/mpv/ - the /g/reatest media player

How do I know if I have performance problems?

>>62459965
video wont look right

>>62459965
You can install status.lua and press SHIFT + I
If you see red you should look into it

>>62460000
I see two red boardered boxes with white in it (Vsync stuff) and frame timings fresh go red when I hit pause but at playback they become white again.

Is that normal?

>>62460066
Yeah, post a screenshot if you're still worried

>>62459850
>Post your system specs
Thinkpad T420 with Intel HD 3000 iGPU.
What's best config for it?

>>62460124
empty config

>>62460124
Try

profile=opengl-hq

and see if you get respectable timings

File: 1505558126910.jpg (23KB, 460x460px) Image search: [Google]

23KB, 460x460px

>>62457938
rossy / akemi / hanna appreciation

What are some visually appealing fonts?
Since you can add font folder in /.config/mpv, and it will use it withput affecting the system's fonts.
Why do I need font.conf ?

File: stats.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62460087
Hmm, what about the dropped frames 30 (output) and delayed and mistimed stuff?

>>62460146
wm4 >>>

>>62460142
I get frame drops, had to disable deband, which helps a lot.
Still CPU temps are very high when playing 1080p.

>>62460162
post specs and config.

>>62460000
No, don't do this

>>62460179
i5-2520M right? You shouldn't have much trouble playing back h264 content
You wouldn't happen to be attempting to watch h265 content would you?

Any vulkan build yet?

>>62460166
This. wm4 so underrated.

>>62460188
CPU: Intel Core i5-4690K @ 4x 3.9GHz
GPU: GeForce GTX 970
RAM: 16GB

config is opengl-hq, interpolation with tscale=oversample and ravu-r4.hook

>>62459965
Well, essentially when you are dropping frames.

hanna: rossy: so does this mean we push vo_gpu or do we wait for wm4 to get back? :^)
rossy: lol
hanna: he *was* pretty drunk
rossy: "just don't fuck it up" "that's all I'm asking for"
hanna: I'm pretty sure we fucked everything up
rossy: sounds like it comes before an always sunny title card
rossy: for an episode where we fuck everything up
hanna: just don't fuck it up
* rossy can't promise anything

>>62460162
>that vsync jitter
Are you on Linux?

>>62460166
>>62460251
Yes but I'm not sure he would even want /g/ to appreciate him...

>>62460274
Yes.

>>62460243
>i5-2520M
Yes, and no I'm watching h264.

>>62460162
It's normal to drop a few frames e.g. when fullscreening or when you activate certain options that change VO stuff. But they should increase over time during normal playback.
Mistimed/Delayed are related to display-sync and your high jitter (should be more like 0.0x) indicates some problem in your setup (not mpv config setup but your graphics environment)

>>62460266
mpvhq coming back soon

>>62460289
Stick with this then >>62460129
Try to stick to the bare minimum in your config
I don't have a thinkpad so I can't really speak but something must be chewing away at your CPU as the i5-2520M has full h264 decoding on the chip

>>62460251
He needs animu avatar

>>62460266
>"just don't fuck it up" "that's all I'm asking for"
>I'm pretty sure we fucked everything up
Yeah so... don't push it...

File: file.png (131KB, 996x1269px) Image search: [Google]

131KB, 996x1269px

>>62460166
>>62460251
Can confirm wm4 does more than any of us by a pretty wide margin. He's the real mvp. As project leader and the only guy who gets paid (by Plex,) he also has to do all the jobs that no one else wants to do.

>>62460244
You can build it yourself

>>62460303
>indicates some problem in your setup
Like what?
I have the latest nvidia drivers installed.
My WM is i3 if that has anything to do with it.

>>62460347
But >>62460330

>>62460321
>full h264 decoding on the chip
I'm using "GNU/Linux", do I need to install something perhaps?

>>62460266
> :^)

>>62460372
Yes agreed.

>>62460374
Honestly it wouldn't surprise me, take a look and perhaps ask in /fglt/

File: 06637240.jpg (23KB, 257x276px) Image search: [Google]

23KB, 257x276px

>>62460347
Holy shit.

>>62460374
Yes of course. Depends on your distro and other things though.

>>62460347
But what if all of this is just doc instead of code.

>>62460374
If you want to use the on-chip decoder you need to enable it in mpv.conf

>>62460407
>>62460426
I'm using manjaro, and /fglt/ isn't so friendly when you say you're using it.

>>62460440
You just have to look at the commit log to know it's mostly code. Also, he's insanely productive. He designed RA when he was supposed to be on holiday.

>>62460426
>>62460451
Honestly it's baffling to me that something like that isn't included universally

>>62460467
He's also been working on mpv for a lot longer than most of the other contributors on that list

File: 1486018446787.png (31KB, 176x217px) Image search: [Google]

31KB, 176x217px

I've fixed wm4

File: screenshot0033.jpg (247KB, 1280x720px) Image search: [Google]

247KB, 1280x720px

>>62460564
>furry
please no

>>62460479
I'm not using Linux as daily drivers but I have my share of experience with installing and using debian and arch.
Problem, especially with video, is that you have options/alternatives even among drivers and there is no best driver but instead each comes with different shortcomings or advantages.
Then there's stuff like compositors that many people don't need, some want them and they can influence a lot as well.
Just look at https://github.com/mpv-player/mpv/wiki/FAQ#Tearing and you see the state of video on linux. There are just too many different cases to enable you a working out of the box solutions. You as a user have to fiddle around and find the best combination of things by trail&error. People that attempt to create a "included universally" solution (e.g. distros like Ubuntu) are considered bloated and every second neckbeard complains about it not running on their dishwashers.

>>62460564
cute but not sure if that ear piercing suits wm4

>>62460600
Just werks for me on open source AMD drivers.

>>62460600
>Intel tears out of the box. Intel users on Linux are going to have a pretty bad time.

>>62460573
>>62460602
Only anime picture I have
>>62460600
Yeah you've got a good point

>>62460679
nice. left eyebrow messed up a bit :)

Where do I set the mpv/youtube-dl quality?

>>62460708
There's no such thing as "quality" with youtube-dl. You can specify formats but that can be rather complex.
By default, youtube-dl tries to go for the best (and mpv furthermore excludes certain things that don't work well with ffmpeg) so you already get better quality.
Aside from that, you can include hardcoded bitrate, fps, resolution and codec limits.

See:
https://github.com/haasn/gentoo-conf/blob/xor/home/nand/.mpv/config#L37
https://github.com/Argon-/mpv-config/blob/master/mpv.conf#L35

>>62460761
*best quality

The linked option is btw just passed to youtube-dl, it's not mpv that created such a complex representation
More information about format strings: https://github.com/rg3/youtube-dl/blob/master/README.md#format-selection

>>62460600
Ironically still less variance than the OS X shitshow

>>62460835
OSX is a completely different topic because by now I'm pretty confident Apple deliberately WANTS their opengl implementation to be as broken as possible to deliver the worst experience possible to developers just so they switch to Metal.
At least I hope it's like that. Otherwise they are just batshit insane.

