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 6.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<ePirat>
durandal_1707, regarding xfade tests, while we can't check the frame content IIRC, how about one that at least verifies the timestamps are proper?
kurosu has quit [Quit: Connection closed for inactivity]
TheSashmo has joined #ffmpeg-devel
balling has quit [Ping timeout: 252 seconds]
TheSashmo has quit [Remote host closed the connection]
Daemon404 has left #ffmpeg-devel [#ffmpeg-devel]
\\Mr_C\\ has joined #ffmpeg-devel
aaabbb has quit [Server closed connection]
aaabbb has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
balling has quit [Client Quit]
balling has joined #ffmpeg-devel
balling has quit [Client Quit]
balling has joined #ffmpeg-devel
balling has quit [Client Quit]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
hbbs has quit [Quit: bye]
hbbs has joined #ffmpeg-devel
hbbs has joined #ffmpeg-devel
thilo has quit [Ping timeout: 268 seconds]
\\Mr_C\\ has quit [Remote host closed the connection]
thilo has joined #ffmpeg-devel
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
lexano has quit [Ping timeout: 260 seconds]
wangbin has joined #ffmpeg-devel
wangbin has quit [Client Quit]
MrZeus_ has quit [Ping timeout: 246 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
compn has quit [Read error: Connection reset by peer]
compn has joined #ffmpeg-devel
wangbin has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
dykai has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
jamrial has quit []
jkqxz has quit [Server closed connection]
jkqxz has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
dlb76 has quit [Server closed connection]
TheSashmo has joined #ffmpeg-devel
dlb76 has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
durandal_1707 has quit [Ping timeout: 240 seconds]
durandal_1707 has joined #ffmpeg-devel
balling has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has quit [Ping timeout: 246 seconds]
dykai has quit [Quit: dykai]
wangbin has quit [Ping timeout: 268 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
MisterMinister has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Roger16 has joined #ffmpeg-devel
Roger16 has quit [Client Quit]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
AbleBacon has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
wangbin has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
MisterMinister has quit [Ping timeout: 256 seconds]
compn has quit [Ping timeout: 252 seconds]
balling has joined #ffmpeg-devel
compn has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
compn has quit [Read error: Connection reset by peer]
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
balling has quit [Ping timeout: 256 seconds]
<courmisch>
kierank: just call it liboringmpeg. bonus points if you can make a backronym from ORING
<courmisch>
or libmmboring
TheSashmo has joined #ffmpeg-devel
balling has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has joined #ffmpeg-devel
<elenril>
yes of course, it's not 'else if', or 'elseif', or 'elif' as you might expect
<elenril>
it's 'elsif' because fuck you
<elenril>
perl is so great
<aaabbb>
just be grateful conditional statements in perl don't have sigils :p
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
kurosu has joined #ffmpeg-devel
<courmisch>
elenril: The Party Line states that Perl is good for writing, and bad for reading. I shall have to refer your unapproved statement to the Discipline Committee.
<courmisch>
also it could be worse: gas doesn't even have an else-if directive
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
<courmisch>
"6 modules and more than 700 API functions of FFmpeg algorithm library are optimized (...) "
<courmisch>
show me the codez
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has quit [Read error: Connection reset by peer]
<Lynne>
courmisch: probably they're only talking about h264, and probably their thought is just use a better vectorizing compiler
Krowl has joined #ffmpeg-devel
<courmisch>
Lynne: AFAIU, it's about writing assembler. the article says nothing about fancy compiler tech
dellas has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<kurosu>
I'd be more interested in some kind of statistical improvement anyway as a metrics. Otherwise, they could have fixed white spaces for all I care, and achieved the same level of interest
<courmisch>
kurosu: it says 8.2, 3.67 and 3.62x depending on algo type on the tested CPU
<kurosu>
Thanks I didn't know where to look for that on this page. And it felt weird the amount fo change was the main metrics for them
gust82 has quit [Remote host closed the connection]
gust82 has joined #ffmpeg-devel
<courmisch>
but the only example code is an interleaved dot product of sorts
<kurosu>
What would explain that kind of scaling? x86 often sees much larger factors. I don't know where neon stands (likely arch-implementation-specific, if that makes sense). Allowed power usage basically?
