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
iive has quit [Quit: They came for me...]
thilo_ has quit [Ping timeout: 255 seconds]
thilo_ has joined #ffmpeg-devel
Livio has quit [Ping timeout: 255 seconds]
System_Error has quit [Remote host closed the connection]
cone-821 has quit [Quit: transmission timeout]
arch1t3cht5 has joined #ffmpeg-devel
arch1t3cht has quit [Ping timeout: 240 seconds]
arch1t3cht5 is now known as arch1t3cht
System_Error has joined #ffmpeg-devel
Marth64 has joined #ffmpeg-devel
<Marth64>
Hi all, hope you are well.
<Marth64>
Picking up work on DVD again (1 trivial bug, and 1 bug to fix a particular disc). Had needed a hiatus but gonna try to be active again.
IndecisiveTurtle has quit [Ping timeout: 246 seconds]
<Lynne>
nice
<Lynne>
do you think you'll be able to pay attention to seeking?
<Marth64>
Lynne: Hi and yes! the "1 trivial bug" is a discovery I made on the journey to making seeking possible
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
<mkver>
courmisch: Your lls test contains a stack-buffer-overflow. The reason for this is that update_lls accesses elements 0..indep_count (inclusive!) and not 0..(indep_count - 1).
AbleBacon has quit [Read error: Connection reset by peer]
<courmisch>
mumble mumble Rust mumble
mateo` has quit [Ping timeout: 256 seconds]
System_Error has quit [Remote host closed the connection]
georgereynolds8 has quit [Ping timeout: 268 seconds]
System_Error has joined #ffmpeg-devel
<courmisch>
mkver: patch on ml
georgereynolds8 has joined #ffmpeg-devel
mateo` has joined #ffmpeg-devel
mkver has quit [Ping timeout: 255 seconds]
<kurosu>
pal: there are subsections in fate, maybe you want to check if there's one for the thing you're checking. Not in front of a computer, but wasn't there a make fate-list? I remember having a bash function that took a string, and ran all matching instances in fate-list
System_Error has quit [Remote host closed the connection]
cone-445 has joined #ffmpeg-devel
<cone-445>
ffmpeg Nuo Mi master:8abdf0a28e01: avcodec/vvcdec: misc, move src offset inside emulated_edge
<cone-445>
ffmpeg Nuo Mi master:2d98786fee56: avcodec/vvcdec: refact, remove emulated_edge_dmvr and emulated_edge_bilinear to simplify code
<cone-445>
ffmpeg Nuo Mi master:a4013e748a79: avcodec/vvcdec: refact out emulated_edge_no_wrap
<cone-445>
ffmpeg Nuo Mi master:685174069f9b: avcodec/vvcdec: misc, reindent inter.c
<cone-445>
ffmpeg Nuo Mi master:f68f40736f1a: avcodec/vvcdec: support mv wraparound
Livio has joined #ffmpeg-devel
<courmisch>
why are ff_c{l,t}z taking signed values :(
ccawley2011 has joined #ffmpeg-devel
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 268 seconds]
rvalue- is now known as rvalue
<courmisch>
FATE subtitle tests assume that iconv succeeds for CP1251
System_Error has joined #ffmpeg-devel
Livio has quit [Ping timeout: 264 seconds]
staceee has quit [Remote host closed the connection]
sm2n has quit [Read error: Connection reset by peer]
<cone-445>
ffmpeg Rémi Denis-Courmont master:3ef5867e4b6d: sws/input: R-V V rgb24ToUV and bgr24ToUV
<cone-445>
ffmpeg Rémi Denis-Courmont master:e0f4d185f1ea: sws/input: R-V V rgb24ToUV_half and bgr24ToUV_half
<cone-445>
ffmpeg Rémi Denis-Courmont master:658439934b25: lavc/vp8dsp: R-V V vp8_idct_add
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
elvis_a_presley has quit [Quit: smoke-bomb ; grapple-hook]
elvis_a_presley has joined #ffmpeg-devel
<courmisch>
According to Steam, 106.85% of processors support SSSE3
<courmisch>
Don't ask me how they came up with the number.