>>62460156
Anyone?

><hanna> Argon]: should we remove timers from stats.lua again
please

>>62460981
What timers?

Is there any way to port my Windows custom display mode to Linux?

>>62460274
What does vsync jitter mean?

>>62460981
It's funny because it was haasn who wanted them in the first place.
I think it's disconcerting how often he changes his opinion about stuff. It's kinda nice how he's speeding up development but at some point I don't feel like renaming half of my config anymore or following all his new ideas that might get implemented, shortly afterwards changed and then maybe removed again.

>>62461079
haasn is for innovations, i like that. wm4 for balance and stability.

>>62461079
To summarize: I tend to look at performance stats quite often just for the fun of it. And now, just because he has suddenly a hunch (once again) he robs users of this feature again?
Oh it's tiring... I want my video player to work and I appreciate new features every now and then. However, I don't feel like reading mpv's commit log every second day just to see what he changed or broke again. Let along removing stuff I use.

>>62460941
vo=metal for macos when

>>62461079
He didn't. Only because mesa devs are being retards and refuse to fix their shit code.

>>62461021
post details of your mode so i can see what you have, and i might be able to tell you how to translate it

>>62461118
Yeah but wm4 only maintains stability for a small set of stuff.
He won't bat an eye when haasn removes the stats now and I suppose he will so I as a normal user will lose something I liked and I can't do anything about it, I'm not a developer.
Well, stay on an old version of course but...

>>62461119
Refer to https://bugs.freedesktop.org/show_bug.cgi?id=102597

>>62461132
I thought it's open source, so fix it yourself...

>>62461155
Yeah, all thousands of amd Mesa users should individually patch their mesa.

>>62461145
So because mesa - something I don't use - has a bug I will loose functionality now?

File: 3142123.jpg (93KB, 659x467px) Image search: [Google]

93KB, 659x467px

>>62461140

>>62460981
Timers? As in frame timings? Possibly the most useful thing in mpv?

>>62459199
But 8x6 is theoretically more powerful/better right?

>>62461267

Modeline "1920x1080" 164.87 1920 1928 1960 2052 1080 1103 1111 1117 +HSync -VSync

i think this is right
to break it down;

Modeline; tells X11 that this is a modeline
"1920x1080"; the name of the mode, doesn't have to be the visible output size, but this is typical
164.87; clock rate in MHz
1920; visible horizontal pixels
1928; visible horizontal + horizontal front porch
1960; horizontal total - horizontal back porch
2052; horizontal total
1080; visible vertical pixels
1103; visible vertical + vertical front porch
1111; vertical total - vertical back porch
1117; vertical total
+HSync -VSync; sync pulse polarities

ps, your mode isn't exactly 72 or 72/1.001 ("71.93"), though it's close enough if paired with display-resample

>>62461171
>>62461273
I am using mesa (on arch) and I also like them

>>62461391
How do i use all of that? I only migrated to Linux recently. Thank you so much anon!

>>62461419
i personally have a file at /etc/X11/xorg.conf.d/10-monitor.conf which contains;

Section "Monitor"
    Identifier "HDMI-0"
    Modeline "1920x1080" 106.82 1920 1928 1952 1960 1080 1083 1087 1090 -HSync +VSync # 50Hz
    Modeline "1920x1080" 132.48 1920 1928 2016 2024 1080 1083 1089 1092 -Hsync +Vsync # 60/1.001Hz
    Modeline "1920x1080" 130.80 1920 1928 1992 2000 1080 1083 1086 1090 -Hsync +Vsync # 60Hz
    Modeline "1920x1080" 164.16 1920 1928 1968 1976 1080 1083 1152 1155 -HSync +VSync # 72/1.001Hz
    Modeline "1920x1080" 156.96 1920 1928 1992 2000 1080 1083 1087 1090 -HSync +VSync # 72Hz
    Modeline "1920x1080" 164.87 1920 1928 1960 2052 1080 1103 1111 1117 +HSync -VSync
    Option "PreferredMode" "1920x1080"
EndSection

should be self-explanatory
you can use "xrandr" to see what your display is labelled as if not "HDMI-0"

also, if you want to quickly test a mode, you can add/switch to one using xrandr without editing anything or reloading X11, using;

$ xrandr --newmode test 164.87 1920 1928 1960 2052 1080 1103 1111 1117 +HSync -VSync
$ xrandr --addmode HDMI-0 test
$ xrandr --output HDMI-0 --mode test

note; you might want to run "xrandr --output HDMI-0 --auto" before running the last one, so you can blindly run it easily if the mode doesn't work and you're stuck on a blank screen

>>62461482
(that last modeline is yours, wrote it in there as a guide, should have removed it)

>>62461482
>>62461391
>>62461499
Thank you! You're the best! *sends kisses*

File: blushing_kotori.gif (443KB, 500x288px) Image search: [Google]

443KB, 500x288px

>>62461545

>>62461673
>I see. BTW you use 72Hz mode, right? Do you have interpolation enabled for 23.976FPS videos?
no, i don't use interpolation (the mpv interpolation which just blends frames) at all, as i have a mode that suits any typical video rate
as far as i'm concerned, it should only be used when a non-multiple display rate isn't available, as it's at least more tolerable than straight dropped/duped frames

>>62461733
Do you use something like autospeed script?

>>62461760
i use xrandr.lua to let mpv switch to a suitable mode automatically

>>62460981
Change it so if it drops a frame or gets a mistimed frame it flashes red. Having it red when you're not dropping any frames causes people to complain about shit when there's nothing wrong.

File: [Coalgirls]_Tamako_Love_Story_(1920x1080_Blu-ray_FLAC)_[ED2903E1].mkv 00_04_15.171.png (462KB, 458x648px) Image search: [Google]

462KB, 458x648px

>>62461378
bump

>>62462079
nnedi3 is deprecated

>>62461378
>>62462079
Its not better on practice, anon. I tested it on normal images and test images. On test images it looked weird sometimes. 8x4 is a safer choice. Or you can use RAVU for a similar look. Or FSRCNN for SUPER sharp.

>>62461786
What does this do for me exactly?

DVI-I-1 connected 1920x1080+0+0 (normal left inverted right x axis y axis) 477mm x 268mm
   1920x1080     60.00*+
   1680x1050     59.95  
   1280x1024     75.02    60.02  
   1152x864      75.00  
   1024x768      75.03    60.00  
   800x600       75.00    60.32  
   640x480       75.00    59.94

Will it only switch on fullscreen?

>>62462144
It's nor SUPER sharp, it's natural.

>>62462189
It looks natural only on very high quality images. On regular ones it looks SUPER sharp.