<courmisch>
kurosu: 128-bit vectors and not so good FPU
<kurosu>
You mean, for that particular risc-v implementation? Or for neon?
balling has quit [Read error: Connection reset by peer]
<courmisch>
kurosu: for that specific Chinese RV IP
<kurosu>
I'm now wondering when we'll get results for sve2 closer to us
ccawley2011 has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
<wbs>
kurosu: I've played a little with SVE1/2; on graviton3, with 256 bit SVE vectors, with some trivial-ish functions from libavutil/float_dsp, the 256 bit SVE doesn't beat the existing NEON code which is manually unrolled to use two 128 bit vectors
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
<kurosu>
Meaning it may be larger vectors, but you don't get as much coissue even if you'd be unrolling sve2 code?
<wbs>
I didn't try unrolling the SVE code, perhaps I should try that
<courmisch>
was that streaming mode or normal?
<wbs>
normal, I think, I don't think I've read/seen how the streaming mode works yet
<kurosu>
But yes, that's the kind of things I'm curious about. AWS is pitching their graviton stuff rather intensely but it feels current results I've seen are biased or flawed
<courmisch>
there's always DDI0487
<wbs>
yes, I haven't finished reading all of it yet ;-)
<courmisch>
it's only doubled in size since the first edition
<wbs>
exactly...
<galad>
there was some sve/sve2 asm merged in x265, but the commit history is a mess and there isn't any mention of how much faster or slower it is
<wbs>
I'm reviewing some SVE/SVE2 asm for x264 currently though
<JEEB>
47
<JEEB>
whoops
<wbs>
it still only uses 128 bit of the vectors (or even less), but mainly seems to benefit from being able to load e.g. 8 bit pixels from memory into the bottom of 16 bit elements, instead of needing to use widening operatings as the next one
<galad>
nice, it's from the same person that did the x265 ones
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<wbs>
the approach seems to have been rewrite everything as far as possible, from NEON to SVE, and only retain some NEON bits where the SVE counterparts become too slow. But for one function I looked at yesterday, it can be even faster if one retains as much as possible in NEON and just do the parts that actually make a difference with SVE
<balling>
SVE2 is supposed to be supported by my 8 Gen 1, but the rumor is it is fused out. Soon SME2 will be a common thing too. There is little to no code that uses SVE2. I only know Intel math library libxsmm
<kierank>
do we support democracy today or do we prefer dictatorship?
<kierank>
hard to keep track
<courmisch>
SVE2 works in Galaxy S22, I verified it at the office. but that was probably the EU variant
<courmisch>
MTE was fused out
TheSashmo has quit [Ping timeout: 246 seconds]
Krowl has joined #ffmpeg-devel
j-b has quit [Server closed connection]
j-b has joined #ffmpeg-devel
<balling>
courmisch: well, EU means what? Exynos?
adema has quit [Server closed connection]
adema has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
elastic_dog has quit [Ping timeout: 260 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
elastic_dog has joined #ffmpeg-devel
navi has joined #ffmpeg-devel
<balling>
courmisch: can you send me an apk file to test SVE2?
TheSashmo has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
<courmisch>
balling: no I just uploaded a static ELF with adb
<courmisch>
and that was over a year ago
<balling>
I mean I can use termux. Is not it more simple to compile one function into SVE2 in mpv-android?