<BBB>
courmisch: that's awesome
<Lynne>
avx2 at 99.85% lets gooooo
Livio has joined #ffmpeg-devel
<BBB>
if it's 99.85% out of 106.85% (or more), then that's not so awesome
Livio has quit [Ping timeout: 240 seconds]
<courmisch>
Corrolary: Arm's market share must be below -6.85%
<courmisch>
can I pay negative €s to get Arm antiprocessors?
<Lynne>
someone just copied the arm chip designs a hundred billion times, causing them to lose all their money so he can buy them for nothing
<Lynne>
give them time to buy then back and delete all the designs they copied
<courmisch>
isn't Arm suing Qualcomm to literally have them delete their Arm server IP?
<courmisch>
then Qualcomm tried to bluff that they would switch their mobile line-up to RISC-V, only to fail miserably because their design's I-fetch can't deal with compressed instructions
<Lynne>
does anyone even use compressed instructions in anything with more than 640kb of ram?
AbleBacon has joined #ffmpeg-devel
<jkqxz>
It's not just for embedded things with very little (or no) external memory; it also increases throughput when you have constrained memory bandwidth (including caches).
deus0ww_ has joined #ffmpeg-devel
deus0ww has quit [Ping timeout: 268 seconds]
PoorvaGaikar2003 has joined #ffmpeg-devel
<Lynne>
right, I fogot about the bandwidth limits of most current dev boards
<Lynne>
the 8gb x60 board should fix that somewhat, they're using two ram chips rather than one
<courmisch>
Lynne: all Linux distros use C, just to save I cache I think
<courmisch>
though me thinks the real reason is just to beat Armv8 at code density metrics
<Lynne>
that is not something I was expecting
PoorvaGaikar2003 has quit [Ping timeout: 272 seconds]
iive has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
Livio has joined #ffmpeg-devel
<courmisch>
yikes, dct_unquantize
<courmisch>
why so much scalar code in the DSP function
<courmisch>
and in C data structures
cone-445 has quit [Quit: transmission timeout]
System_Error has joined #ffmpeg-devel
<courmisch>
kind of explains why the NEON code is written in intrinsics
Krowl has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
darkapex has quit [Remote host closed the connection]
darkapex has joined #ffmpeg-devel
<mkver>
courmisch: Stay out of mpegvideo-hell or you'll lose your sanity.
<compn>
yeah people get the idea to review the old code. then decide its too bad to save. then they try to write their own mpegvideo decoder. then no one ever hears from them again.
<compn>
better to spend time training an ai to write the code
<compn>
let the ai lose its mind
<iive>
courmisch, probably because the code does rle, zigzag and matrix at the same time.
<courmisch>
mkver: my what now?
<iive>
when you have few non-zero coefficients, scalar code might be faster than mixed scalar and simd.
<courmisch>
each functions has a scalar prologue that really doesn't want to be in assembler
<courmisch>
then some very vectorisable (and vectorised) code
<elenril>
mkver: did you?
<courmisch>
result: every architecture has copied the prologue and written the loop in intrinsics
<michaelni>
"<compn> let the ai lose its mind" ... why do i have a bad feeling here
<courmisch>
and it's basically unfixable because who's going to revector the Alpha and MIPS?
<elenril>
git rm?
<courmisch>
elenril: one at time. We're trying to remove SuperHitachi atm
<compn>
i dont think its unfixable. could always part it out. whatever.c for alphamips and then whatever-fixed.c for the rest.
<compn>
i dont have problem with duplication :D
<courmisch>
I mean, yes, basically we have to create DSP callbacks around DSP callbacks now
<iive>
courmisch, i took a look. Probably the prologue could be moved into the codec and the asm block to be done into it's own function.
<courmisch>
iive: ofc, that's how it should be. But that would require updating the 5 different architectures that implement the callbacks already
<courmisch>
inb4 what-if somebody will be running FFmpeg 8.0 on a DEC Alpha
<courmisch>
as far as I am concerned, we can unleash a wild starved elenril at those optimisations but ...
<iive>
courmisch, you don't have to touch the assembler blocks for these architectures.
<courmisch>
at least Arm has some actually assembler
<courmisch>
s/ly //
<iive>
gah, they are all kind of different. some use intrinsics