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
___nick___ has joined #ffmpeg-devel
___nick___ has quit [Client Quit]
___nick___ has joined #ffmpeg-devel
thilo has quit [Ping timeout: 248 seconds]
thilo has joined #ffmpeg-devel
mkver has quit [Ping timeout: 248 seconds]
arch1t3cht3 has joined #ffmpeg-devel
arch1t3cht has quit [Ping timeout: 252 seconds]
arch1t3cht3 is now known as arch1t3cht
IndecisiveTurtle has quit [Ping timeout: 252 seconds]
hbbs has quit [Ping timeout: 252 seconds]
hbbs has joined #ffmpeg-devel
hbbs has quit [Changing host]
hbbs has joined #ffmpeg-devel
sepro0 has joined #ffmpeg-devel
sepro has quit [Read error: Connection reset by peer]
sepro0 is now known as sepro
kasper93 has quit [Ping timeout: 252 seconds]
lexano has quit [Ping timeout: 260 seconds]
Warcop has quit [Remote host closed the connection]
Martchus has joined #ffmpeg-devel
Martchus_ has quit [Ping timeout: 258 seconds]
jamrial has quit []
jarthur has quit [Quit: jarthur]
<Lynne>
its a step in the right direction
blb has quit [Ping timeout: 252 seconds]
blb has joined #ffmpeg-devel
<elenril>
Marth64: not coming?
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg-devel
microchip_ has quit [Remote host closed the connection]
microchip_ has joined #ffmpeg-devel
<Marth64>
elenril: i have a major project due that week and tickets are $2200+
<Marth64>
possibly next time if it is in europe can go
<Marth64>
but who knows, will monitor airfare price changes
emersion has quit [Remote host closed the connection]
emersion has joined #ffmpeg-devel
<elenril>
sad
<elenril>
why is it that expensive anyway
<elenril>
my gf is flying from boston to taiwan and it's like 600 USD one way
<Marth64>
non-stop flight
<Marth64>
that's a crazy good deal :o
AbleBacon has quit [Read error: Connection reset by peer]
<Marth64>
elenril: thx for the -map fixes
<elenril>
?
<elenril>
oh, you mean the stream specifier stuff?
<Marth64>
ya
<Marth64>
fftools/cmdutils: split stream specifier parsing and matching
<elenril>
is it relevant to something you're doing?
<elenril>
for me it's mainly preparatory work for MV-HEVC
<Marth64>
just some scripting on top of CLI. mostly appreciate the fact its being iterated on and improved
<Marth64>
i can test it out tomorrow
IndecisiveTurtle has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 245 seconds]
<courmisch>
Free Hawaii from US imperialists
IndecisiveTurtle has joined #ffmpeg-devel
<courmisch>
elenril: BOS-ICN is kind of veeery long, especially if Mordorian Siberia is no fly zone
<courmisch>
BOS-TPE even worse
<Marth64>
i was just in hawaii few months ago
<courmisch>
to TPE, you have no competition from Better Chinese airlines to keep prices low
<Marth64>
me-ICN is 15 brutal hrs ;-;
<courmisch>
12 hours here via South route
<courmisch>
used to be 7 hours
<courmisch>
or 8 maybe
<Marth64>
when I did it in 2017, it was over canada route, got a glimpse of northern lights
<Marth64>
oh wow why route became longer?