<courmisch>
I am not familiar with the mpv build system
<courmisch>
I just wrote a main function with the trivialest SVE insn inline
<courmisch>
don't remember details by now
<wbs>
galad: oh, also, re x265, I'm not sure about what he has contributed there, but the stuff he contributed to x264 was kinda broken in the first iteration, so I hope x265 has had a good thorough review of it. in particular, it would break on 256 bit SVE2 HW; so it would seem fine on current HW but break in the future
<wbs>
(I've access to test it on 256 bit SVE1 and 128 bit SVE2)
TheSashmo has quit [Ping timeout: 246 seconds]
<galad>
I highly doubt the x265 patches went thru a good review
<wbs>
assembly without a good enough test framework (checkasm, and in the case of SVE, qemu testing with larger vector sizes) is just a ticking time bomb
<wbs>
not saying that QEMU testing on its own is good enough for everything, testing on real HW clearly is essential anyway. But with SVE there's this extra possibility of writing assembly that only works on one specific vector length, so one should make sure to test it on multiple lengths
<wbs>
like, there's no public SVE2 HW with 256 bit vectors available yet, afaik
<courmisch>
yeah unlord is doing that for dav1d RVV too
<courmisch>
though QEMU only supports 128-1024 bits, while specs says 32-65536 bits
<wbs>
cases where you use the scalable vector instructions "properly" for algorithms where you operate on longer inputs, where you naturally have such a loop anyway, probably works fine unless you really screw things up. but for squeezing it into use in fixed block sized video codecs, there's quite a lot of possible ways of silently assume things
<wbs>
ouch, making the code work fine for 32 bit vectors while also handling longer ones, sounds super fun
<wbs>
luckily SVE has a reasonable lower bound at 128
<wbs>
and they dropped the requirement of supporting non-power-of-two lengths
<courmisch>
RVV was always power of two. 32 and 64 bit are only for hypothetical embedded subset hardware. V requires 128
<wbs>
ah, good
<wbs>
in that case, writing RVV code that works for the vector lengths that QEMU supports, but breaks on the theoretical cases outside of that, sounds like a tricky challenge ;-)
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<Gramner>
non-power-of-two vectors sounds like a recipe for obscure edge case bugs, so it's a good thing they ditched that idea
<wbs>
yep
<Lynne>
would've made for interesting nptwo FFT asm though
<elenril>
why do non-power-of-two numbers even exist
<elenril>
they are so ugly
<JEEB>
lol honeywell with please do the needful
<JEEB>
<3
<courmisch>
elenril: I hear zero is a very useful mathematical construct, also why Romans sucked at maths
<elenril>
2^-∞
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
NaN
<elenril>
lim_{x→-∞} 2^x for anal-retentive people
<courmisch>
negative numbers aren't power of two, so this is absurd
<elenril>
∞ is a notational tool, not a number
<another|>
JEEB: I'm seriously wondering how many signs it needs to point people to the correct mailinglist
<another|>
it's mentioned quite a few times in documentation you're supposed to read
<courmisch>
another|: no signs will deter entitled idiots
Krowl has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
patricia_ has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
Dariusz has quit [Server closed connection]
Dariusz has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has quit [Ping timeout: 268 seconds]
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashm_ has joined #ffmpeg-devel
TheSashm_ has quit [Read error: Connection reset by peer]
TheSashm_ has joined #ffmpeg-devel
TheSashm_ has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
kekePower has quit [Server closed connection]
kekePower has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
lexano has joined #ffmpeg-devel
patricia_ has quit [Ping timeout: 268 seconds]
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
wangbin has quit [Ping timeout: 244 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
MrZeus_ has joined #ffmpeg-devel
user23 has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
dellas83 has joined #ffmpeg-devel
dellas has quit [Ping timeout: 256 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
"Cobol also has quite requirements for numerical precision. To that end, gcobol now uses C _Float128 datatype."
<courmisch>
elenrils can use Cobol
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
lexano has quit [Ping timeout: 256 seconds]
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
oh, there comes the guy that sent me my own patch yesterday
aljazmc has joined #ffmpeg-devel
cone-969 has joined #ffmpeg-devel
<cone-969>
ffmpeg Rémi Denis-Courmont master:90a779bed6a2: lavc/huffyuvdsp: basic R-V V add_hfyu_left_pred_bgr32
<cone-969>
ffmpeg Rémi Denis-Courmont master:6b708cd78313: checkasm/huffyuvdsp: test for add_hfyu_left_pred_bgr32
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
I guess the Chinese paper from above is showing the sbrdsp.sum_square function as their example
<courmisch>
I somehow doubt that they did port 700 functions in 6 modules
TheSashmo has joined #ffmpeg-devel
aljazmc has quit [Quit: Leaving]
TheSashmo has quit [Remote host closed the connection]
aljazmc has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<courmisch>
my elenrilbait failed; I need to change supplier
<BtbN>
This still built fine yesterday, and today specifically shared builds on windows from master are broken
<BtbN>
a1a6a328f0f58af23c174828775754b28ac64b2d seems like the most likely culprit, but I don't see how
<BtbN>
Yeah, it's probably a result of adding "libplacebo vulkan" to ffplay_suggests
<BtbN>
I'm a bit confused why libavfilter is exporting those symbols though
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<BtbN>
Hm, on Linux PL_API is forced to be visibility("default"), which is problematic for static linking. But on Windows it should be nothing.