>>62462151
you only have a 60Hz mode for 1920x1080, so xrandr.lua won't do anything (unless you use one of the modes that has multiple rates, like 1280x1024)
you could try adding additional modes (see >>62461482)
note: these modes i use are with a monitor which also only shows 60Hz normally, yet works fine with 50/72/71.93/59.94

>>62462151
>>62462235
>>62461545
also, you may be able to query your monitor's limits using;

$ grep "Ranges: " /var/log/Xorg.0.log
[ 69484.264] (II) RADEON(0): Ranges: V min: 50 V max: 76 Hz, H min: 30 H max: 83 kHz, PixClock max 175 MHz

>>62462235
>>62461482
Tells me

X Error of failed request:  BadMatch (invalid parameter attributes)
  Major opcode of failed request:  140 (RANDR)
  Minor opcode of failed request:  18 (RRAddOutputMode)
  Serial number of failed request:  39
  Current serial number in output stream:  40

when I try to addmode.

>>62462338
Didn't work. I read through the file too, doesn't seem like there's anything in there.

>>62462441
>Tells me
your monitor or video card might not like that particular modeline
>Didn't work. I read through the file too, doesn't seem like there's anything in there.
could be a driver-specific thing, not sure where to find this elsewhere (well, besides your monitor manual)

>>62460162
the pain of no madVR....why would anyone do it?!

>>62461171
>So because mesa - something I don't use - has a bug I will loose functionality now?
It's a fucking script, you won't lose shit if you don't update it, retard.

>>62460347
>9k commits
>still not up to the level MPC-HC was 10 years ago.

I see mpv is on the qBittorrent development cycle.

>>62462441
oh, if you're using the xrandr lines, you are changing "HDMI-0" to your "DVI-I-1" (as in >>62462151), right?

>>62462487
Not everyone can afford Windows.

>>62462500
Hi rukia

>>62462516
Of course I am.

File: laughing.gif (1MB, 400x225px) Image search: [Google]

1MB, 400x225px

>>62462520
He pays for software

>>62462534
did you try one of the modes besides the last one in my 10-monitor.conf?
the last one was a translation of someone else's custom mode, and i haven't tested it myself

File: 1505519770394.png (191KB, 600x600px) Image search: [Google]

191KB, 600x600px

>>62460347
If he's paid by plex, who pays plex?

>>62462584
Plex Pass subscribers.

>>62462567
I tried the last 3 and all failed.

File: 2017-09-16_19-06-17-sel.png (17KB, 746x107px) Image search: [Google]

17KB, 746x107px

>>62462478
>well, besides your monitor manual
This part?

http://www.lg.com/us/business/download/resources/BT00001837/BT00001837_638.pdf
I have 27MB65PY

>>62462593
are you using nvidia proprietary drivers by chance? i didn't have much luck with that a few years ago before i switched to a radeon card (pretty much so i could use the free drivers, which just work much nicer, and nowadays isn't even slower than the proprietary driver)

>>62462677
>are you using nvidia proprietary drivers by chance?
Yes, I have a GeForce GTX 970.
Shouldn't I use them?

>>62462494
You don't understand. They want to remove the information from mpv as it's currently transmitted to the script.
Also, not upgrading doesn't work well in mpv-land.

>>62462522
Still doing nothing but shilling a free player 24/7, eh?

>>62462728
>They want to remove the information from mpv
citation needed

>>62462757
stop stealing subs or at least give credits to the authors

>>62462776
No.

>>62462650
well that rules out 50Hz as it's minimum v-freq is 56
there's no mention of the max pixel clock rate, which doesn't help

>>62462697
the free nvidia driver, nouveau, last i checked, didn't do reclocking (core/memory frequency) well, so it's not really a feasable choice for high end use
there might be other ways to get custom modes to work with the nvidia driver besides using xrandr though
at the time i was trying to make modes for 15KHz analog output, to a CRT SDTV, so just because i couldn't get that to work, doesn't mean normal PC 30KHz+ custom modes can't be done with it

>>62462697
>>62462796
oh, it's just gotten worse
>"Little hope of reclocking becoming available for GM20x and newer GPUs as firmware now needs to be signed by NVIDIA to have the necessary access."
your GTX 970 is a GM204

>>62462697
if you have a /etc/X11/xorg.conf or /etc/X11/xorg.conf.d/*-nvidia.conf, could you post it?

>>62463041
I just auto generated it with nvidia-settings.

Section "ServerLayout"
    Identifier     "Layout0"
    Screen      0  "Screen0" 0 0
    InputDevice    "Keyboard0" "CoreKeyboard"
    InputDevice    "Mouse0" "CorePointer"
    Option         "Xinerama" "0"
EndSection

Section "Files"
EndSection

Section "Module"
    Load           "dbe"
    Load           "extmod"
    Load           "type1"
    Load           "freetype"
    Load           "glx"
EndSection

Section "InputDevice"
    # generated from default
    Identifier     "Mouse0"
    Driver         "mouse"
    Option         "Protocol" "auto"
    Option         "Device" "/dev/psaux"
    Option         "Emulate3Buttons" "no"
    Option         "ZAxisMapping" "4 5"
EndSection

Section "InputDevice"
    # generated from default
    Identifier     "Keyboard0"
    Driver         "kbd"
EndSection

Section "Monitor"
    # HorizSync source: edid, VertRefresh source: edid
    Identifier     "Monitor0"
    VendorName     "Unknown"
    ModelName      "LG Electronics 27MB65"
    HorizSync       30.0 - 83.0
    VertRefresh     56.0 - 75.0
    Option         "DPMS"
EndSection

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce GTX 970"
EndSection

Section "Screen"
    Identifier     "Screen0"
    Device         "Device0"
    Monitor        "Monitor0"
    DefaultDepth    24
    Option         "Stereo" "0"
    Option         "metamodes" "nvidia-auto-select +0+0"
    Option         "SLI" "Off"
    Option         "MultiGPU" "Off"
    Option         "BaseMosaic" "off"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

>>62463162
this is somewhat stab-in-the-dark, as i don't have an nvidia card at hand to test with, and i'm not sure what the defaults for some things are (which are also not mentioned in your generated conf), but it almost certainly won't hurt to try some stuff

try replacing your "Monitor" section with this;

Section "Monitor"
    # HorizSync source: edid, VertRefresh source: edid
    Identifier     "Monitor0"
    VendorName     "Unknown"
    ModelName      "LG Electronics 27MB65"
    HorizSync       30.0 - 83.0
    VertRefresh     56.0 - 75.0
    Option         "DPMS"
    Option         "ExactModeTimingsDVI" "True"
    Option         "UseEdidFreqs" "False"
    Option         "IgnoreEDIDChecksum" "DFP"
    Option         "ModeValidation" "NoMaxPClkCheck, AllowNon60HzDFPModes, AllowNonEdidModes, NoEdidMaxPClkCheck"
    Modeline       "1920x1080" 132.48 1920 1928 2016 2024 1080 1083 1089 1092 -Hsync +Vsync # 60/1.001Hz
    Modeline       "1920x1080" 164.16 1920 1928 1968 1976 1080 1083 1152 1155 -HSync +VSync # 72/1.001Hz
    Modeline       "1920x1080" 156.96 1920 1928 1992 2000 1080 1083 1087 1090 -HSync +VSync # 72Hz
EndSection

(you will need to restart Xorg, also be sure the generated/edited conf is at "/etc/X11/xorg.conf")

>>62463303
>>62463162
oh, do you do know how to log into a tty and remove/move/rename this file if there's trouble?

