the Internet
<elenril>
Lynne: PEBKAC?
<elenril>
email has a lot of cruft, but everything else is worse
<Lynne>
even astronomers got rid of it and replaced it with something horrible yet better - kafka
iive has joined #ffmpeg-devel
HarshK23 has quit [Quit: Connection closed for inactivity]
Krowl has joined #ffmpeg-devel
<Traneptora>
texting be like: yo can you come over, nah busy, aight np
<Traneptora>
email be like: enclosed please find an invitation for your arrival at my location....
Traneptora has quit [Quit: Quit]
<Lynne>
only both the sender and the receiver are members of the british aristocracy, and they need at least 10 tea parties to discuss it before cordially accepting a message
<Lynne>
ten tea parties of talking about the weather, lord allenby, lord allenby's wife's brother's affair, sir johnathan's new statue on shaftesbury avenue and lord henford's recent cruise to egypt
Krowl has quit [Read error: Connection reset by peer]
* Sean_McG
giggles.
MrZeus has joined #ffmpeg-devel
MrZeus_ has joined #ffmpeg-devel
<thardin>
can someone tell me what the actual purpose of srt:// is? it seems strictly worse than a well tweaked TCP implementation
MrZeus has quit [Ping timeout: 252 seconds]
<Lynne>
mpegts + fec
<thardin>
those aren't transport protocols
<Lynne>
that's what it was designed to carry
<elenril>
thardin: welcome to webscale?
<thardin>
or is it possible to tweak it to guaranteed in-order delivery?
<elenril>
reinventing strictly worse implementations of the wheel, since 2016
<thardin>
there's something called "Too-Late Packet Drop" that seems possible to disable
<thardin>
in this specific application it makes little sense to drop packets because the stream is split up into 30 second chunks or so sent to an AI backend, and the entire thing has maybe 2-3 minutes of latency
<Lynne>
you get frames out of order?
<Lynne>
shouldn't the decoder figure out when to decode them based on DTS?
cone-941 has quit [Quit: transmission timeout]
<thardin>
I doubt SRT is frame aware. I get garbage packets
<thardin>
I guess it fills missing data with NUL
<thardin>
-tlpktdrop 0 perhaps
<thardin>
or &tlpktdrop=0 on the URL even
<JEEB>
elenril: but SRT (protocol) was not a web thing :D
<Lynne>
use avtransport, rule number 1 is you will never ever get garbage
BtbN has quit [Server closed connection]
BtbN has joined #ffmpeg-devel
<thardin>
so far setting that seems to work. it seems with SRT you just have to know the right incantation
<thardin>
I sense.. another niche to get into
Kwiboo has quit [Server closed connection]
Kwiboo has joined #ffmpeg-devel
SystemError has joined #ffmpeg-devel
osvein has quit [Server closed connection]
osvein has joined #ffmpeg-devel
ergZay has quit [Quit: This computer has gone to sleep]
witchymary has quit [Remote host closed the connection]
witchymary has joined #ffmpeg-devel
witchymary has quit [Remote host closed the connection]
witchymary has joined #ffmpeg-devel
graphitemaster has quit [Server closed connection]
graphitemaster has joined #ffmpeg-devel
jamrial_ has quit [Read error: Connection reset by peer]
jamrial has joined #ffmpeg-devel
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
psykose_ has joined #ffmpeg-devel
psykose has quit [Ping timeout: 244 seconds]
psykose_ is now known as psykose
ccawley2011 has quit [Read error: Connection reset by peer]
MrZeus_ has quit [Ping timeout: 252 seconds]
Sebastinas has quit [Server closed connection]
Mikhail_AMD has joined #ffmpeg-devel
Sebastinas has joined #ffmpeg-devel
MikhailAMD has quit [Ping timeout: 246 seconds]
jamrial has quit [Read error: Connection reset by peer]
jamrial_ has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
aaabbb has quit [Server closed connection]
aaabbb has joined #ffmpeg-devel
michaelni has quit [Ping timeout: 246 seconds]
thardin has quit [Server closed connection]
thardin has joined #ffmpeg-devel
witchymary has quit [Remote host closed the connection]
witchymary has joined #ffmpeg-devel
jamrial_ has quit [Read error: Connection reset by peer]