<BtbN>
Yeah, libplacebo also always builds with every API forced as __declspec(dllexport). This is breaking this here.
<psykose>
why would it be problematic? you're meant to set default-visibility on exported symbols since -fvisibility can arbitrarily change the default (and e.g. -fvisibility=hidden is required for building with CFI, etc)
<BtbN>
Oh, setting "default" in the code does not override =hidden on the cli?
<BtbN>
That's fine then
<BtbN>
The unconditional dllexport on Windows is broken though
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<BtbN>
Though arguable, if you are making a static lib, you can probably safely set the visibility to hidden
<psykose>
it does override =hidden, the point is to start with no symbols and explicitly mark default what is exported
<psykose>
this doesn't break any static linking, what scenario are you thinking of?
<BtbN>
It makes libplacebo symbols appear on libavfilter
<BtbN>
And then explodes linking ffplay, since that now itself links against libplacebo
<ePirat>
there should probably be a define added, used to guard that the visibility is only set when actually building libplacebo
<BtbN>
That part is working fine
<BtbN>
the problem is that it's set to __declspec(dllexport) when building libplacebo itself
<BtbN>
And when libplacebo is then linked into libavfilter, the exported symbols appear on libavfilter.
<ePirat>
I guess it needs a guard to set that only when not static building then
<ePirat>
for dllexport
<psykose>
isn't the issue there that libplacebo is linked into libavfilter prematurely
<BtbN>
no?
<psykose>
even if you had no symbol conflict you would be linking the same code twice
<BtbN>
That's fine, the linker figures that out
<BtbN>
But it _is_ problematic for shared builds for sure. Cause then you have two copies of libplacebo. One in ffplay, one in libavfilter
<BtbN>
so if there is any global state, it might exist twice independently
<psykose>
yes
<psykose>
idk, seems better to just link libplacebo in last step only
<BtbN>
"last step only"?
<BtbN>
Both libavfilter and ffplay depend on libplacebo now
<BtbN>
so both need to link against it
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
tufei has joined #ffmpeg-devel
<Gramner>
__declspec(dllexport) should only be used when building a dll, not for a static lib
<BtbN>
Yeah, the build system of libplacebo is having that wrong
mathiscool888 has joined #ffmpeg-devel
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
MisterMinister has joined #ffmpeg-devel
<courmisch>
can we switch to Gitlab so my stupid mistakes are less obvious
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
<another|>
how would gitlab hide your inadequacies ?
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
nobody will pay attention if you just push a new version of the MR
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<another|>
i mean... kinda true
<JEEB>
:)
<mathiscool888>
is there a list of open issues in ffmpeg? any easy wins for someone new looking to contribute?
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<elenril>
yes there is
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
<mathiscool888>
link please?
TheSashmo has quit [Remote host closed the connection]
<elenril>
no
<courmisch>
oh great, I have a pseudorandomly failing checkasm test now
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
okay so... how do I benchmark code that requires 256-bit vectors or larger?
TheSashmo has joined #ffmpeg-devel
<elenril>
get a haswell or later?
TheSashmo has quit [Remote host closed the connection]
<another|>
qemu?
<courmisch>
benchmarking with QEMU, verrry funny
<courmisch>
elenril: yeah sure, do you have RVV to AVX256 JIT?
<elenril>
you did not specify an ISA
<courmisch>
elenril: why would I? no other ISA matters obviously
<elenril>
meh, seems ddagrab does not work through wine
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<another|>
courmisch: so you want to bench non-existing HW?