>>62463341
Yeah, I should be good.

Trying it now.

>>62463303
>>62463341
>>62463385
I fucking love you. Also the 72Hz webm is purple for me now.

DVI-I-1 connected 1920x1080+0+0 (normal left inverted right x axis y axis) 477mm x 268mm
   1920x1080     60.00+  72.00*    71.93    59.94  
   1792x1344     60.01  
   1680x1050     59.95  
   1600x1200     65.00    60.00  
   1400x1050     74.76    59.98  
   1280x1024     75.02    60.02  
   1280x960      60.00  
   1152x864      75.00  
   1024x768      75.03    70.07    60.00  
   832x624       74.55  
   800x600       75.00    72.19    60.32    56.25  
   700x525       74.76    59.98  
   640x480       75.00    72.81    59.94  
   512x384       70.07    60.00  
   400x300       72.19  
   320x240       72.81    60.05

>>62463482
nice, enjoy the new modes

File: 23477408.jpg (157KB, 389x401px) Image search: [Google]

157KB, 389x401px

OMG. That guy made two people happy today!

>>62463510
Gonna need to have a look at xrandr.lua now.

>>62463522
making others happy is how i make myself happy

>>62463529
just drop it in your scripts folder (which is alongside mpv.conf, create if necessary), at least, i'm pretty sure that's it, i got it ages ago

Is there really no developer interested in supporting DVD menus?

Would really like this feature.

>>62463556
Should I still use interpolation with this?

I'm a bit confused.

>>62463574
It was removed because nobody wanted to support the half broken existing ones.

>>62463574
back when dvd was the norm, it was a feature to skip over dvd menus, both in hardware and in software
most people don't really want the menus

>>62463593
not if the display rate is a multiple of the video rate, that is;
~29.97/~59.94fps video - ~59.94Hz
~23.976fps video - ~71.93Hz
24fps video - 72Hz
30/60fps video - 60Hz
you'll still need it for 25/50fps video (your monitor can't go under 56Hz) or other strange video rates which don't multiply cleanly into one of those

NGU optimized for anime soon. /mpv/ on suicide watch once again.

>>62463662
>not if the display rate is a multiple of the video rate
I see, as I expected.
But would it be detrimental to leave it on?

Also if I switch the modeline order will X default to the 72Hz one?

>>62462136
t.Pentium D

>>62463734
>But would it be detrimental to leave it on?
i'm not sure what it does in the case of a clean multiple, it might do nothing, making it harmless to leave on

>>62462520
>Windows
>expensive
You can get $5 keys from ebay ,just check the seller reputation first.

>>62463728
source or gtfo

>>62463734
also, i'm not sure how to change the default refresh rate
i suppose you could give the preferred mode a unique name, by which you can reference it using "preferredmode"

...
    Modeline       "1920x1080p72" 156.96 1920 1928 1992 2000 1080 1083 1087 1090 -HSync +VSync # 72Hz
    Option         "PreferredMode" "1920x1080p72"
EndSection

the order you specify the modelines in xorg.conf doesn't affect anything afaik, at least with my driver

>>62463944
>>62463734
ok, that doesn't really work out, as that mode becomes separate from the rest of "1920x1080", so things like xrandr.lua won't see rates besides 72Hz

>>62464113
Yeah I noticed.

I just put this in my i3 config.

exec xrandr --output DVI-I-1 --mode 1920x1080 --rate 72

Though I should probably put it into my lightdm config file.

>>62464169
that'll do

also, i just tested interpolation, and it does indeed still blend frames even with clean multiples, it effectively becomes a cheap form of motion blur in this case
whether you want it in these cases or not is up to you

>>62464313
Someone told be long time ago that interpolation creates only intermediate frames or something in case of clean multiples.

>>62464313
I tried to use cvt before to get a modeline for 72Hz, but it looks very different compared to yours.

Modeline "1920x1080_72.00"  210.25  1920 2056 2256 2592  1080 1083 1088 1128 -hsync +vsync

I wonder why yours is working and this one doesn't.

>>62464430
it is creating intermediate frames, that's the problem
the additional frames are a blend of the previous and next frames, the amount of each being a function of the additional frames' temporal distance from each real frame (closer to old frame, display a stronger blend of the old frame...)

>>62464456
that pixel clock is likely too high. my monitor for example, won't take anything over 165MHz
i'm simply using this page to generate pixel clock + total values;
https://www.monitortests.com/pixelclock.php
then i just use 3px for front/back vertical porch and 8px front/back horizontal porch

>>62459850
What anime is this?

>>62464581
Gurren no Pico: Shippuden

I've been searching for the right video player for some time now, spent a while on VLC, MPV and bomi. The latter is my favourite simply for the way it generates and displays playlists but it no longer seems to have developer support.

I want to migrate back to MPV or perhaps a fork of it but I am wondering how difficult it would be to get a slick looking playlist on the gui such as bomi's.

>>62464553
>it is creating intermediate frames, that's the problem
Shouldn't it produce less blur than when there is no clean multiple? I dunno if its a problem since even in worse case scenario smoothmotion produces very low amount of blur.

>>62464581
read the file name!
>>62464595
Nice spoonfeed :<

>>62464604
>Shouldn't it produce less blur than when there is no clean multiple?
it should, as at least all the real frames should be displayed unmodified, as they all line up with display frames in that case
where poorly matched rates might be mostly blended frames with few fully unmodified frames

i would say assign it to a keyboard button and toggle it on and off to see the differences, such as with (input.conf);

3 cycle interpolation

(press "3" to turn it on/off during playback)

Why is this happening? Google didn't find anything.

Unknown profile 'opengl-hq'.
Error parsing option profile (option could not be parsed)
/home/tsuribori/.config/mpv/mpv.conf:2: setting option profile='opengl-hq' failed.                                                                        
Error parsing option scale (option not found)                                
/home/tsuribori/.config/mpv/mpv.conf:3: setting option scale='ewa_lanczossharp' failed.                                                                   
Error parsing option cscale (option not found)                               
/home/tsuribori/.config/mpv/mpv.conf:4: setting option cscale='ewa_lanczossharp' failed.

>>62464801
do you have some ancient build of mpv?

>>62464821
Well I'm on Ubuntu 16.04 LTS so probably.

>>62464688
I have a 72Hz display that passed your test from the last thread. I have no motion judder during the panning scenes but some parts of the picture are stuttering. Not whole picture like during 3:2 pulldown but parts of it. Interpolation helps to smoother that shit. What do you think? What might be causing that? High panel response time?