<Marth64>
no fly zone i guess
<courmisch>
because of Demon Lord Vlad
<Marth64>
just caught the Mordor reference
<Marth64>
i literally just finished the 3rd film 20 mins ago
kasper93 has joined #ffmpeg-devel
<Marth64>
far too late to be awake but had a late night shift and needed background tv
<Marth64>
*sigh* tomorrow its marathon of the hobbit
kepstin has quit [Remote host closed the connection]
kepstin has joined #ffmpeg-devel
kepstin has quit [Remote host closed the connection]
kepstin has joined #ffmpeg-devel
microchip__ has joined #ffmpeg-devel
microchip_ has quit [Ping timeout: 272 seconds]
mkver has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 248 seconds]
IndecisiveTurtle has joined #ffmpeg-devel
novaphoenix has quit [Quit: i quit]
novaphoenix has joined #ffmpeg-devel
<Lynne>
indecisiveturtle: ping
MrZeus has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 245 seconds]
MrZeus has quit [Read error: Connection reset by peer]
<another|>
heh. great circle lines are still surprising
rvalue has quit [Ping timeout: 244 seconds]
rvalue has joined #ffmpeg-devel
IndecisiveTurtle has joined #ffmpeg-devel
MrZeus_ has joined #ffmpeg-devel
MrZeus__ has joined #ffmpeg-devel
MrZeus_ has quit [Ping timeout: 252 seconds]
Teukka has quit [Read error: Connection reset by peer]
Teukka has joined #ffmpeg-devel
Teukka has joined #ffmpeg-devel
Teukka has quit [Changing host]
<IndecisiveTurtle>
Lynne: Hi. Finished Legall yesterday. Took me 2 tries cause at first I tried to make it fully parallel but it didn't work out at all, so second approach each invocation process a hor/ver line at a time. Now doing Deslauriers-Dubuc filter looks quite similar honestly
<IndecisiveTurtle>
It seems like it had issues again with workgroup sync not sure
lexano has joined #ffmpeg-devel
<IndecisiveTurtle>
I think tomorrow the window for final submissions will open. Might need some deadline expansion depending on how much is there to complete. But I hope I can finish it in time
<Lynne>
sure, np
<Lynne>
what's the performance like?
<IndecisiveTurtle>
It seems to be about on par with cpu encoder, tho I'm not sure what gpu ffmpeg is using (my vega igpu or nvidia one)
<IndecisiveTurtle>
After filters are done I want to discuss possible optimization options for slice calculation and encoding
<IndecisiveTurtle>
Because atm each invocation processes a slice so similar to cpu code. But it would be nice to have each invocation perform per pixel work and elect one invocation to do the auxilary work
<Lynne>
I think you should prioritize speed for now, legall is at most all you'll need
<Lynne>
you don't get much compression gains with better wavelets, legall is pretty optimal
<IndecisiveTurtle>
I will try to optimize legall in that case as much as possible
<IndecisiveTurtle>
Possibly have another go at the fully parallel solution
<IndecisiveTurtle>
The issue that seemed it had was again workgroup sync, but I didn't quite understand why. The dumps it produced vs cpu were very wrong
<IndecisiveTurtle>
Lynne: So to recap you think having slice size/encoding shaders do slice-per-invocation is fine perf wise?
<Lynne>
yeah, definitely
<Lynne>
maybe not for the actual transform step, since you have it as a separate shader
<Lynne>
but you can easily run queries and see which shader takes the most amount of time
<IndecisiveTurtle>
Ah alright then. Btw is there a way to control which gpu ffmpeg picks for vulkan so I can test/profile for both vendors
<IndecisiveTurtle>
BDAs have quite picky behavior wise if they have bugs depending on which vendor
<IndecisiveTurtle>
Some people told me AMD hw cannot recover from page faults and it seems to make sense, during earlier stages of the haar shader when it was wrong it could lock up my laptop when radv used
<Lynne>
use -init_hw_device "vulkan=vk:<gpu>" with gpu being either a 0-indexed number of the GPU, or a direct name (use -loglevel verbose or vulkaninfo to see a list) or a UUID or a PCIE vendor
<Lynne>
these days with the latest kernel and a compositor like sway amdgpu will recover in most cases, with maybe your wallpaper client being broken, but who looks at wallpapers, and you can restart it in the worst case
<Lynne>
if your amd gpu doesn't handle displaying you can crash it all you want, won't affect your system and you won't have to wait for it to recover
<Lynne>
I run risky stuff on a separate machine, then when I'm confident enough I move them to my main machine, and then I immediately crash them and go back, is what I always end up doing
<IndecisiveTurtle>
I'm using KDE and sometimes it resets immediately other times there was a freeze, probably while it reset the gpu