TheSashmo has joined #ffmpeg-devel
TheSashmo was kicked from #ffmpeg-devel by elenril [TheSashmo]
TheSashmo has quit [Remote host closed the connection]
<courmisch>
another|: no, I want non-existing HW to run benchmarks on
<courmisch>
you will appreciate the every so subtle difference
<j-b>
good morning
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
mathiscool888 has quit [Quit: Connection closed]
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
user23 has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
user23 has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
uau has quit [Server closed connection]
uau has joined #ffmpeg-devel
AbleBacon has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has quit [Read error: Connection reset by peer]
<courmisch>
elenril: you know you can ignore PARTS and JOINS?
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
gust82 has quit [Remote host closed the connection]
<elenril>
I do, but it's still spam
<jamrial>
courmisch: doesn't make it any less annoying
<jamrial>
courmisch: also, i get a segmentation fault with your flac lpc checkasm test
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
ccawley2011 has quit [Read error: Connection reset by peer]
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
ccawley2011 has joined #ffmpeg-devel
cone-969 has quit [Quit: transmission timeout]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
lexano has quit [Ping timeout: 268 seconds]
<courmisch>
must be cosmic rays reproducibly breaking the test case
<courmisch>
jamrial: works fine here but I don't have XOP?
<jamrial>
Gramner: ^ that shouldn't be necessary, right? it's an argument in a reg, not taken from stack
<courmisch>
jamrial: ok I have no idea what that means TBH. Am I testing invalid pred_order values?
<jamrial>
courmisch: no, it's just garbage in the upper 32 bits of the 64bit gpr. just windows things
<jamrial>
but afaik it shouldn't happen unless it's an argument taken from stack
<wbs>
jamrial: 32 bit arguments on registers also have undefined upper halves
<wbs>
checkasm should try to trigger it on all platforms, but the trick/hack for doing that might need an update from dav1d
<BBB>
jamrial: I believe upper half of int can be crap also
<BBB>
at least in checkasm harnass
<BBB>
I believe in real life also
<BBB>
this is why we use intptr_t/ptrdiff_t where possible
<BBB>
it seems silly but makes life easier sometimes :)
<courmisch>
well there's plenty of places that should use ptrdiff_t or size_t but I didn't dare to touch them because I can't possibly fix all the other archs
<wbs>
yes. it's just much more rare in real life, causing annoying submarine bugs - that's why checkasm tries to trigger it everywhere
<jamrial>
yes, but i was under the impression it was only starting with the fifth argument, which is the first one to be taken from stack
<jamrial>
oh well
<wbs>
in real life, it probably hits more often for stack arguments than register arguments yes
<courmisch>
I thought the RV64 ABI was weird
<wbs>
courmisch: so, thanks - your checkasm test found a latent but real bug in the x86 asm
<JEEB>
nice
MrZeus_ has quit [Ping timeout: 256 seconds]
<courmisch>
pl well there's also a XOP optimisation for that function if somebody cares to check it
<BBB>
it's the same code
<BBB>
macro
<BBB>
so fixing one will fix both
<courmisch>
Oh, I underestimated just how much black magic the x86 forbidden arts involved
<wbs>
courmisch: what does the RV64 ABI say about 32 bit arguments passed in 64 bit registers?
user23 has quit [Remote host closed the connection]
<courmisch>
wbs: narrow values are signedness-extended to 32-bit, then sign-extended to 64-bit
<wbs>
upper half undefined, zero or sign extended?