>>62464839
Install the latest version from the PPA. Links in the OP.

>>62464841
a 24fps video won't be smooth no matter what you do (well, besides full motion interpolation such as with SVP/mvtools, but that's a very imperfect science)
using 72Hz for 24fps media just eliminates inconsistent frame times and drops/dupes due to misaligned video frames versus display frames
it doesn't make 24fps smooth, 24fps is 24fps. it's a low frame rate, simple as that

>>62464906
>>62464841
put simply, 24fps-on-72hz is what 24fps is supposed to look like, on an LCD (anything sample-and-hold), the picture is absolutely identical to if you used a 24Hz TV
it still not looking smooth is simply 24fps not being high enough in the first place

File: f52ce1f5a6f2e78f752a059b55523d0a.jpg (149KB, 1451x816px) Image search: [Google]

149KB, 1451x816px

>>62464861
Thanks Anon, works now

>>62464841
use display-sync=resample

File: 1477978033841.jpg (12KB, 280x318px) Image search: [Google]

12KB, 280x318px

There threads are becoming /desktop/ threads with the amount of attention whoring

>>62465039
I do. I guess its just my panel being slow.

The difference between watching animu in 60Hz and 72Hz is so immense, I can't even believe my eyes.

>>62465181
What do you mean?

Is there a way to actually unbind a key? Setting the command to "ignore" doesn't work, mpv still captures the key event and prevents other applications from using it.

>>62465382

no-input-default-bindings

>>62465276
It's like the first time in your life seeing a naked girl.

>>62465427
What do you mean?

>>62465181
>>62465427
I don't think so. Probably placebo.

Is rossy's D3D11 renderer going to be released soon? (less than a month)

so if I understand it correctly ravu/nnedi are "trained" (?) shaders but trained using anime, is there any way I can train my own version with movies or some other medium?

>>62465582
Why do you care?

>>62465610
ravu is trained with anime. nnedi3 trained with everything.

Is haasn's Vulkan renderer going to be released soon? (less than a month)

>>62465611
To get better performance with my toaster.

>>62465610
1. training is pretty intensive so rented computing power is usually used
2. you need a LOT of training data to train a neural network, and in RAVU's case the neural network isn't deep enough to really benefit from different types of training data (deeper neural networks = slower, RAVU has to be fast)

>>62465684
Does slower also mean objectively better?

>>62465865
yes

>>62465684
wew lad

>>62465907
So which prescaler is best?

I'm using nnedi3-nns256-win8x4.hook

>>62466088
>I'm using nnedi3-nns256-win8x4.hook
no you're not

File: 1487565775351.png (89KB, 630x686px) Image search: [Google]

89KB, 630x686px

>720p movie with shaders look better than 1080p movie without
>too much of a brainlet set up shaders correctly so they still work on 1080p

>>62466139
you're probably retarded, but what shaders?

>>62466088
The tutorial I saw some time ago said anything above nnedi3-nns32-win8x4 is too much

>>62466088
r u retarded

>>62466139
Just use SSSR, brainlet. It wont work on videos with native res to youre monitor.

>>62466131
I realized I'm dropping a lot of frames with this.

>>62466158
>anything above nnedi3-nns32-win8x4 is too much
This one works better, but is it really better than ravu?

Meme config for 1950x threadripper?

>>62466263
https://github.com/bjin/mpv-prescalers/tree/master
use nnedi3 nns256, I'm not sure which window size is better though

File: Untitled-2.jpg (1MB, 1863x1165px) Image search: [Google]

1MB, 1863x1165px

>>62466157
>>62466185

File: VLC.png (79KB, 500x500px) Image search: [Google]

79KB, 500x500px

I am ready to move from vlc to mpv but I am not windows and where do I even start? which build do I use? please spoon feed me

>>62466355
I meant I am using windows (8.1)

>>62466355
>>62466380
https://sourceforge.net/projects/mpv-player-windows/files

https://github.com/Argon-/mpv-config/blob/master/mpv.conf

https://mpv.io/manual/master/#files-on-windows

>>62466355
>>62466380
https://sourceforge.net/projects/mpv-player-windows/files/
pick 64 or 32 bit
pick latest build
open .7z archive
run install.bat as admin
run configure opengl-hq.bat as admin

>>62466420
don't do this, just extract, run the install shit and use portable_config folder in the extracted location, fuck AppData shit

>>62466435
>just extract
that was implied in "open .7z archive
>run the install shit
ok
>use portable_config
configure opengl-hq already makes a portable mpvp.conf, the fuck are you talking about

File: 1497129904079.jpg (6KB, 206x210px) Image search: [Google]

6KB, 206x210px

>>62466469

>>62466420
ok I have done this
this is the contents of opengl-hq.conf
# High quality video rendering for fast computer.
profile=opengl-hq

is there anything else I should add there?

File: 0-raw.png (147KB, 448x256px) Image search: [Google]

147KB, 448x256px

Found an old 448x256 avi.

Let's put ravu, nnedi and sssr to the test.

1/4

File: 1-ravu.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62466516
opengl-shader=~~/shaders/ravu-r4.hook

2/4

File: 2-nnedi.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62466526
opengl-shader=~~/shaders/nnedi3-nns32-win8x4.hook

3/4

>>62466503

profile=opengl-hq

scale=ewa_lanczossharp
tscale=oversample
video-sync=display-resample
interpolation

screenshot-format=png
screenshot-high-bit-depth=yes
screenshot-png-compression=9

my personal config
you should read https://mpv.io/manual/master/ though, especially if you have a weird setup

>>62466306
I dont get it. Whats your point?

File: 3-sssr.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62466548
opengl-shader=~~/shaders/SSimSuperRes.glsl

4/4

>>62466516
>>62466526
>>62466548
>>62466566
did you set the max-downscaling-ratio appropriately?

>>62466558
>what shaders?

>>62466580
I don't even know what that is.

I have dscale=catmull_rom in my config though.

>>62466589
Im the other anon you quoted. What your picture was supposed to show?

why newfags always post their retarded incorrect comparisons?

>>62466601
https://github.com/bjin/mpv-prescalers
neural upscalers (RAVU and nnedi3) can only double resolution
you have to apply RAVU or nnedi (literally put multiple lines of

opengl-shaders-append=blahblah[/code in your config the necessary amount of times) the appropriate amount of times to reach the output resolution

>>62466631
That sounds unflexible as fuck.

>>62466648
it's a common compromise for neural upscalers because you either have to train it to double/triple the image (consistent transformation) or train to upscale to an arbitrary resolution (consistent output)
because of this, RAVU and nnedi3 just double the image and downscale it to your monitor resolution
how often do you watch videos that are more than 1.8x (default max-downscaling-ratio) smaller than your monitor resolution?

>>62466628
SSSR objectively won.

I'm gonna use that now.

>>62466737
You wont even notice any difference on such a horrible source image. Youre right with SSSR though. Its the least complicated to use shader and produces the best result (subjectively) for most people.

>>62466806
I tested it on 720p anime too, to make sure.
Ravu comes pretty close to SSSR, but nnedi is a blurry mess.

>>62466686
Everything from DVDs should be easily go >2. And for this content a good scaler is even more important.

>>62466888
>DVDs in UNIX epoch + 47
lmaoing at ur life

File: 0-raw.png (912KB, 1280x720px) Image search: [Google]

912KB, 1280x720px

>>62466861

File: 1499479077155.png (50KB, 255x204px) Image search: [Google]

50KB, 255x204px

Think I'm satisfied with my windows config, based it on the Argon one:
https://pastebin.com/Q9hBMHJy

profile suggestions always appreciated, not sure if everything is correct, I'd like some help with the SD shit and HD/UHD

File: 1-sssr.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62466993

>>62460266
when is he back?

File: 2-nnedi.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62467010

File: 3-ravu.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62467030

>>62467030
>>62467049
>and then you noticed that nnedi3 and RAVU dont work with png's out of the box
Oh wait, you didn't. Use stats.lua, newfag.

>>62467124
What do you mean?
I use stats.lua

I realized just now using SSSR my estimated display fps caps at 60 instead of going to 72 like with ravu. That's horrible, but why?

>>62467175
You used .png as source, correct? Also >>62467030 looks like its not even the same frame.

>>62467247
>You used .png as source, correct?
No, the video file. I just put the 0-raw.png there for reference.
I used Q to quit and reopen at the same frame when I changed the shaders. then used "SHIFT+s screenshot window" to take the screenshots.

>looks like its not even the same frame
I don't know why nnedi does this weird distortion, it really sucks.

File: 85002720.jpg (17KB, 350x259px) Image search: [Google]

17KB, 350x259px

>>62467355

>>62467355
>SHIFT+s screenshot window
You mean ctrl+s?

How do i make it so i can open ISO's with mpv from the file manager menu?
I have a .desktop file that has --dvd-device=$f as teh exec part but it dosent start the iso in mpv it just dosent open mpv at all either
Is there an easier way to open iso's with mpv that isnt constantly using the cli?

>>62467444
Here's another one (got tired of uploading 4 screens to 4chan):
https://diff.pics/jKH6yXRJuZ3f/1

This time I used opengl-hq instead of raw.
Again, SSSR wins objectively!

Also, how am I gonna learn anything if I don't make mistakes. It's not like the correct way to do this is written somewhere.

>>62461119
Protip: https://raw.githubusercontent.com/mpv-player/mpv/master/DOCS/interface-changes.rst

This documents all backwards-incompatible changes. Also, mpv tries really hard to make changes backwards compatible, so you only get a deprecation warning. The new vo_gpu shit is a good example, the only thing that actually got hard-removed was --opengl-es=force2, everything else has backwards compatibility aliases in place.

>>62461132
It's not just mesa, timers are also completely useless on OS X and ANGLE; and possibly WGL. Heck the only platform I know that implements timers properly is nvidia linux.

>>62467791
How to measure performance then? ;___; Its much easier to increase or decrease your settings when there is some number estimates.

>>62467763
hardly a good example when it hasn't been merged yet :^)

>>62464313
>>62464430
tscale=oversample and tscale=nearest are the only two modes that don't do anything on clean multiples. Turning it off in those cases might save you a tiny bit of performance, especially on implementations with a low OpenGL latency.

>>62467867
Do you think any of the past uber-refactors are any different?

Fucking --vo=opengl-hq is still accepted and that was deprecated over a year ago

>>62467908
>tscale=oversample doesn't do anything on clean multiples
That explains why I didn't notice anything.

>>62467833
Here are two things I can offer:

1. To estimate raw throughput (e.g. for shader performance tuning and stuff), run with `--no-audio --opengl-swapinterval=0 --video-sync=display-desync` on a clip that matches your framerate (e.g. 60 Hz for a 60 Hz monitor). If you don't have a clip, you can use --speed to “generate” one yourself (e.g. by playing a 24 Hz clip with --speed 2.5 to make it 60 Hz, or --speed 3 to make it 72 Hz). Then watch the “estimated display fps” property. This is essentially like turning vsync off in games and doing benchmarks by measuring your overall FPS. That should give you a very reliable estimation about how taxing various settings actually are, with no room for error due to bad timing figures.

2. To make sure it's actually working properly at normal speeds, watch the “delayed:” and “mistimed:” properties. If those go up, you're *actually* losing frames.

To give you an example, mesa (*with* the mesa timing fix applied, and the GPU set to 'auto' frequency) gives me frametimes of >30k ms (with spikes to 50k ms) but I can reliably play 24 Hz content (41ms per frame) with interpolation on a 60 Hz monitor; because the OpenGL frame latency is enough to hide the 30k ms rendering. Without the mesa timing fix, it's even more stupid and reports like 50k-100kms frame times. Yet I still play with no framedrops whatsoever. This is because the timers can just lie, and the whole “color in red if above 1/vsync thing” is just a complete myth.

>>62468001
>That should give you a very reliable estimation about how taxing various settings actually are, with no room for error due to bad timing figures.
Oh, but it requires you to actually decode the video at that speed. Personally I can decode 1080p H.264 at several hundred FPS anyway so it's not a big deal, but you could also use one of ffmpeg's built-in test pattern generators if it becomes an issue.

(You can also test with --untimed -vo null to see if decoding is the bottleneck)

Is nnedi3 for chroma good?
I have the resources to spare.

>>62468173
If your content has chroma aliasing it can help but in most cases it's probably placebo (and you might just want to use cscale-blur or something on aliased chroma anyway).

File: 1502311941322.jpg (40KB, 960x743px) Image search: [Google]

40KB, 960x743px

>>62462556
>He watches anime

>>62468579
>he uses an anime website

>>62467908
>tscale=oversample and tscale=nearest are the only two modes that don't do anything on clean multiples
But they do! The difference is very clear.
>>62468001
So TLDR for regular users is to watch for “delayed:” and “mistimed:” properties?
>>62468173
Good but id rather use KrigBilateral.

File: Captura de tela de 2017-09-16 21-00-04.png (835KB, 1366x768px) Image search: [Google]

835KB, 1366x768px

Anons, help.
Why this happens when I capture (pressing S) a screenshot with subtitles? The image comes out ok, but the subtitles in the video gets fucked until I disable the option or re-open the file.
MPV, SMPlayer and codecs are all updated, this happens since over a year now.