<courmisch>
it's a very convoluted way to say that everything is signedness extended except u32 which is sign extended
<jamrial>
BBB: no, i'll push it directly since i got your lgtm
Guest6852 has quit [Quit: WeeChat 3.5]
<courmisch>
so everything works fine except "unsigned int"
<BBB>
ok
cone-217 has joined #ffmpeg-devel
<cone-217>
ffmpeg James Almer master:78f55457c9be: x86/flacds: clear the high bits from pred_order in lpc_32 functions
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has joined #ffmpeg-devel
Daemon404 has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
elastic_dog has quit [Ping timeout: 256 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
<courmisch>
RISC-V has unlocked the 100 assembler functions achievement
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
elastic_dog has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
Krowl has quit [Read error: Connection reset by peer]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
mkver has quit [Ping timeout: 268 seconds]
user23 has quit [Remote host closed the connection]
user23 has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
rcctl has quit [Ping timeout: 264 seconds]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
mkver has joined #ffmpeg-devel
<courmisch>
uh, just saying but
<courmisch>
add_gradient_pred_c: 3591.0
<courmisch>
add_gradient_pred_ssse3: 123.5
<courmisch>
add_gradient_pred_avx2: 494.0
<JEEB>
fun
<courmisch>
add_left_pred_rnd_acc and add_left_pred_zero similar
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
<courmisch>
suddenly x86 people are all gone
TheSashmo has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
inb4 "you don't have the right kind of x86"
aljazmc has quit [Quit: Leaving]
TheSashmo has joined #ffmpeg-devel
balling has quit [Read error: Connection reset by peer]
TheSashmo has quit [Remote host closed the connection]
rcctl has joined #ffmpeg-devel
<BBB>
I have no idea :)
<BBB>
is this lossless stuff?
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
I don't know, it's clearly loosing my CPU cycles
<courmisch>
but yes
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
<courmisch>
I see Arm does not have the exclusivity of slowing things down with newer extensions
TheSashmo has quit [Remote host closed the connection]
<courmisch>
RV solved that problem by not having enough coding space left to add a new vector instruction set
<courmisch>
(well, no not really)
<elenril>
courmisch: 30x speedup for ssssssssssssse3?
TheSashmo has joined #ffmpeg-devel
<elenril>
that sounds like the c version is broken
TheSashmo has quit [Remote host closed the connection]
<courmisch>
elenril: it sounds like the AVX version is broken
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<kurosu>
add_gradient_pred seems to assume that width is a multiple of 16, if not 32. Maybe processing garbage pixels doesn't matter
<Gramner>
jamrial: upper bits are undefined. dav1d has a checkasm feature on x86 that should 100% reliably detect such bugs, but ffmpeg checkasm only have an older hacky variant of that feature that doesn't really work anymore
Son_Goku has quit [Server closed connection]
Son_Goku has joined #ffmpeg-devel
<courmisch>
dav1d hoards all the good things
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<wbs>
ffmpeg's checkasm is still one magnitude more mature than the one in x264
<wbs>
the x264 one relies on nonstandard statement expressions, which don't build with msvc
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
<courmisch>
sounds very rusty
<courmisch>
sooooo, I need to recompile every C function that calls mid_pred() with different flags
<courmisch>
this is going to be fun, not
markh has quit [Server closed connection]
markh has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
<durandal_1707>
FFmpeg is dead.
<Lynne>
long live ffmpeg
Wessie has quit [Server closed connection]
Wessie has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
TheSashmo has quit [Remote host closed the connection]
markh has quit [Read error: Connection reset by peer]
user23 has quit [Remote host closed the connection]
user23 has joined #ffmpeg-devel
TheSashmo has joined #ffmpeg-devel
ccawley2011 has quit [Read error: Connection reset by peer]
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
<pross>
durandal_1707: what comes after ffmpeg?
<elenril>
fffeet?
user23 has quit [Remote host closed the connection]
user23 has joined #ffmpeg-devel
cone-217 has quit [Quit: transmission timeout]
markh has joined #ffmpeg-devel
user23 has quit [Remote host closed the connection]
balling has quit [Read error: Connection reset by peer]
balling has joined #ffmpeg-devel
gust82 has joined #ffmpeg-devel
Compnn has joined #ffmpeg-devel
hbbs has quit [Quit: bye]
balling has quit [Remote host closed the connection]
balling has joined #ffmpeg-devel
hbbs has joined #ffmpeg-devel
hbbs has joined #ffmpeg-devel
xvaclav has joined #ffmpeg-devel
Teukka has quit [Read error: Connection reset by peer]
Teukka has joined #ffmpeg-devel
Teukka has joined #ffmpeg-devel
Teukka has quit [Changing host]
s55 has quit [Server closed connection]
s55 has joined #ffmpeg-devel
gust82 has quit [Remote host closed the connection]
gust82 has joined #ffmpeg-devel
<durandal_1707>
FFmpeg is supressing knowledge
Gramner has quit [Remote host closed the connection]