>>62468666
>So TLDR for regular users is to watch for “delayed:” and “mistimed:” properties?
Yes. Also, the timers are only reliable if your GPU clock is constant. Most GPUs implement shader clock frequency scaling, which completely throws off the timers. (Since they're based on the number of elapsed GPU ticks)

File: [HorribleSubs] Tenshi no 3P! - 10 [720p] - [00_19_22.453].jpg (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

>update mpv
>timings disappear

>>62465641
Almost surely. The only thing that's still `WIP` about the vulkan renderer is the dumb glslang hack, but that should take no more than a day to fix properly; and rossy might also be working on it (he needs the same thing for SPIRV-Cross)

Then it could at least get merged, although it's an open question as to when it will actually provide benefits compared to opengl. There are *several* things that still need to get improved to make vulkan start really shining compared to opengl. But here are three key things that could be coming soon-ish:

1. array of arrays support in glslang. This is something mpv can do basically nothing about except wait. But it's required for our EWA kernels. So right now, the fast compute shader versions of the EWA kernels are not supported by glslang (and therefore vulkan, except on nvidia). But if glslang won't implement it for some reason, we can still avoid the use of arrays of arrays in mpv. (It was just slightly slower to do so iirc)

2. Support for async compute. This mostly depends on #1 for it to actually be useful, so I won't implement it until I can test it. This will also help for #3 in a huge way.

3. Support for time-sliced rendering. This is going to be massive, and will solve many of the problems that vulkan was originally intended to design. Basically the idea is that instead of rendering the entire frame at once, we can split it up into different rendering “chunks” and submit those in any order we want. This will make it so that we can basically be rendering small bits and pieces of future frames “ahead of time”, interleaving this with stuff like interpolation and presenting to the screen. We can also offload compute shaders to the async compute queues and have ongoing rendering happen *while* the compute shader is working, because there will no longer be explicit ordering dependencies between everything due to the time slicing across frames.

>>62469196 (cont)
Time-sliced rendering is pretty much the “killer feature” of vulkan, and it's something that's just straight up impossible with opengl unless you want to use very unrealiable, buggy and platform-specific hacks (like shared contexts), and even then it would require a lot of difficult refactoring work for opengl. (vulkan makes it much easier)

>>62468001
>angle ~145fps
>dxinterop ~250fps
>win 60fps
That's cool.

>>62469248
If you're getting exactly 60 fps, then --opengl-swapinterval=0 might not be working (and you need to find some other way of disabling vsync)

>Stop coloring timing values
Good start. Very nice change.

>>62469036
L-lewd.

File: 46892809.jpg (48KB, 444x366px) Image search: [Google]

48KB, 444x366px

>>62469196
Good luck with all of that. I believe in you!

>>62469292
Disabling it on gpu doesn't work for win backend too.

>>62469212
Actually it would be possible with OpenGL by simulating command buffers using gl sync objects and virtual lists of commands to run; but not on all platforms. And it would be more difficult anyway, since you'd also need to assemble dependent state changes (like texture uploads etc.) into the same command buffers.

>>62469036
you don't need them :^)

What cscale and dscale should I be using with ravu-r3-smoothest1?
Currently using cscale=ewa_lanczossoft and dscale=mitchell but the cscale is using half of the rendering time each frame and it's only just managing to get under 16ms.

Using 290X + 1440p monitor

>>62469539
read
>>62467791
>>62468001
>>62469027

why madvr has rendering timers and mpv not?

>>62469685
madvr is a professional software.

>>62469699
then why its free?

>>62469738
It's on beta, wait for 1.0.

>>62469036
Report the issue...?

File: screenshot0001.jpg (678KB, 1920x1080px) Image search: [Google]

678KB, 1920x1080px

>>62469589
>Watch out for dropped frames instead
I sometimes get noticeable dropped frames, like 2-4 in a row.
But it happens very very rarely, just a couple of times when I watch a 45 minute episode of something, so I have a total of 30 dropped frames in the end.

Doesn't make sense to me why this would happen. What causes this, or is that totally normal?
It annoys me whenever I see it happening.

>>62469738
It won't be for long. Madshi has plans to monetize madVR for when it's out of beta. He thought about starting with premium features some time ago. He might do this once the NGU stuff is stable and mature and also when he has his display-sync alternative.

>>62469869
Dropped frames on certain events (e.g. fullscreen) are normal. What you need to watch out for is a constantly increasing amount of them.

>>62469910
>display-sync alternative
But he has custom resolutions.

>>62469919
It's not an event though, it just happens randomly during playback.

>>62470074
That's not supposed to happen. Doesn't mean your config is too hard for your hardware but something doesn't sit well.

Will haasns activity go down next month?

>>62470099
He need to do some work on madvr or people will start to complain.

File: gpuvis.png (47KB, 1841x530px) Image search: [Google]

47KB, 1841x530px

>>62468001
Example. Frame-times for fresh frames were ~25ms, yet no dropped frames. Because even though it renders “through” a vsync, the swapchain depth of 2 (default) covers up the fact that we “skipped” a vsync.

Another day, another release with linked mkvs.

File: async.png (197KB, 4096x2128px) Image search: [Google]

197KB, 4096x2128px

>>62470615
also lol, apparently async transfer just plain wasn't working

File: async-transfer.png (324KB, 4096x2128px) Image search: [Google]

324KB, 4096x2128px

>>62470881
fixed

>The maximum pixel clock rate for HDMI 1.0 is 165 MHz
>HDMI 1.3 increases that to 340 MHz
Should I get it lower than 340mhz?

anyone know what specific libs are required to get youtube live streams with mpv on ubuntu? I've got it on a second machine and everything in the main repos are out of date but PPAs tend to blow shit up for me. I'd rather not resort to source.

Youtube-dl is latest cause it has a built in updater.

is krigbilateral still relevant? Im using it on native res.

>>62465420

how do i post like this?
how to post text inside a white box?

>>62471256
go read the sticky

>>62471151
literally a meme

It would be nice to get a statistics viewer for when you want to see framerate, jitter, sync offset, bitrate. etc. since that doesn't seem to be in the default program

>>62471266
OKAY

and

File: Capture.png (5KB, 567x47px) Image search: [Google]

5KB, 567x47px

>>62471489
can't you fucking read?

>>62471367
There's one in the TOOLS/ dir

It's called stats-something. (use it on the file generated by --dump-stats)

>>62471531
Oh sorry you were talking about something else. Yeah I've approached Argon about merging stats.lua into mpv itself. I think nobody was against it, so it's just a question of time. (Argon said he wanted to improve some things first)

>>62471512
i can read but i dont understand.

>>62471538
basically what i was thinking. it was really nice and handy having a constantly updating info while playing like media player classics >statistics view. but the stats.lua is just a snapshot. i was trying to test out some examples to figure out if the doubling of the framerate with interpolation was working but i couldn't tell. thus the hoping for statistics that show it.

>>62459965
When you can no longer keep an erection without your cock ring

>>62471147
Nevermind this, I got it from mc3man's mpv-tests & xerus-media repos, just had some trouble reinstalling shit with apt-get. mpv wins again

>>62471589
stat.lua does update though???

>>62471978
oh that was totally my bad. didn't notice the part that said "i" versus "I" to make it stick.

>>62467000

volume=70
volume-max=150
audio-pitch-correction=yes          # automatically insert scaletempo when playing with higher speed
ad-lavc-downmix=yes                 # enables downmixing
#audio-normalize-downmix=yes        # makes sound quiet (enable/disable normalization if surround audio is downmixed to stereo)
audio-channels=mono
af=format=channels=1

[HD]
profile-desc=cond:get('height', 0) < 2000 and get('height', 0) >= 700
# doesnt fuckin work

why

>>62472613
do you have auto-profiles.lua?

>>62467524
shaders are so worth it look at that
>>62468873
MPV is bad at playing "broken" subs or files, in fact out of all the players is probably the worst in that aspect.
>>62472636
yes

Can someone help me watching a DVD movie with MPV?

>https://github.com/mpv-player/mpv/issues/4841

Seriously this is the betrayal of a lifetime, how could MPV do this to me? I literally have no way to watch this movie with MPV, what the fuck my dude...

I'm worried about haasn. When does he sleep?

has anyone experience random flickering lately?

can this player play BDMV bluray discs with menus?

If not which player can acomplish this?

I could rip the thing into a mkv but it has like 100 streams and i cant know which audio is supposed to go where without acess to the menus.

>>62472646
>MPV is bad at playing "broken" subs or files, in fact out of all the players is probably the worst in that aspect.
??? mpv just uses ffmpeg like every other player except VLC. (and VLC isn't exactly good at playing broken files either)

I have no idea where you got this from. Also in his case, it's probably not an mpv issue. I would guess it's an SMPlayer issue becuase >lolsmplayer

File: async-tf.png (309KB, 4096x2128px) Image search: [Google]

309KB, 4096x2128px

async compute is back on the menu, boys

Now my vulkan performance matches my opengl performance with radv/mesa! \o/

>>62473681
why are you even bothering with that meme hardware

Is this going to use Vulkan ever?

>>62473815
Haasn said it was almost done weeks ago.

>>62473744
1. Because they have tools like gpuvis, R600_DEBUG, VGPR statistics etc.
2. Because they have open source drivers
3. Because they have a helpful IRC channel
4. Because they openly document their hardware down to the microassembly level, document how the underlying execution engine works, etc.
5. Because they publish shittons of articles about their hardware, vulkan and how to optimize for it
6. Because changing amdgpu driver settings etc. to debug stuff is much easier than trying to do anything with nvidia

Only thing that's annoying about AMD hardware so far is the fact that they only have one graphics queue. Fortunately, I could actually work around it by using making every single pass a compute shader (+ submitting them to the compute queues). This could improve pipelining.

Honestly it's so comfy that I'm keeping my RX 560 plugged in even though it's slower than my GTX 970 (means I can't use RAVU or play DOOM)

>>62473849
Oh, also, it's funny because RADV claims (via stdout) it's an experimental non-conforming implementation but it actually works _WAY_ better than the nvidia vulkan driver. It's like nvidia just doesn't even care about implementing the specs properly. It's a miracle they passed the khronos conformity test suite.

>>62473849
But none of that matters when the hardware itself is a poor performing spaceheater.

>>62473833
Vulkan has been working and feature complete(*) for weeks. Everything since then is just performance improvements, enhancements etc.

(*) except for screenshot support :^)

>>62473859
>Oh, also, it's funny because RADV claims (via stdout) it's an experimental non-conforming implementation but it actually works _WAY_ better than the nvidia vulkan driver.
What happened to mesa devs being retarded and not implementing simple features or following spec like you were complaining about a week ago?

>>62473867
WTF! So I can already compile mpv from the vulkan branch and use the vulkan renderer!

>>62473892
on linux

>>62473899
Ah! Is it going to work/compile on windows soon?

>>62473925
not until someone cares enough to get it working there because haasn sure as hell doesn't

>>62473930
What about the rossy's d3d11 renderer?
Does it work/compile?

>>62473892
Yes.

Although if you're on windows and not-nvidia, you probably need to change one of the paths in the glslang hack to make it a valid windows filepath. I'll change that to use a tmpfile path soon-ish (maybe tomorrow)

>>62473999
Oh right, you also need to add support for windows surfaces :^)

>>62473947
That one is WIP: https://github.com/rossy/mpv/commit/92424352e58c985

Should compile and work, give it a try?

>>62474010
Oh great!

>>62473888
That issue is limited to OpenGL TIME_ELAPSED timers. Vulkan (RADV) isn't even affected by it. It's also just cosmetic. Meanwhile in nvidia land you usually just randomly segfault the nvidia driver, deadlock your device for a few seconds while the display server has to restart, or get garbage output because they can't implement buffer alignment properly.

Oh, and the nvidia vulkan context takes like half a second to create. With RADV it's basically instant.

>>62474169
mesa people also reply to my bug reports, nvidia just ignores them completely

>>62470615
>>62470881
>>62470901
>>62473681
Pretty pictures.
>>62471151
>is krigbilateral still relevant?
Yes.
>>62472957
He goes to sleep when i go to sleep :^) (just a coincidence!!!)

>>62474528
Yes, it's very pretty; and the UI is super responsive and nice to use. Way better than mpv's python stats shit. I might try making mpv dump its stats graphs to this instead.

It uses one of those meme imGUI toolkits (that I still think mpv should be using for its OSC etc. instead of stupid libass crap)

>>62474552
what would it take to convince wm4 though

>>62474667
more importantly, it would be pretty much impossible to plumb all the way through the layers

I don't think it would even be possible to integrate with LUA, so the OSC would have to be written in C. It would probably also only work with --vo=gpu and nothing else

>>62473861
>nvidia employee received $0.04 for this post

>>62473054
What kind of random flickering? Old flashes of previous frames kind of stuff?

I get that in some of my tests, but never during actual (“normal”) playback. Might still be a bug lurking somewhere nasty

>>62473947
>>62474010
Yeah, in theory this should compile and work on computers other than mine. Make sure you have a git master build of glslangValidator.exe and spirv-cross.exe in your %PATH%.

--sub-ass-override=strip

is lewd

>>62475082
>ass
>strip
wew lad

>>62475160
you forgot
>sub

>>62473601
It's from my experience, has nothing to do with SMPlayer.

>>62475200
What's lewd about sub?
You tryna send some subliminal message here?

>>62475271
common abbreviation for a submissive

>>62472646
>shaders are so worth it look at that
Here's another one for you, this time non-anime:
https://diff.pics/sy84pO5VqSOe/1

I dropped nnedi3 cause it's a load of crap.

SSSR is magical. Would it perform better if mpv included it natively somehow?

>>62475331
MadVR SuperRes algorithm is even more sharper with less bloat. However NGU Sharp Medium+ is even more magical.

File: fsrcnn.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>62475388
I made one for FSRCNN_24_4_4_ssim.glsl too.

It does some weird things.

>>62475388
go away retard
MadVR SuperRes is a load of crap.

>>62475331
>>62475408
Please don't post your retarded comparisons again.

>>62475457
I will continue unless you teach me how to make them not retarded.