picom considerably slower than xcompmgr

[Hippo0o]

Platform

Platform: Arch linux amd64 kernel 5.4

GPU, drivers, and screen setup

Radeon RX 580, amdgpu
xrandr --output DisplayPort-2 --primary --mode 5120x1440 --rate 120 --pos 0x0 --rotate normal --set TearFree on

$ glxinfo -B
name of display: :0
display: :0  screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: X.Org (0x1002)
    Device: Radeon RX 580 Series (POLARIS10, DRM 3.35.0, 5.4.23-1-MANJARO, LLVM 9.0.1) (0x67df)
    Version: 19.3.4
    Accelerated: yes
    Video memory: 8192MB
    Unified memory: no
    Preferred profile: core (0x1)
    Max core profile version: 4.5
    Max compat profile version: 4.5
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 3.2
Memory info (GL_ATI_meminfo):
    VBO free memory - total: 7352 MB, largest block: 7352 MB
    VBO free aux. memory - total: 8131 MB, largest block: 8131 MB
    Texture free memory - total: 7352 MB, largest block: 7352 MB
    Texture free aux. memory - total: 8131 MB, largest block: 8131 MB
    Renderbuffer free memory - total: 7352 MB, largest block: 7352 MB
    Renderbuffer free aux. memory - total: 8131 MB, largest block: 8131 MB
Memory info (GL_NVX_gpu_memory_info):
    Dedicated video memory: 8192 MB
    Total available memory: 16384 MB
    Currently available dedicated video memory: 7352 MB
OpenGL vendor string: X.Org
OpenGL renderer string: Radeon RX 580 Series (POLARIS10, DRM 3.35.0, 5.4.23-1-MANJARO, LLVM 9.0.1)
OpenGL core profile version string: 4.5 (Core Profile) Mesa 19.3.4
OpenGL core profile shading language version string: 4.50
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

OpenGL version string: 4.5 (Compatibility Profile) Mesa 19.3.4
OpenGL shading language version string: 4.50
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile

OpenGL ES profile version string: OpenGL ES 3.2 Mesa 19.3.4
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20

Environment

i3-gaps 4.18

picom version

**Version:** v7.5

### Extensions:

* Shape: Yes
* XRandR: Yes
* Present: Present

### Misc:

* Use Overlay: Yes
* Config file used: ~/.config/compton.conf

### Drivers (inaccurate):

AMDGPU, Radeon 

Configuration:

# Shadow
shadow = true;          # Enabled client-side shadows on windows.
shadow-radius = 3;      # The blur radius for shadows. (default 12)
shadow-offset-x = -3;       # The left offset for shadows. (default -15)
shadow-offset-y = -3;       # The top offset for shadows. (default -15)
shadow-opacity = 0.75;      # The translucency for shadows. (default .75)
#shadow-red = 0.58;     # Red color value of shadow. (0.0 - 1.0, defaults to 0)
#shadow-green = 0.88;       # Green color value of shadow. (0.0 - 1.0, defaults to 0)
#shadow-blue = 0.89;        # Blue color value of shadow. (0.0 - 1.0, defaults to 0)
shadow-exclude = [ ];       # Exclude conditions for shadows.
shadow-ignore-shaped = true;    # Avoid drawing shadow on all shaped windows
                # (see also: --detect-rounded-corners)

# Fading
fading = false;         # Fade windows during opacity changes.
fade-delta = 10;        # The time between steps in a fade in milliseconds. (default 10).
fade-in-step = 0.05;        # Opacity change between steps while fading in. (default 0.028).
fade-out-step = 0.06;       # Opacity change between steps while fading out. (default 0.03).
# no-fading-openclose = true;   # Avoid fade windows in/out when opening/closing.
fade-exclude = [ ];     # Exclude conditions for fading.

# Other
backend = "glx"         
#refresh-rate = 0;      
vsync = false;
glx-no-rebind-pixmap = true
glx-no-stencil = true
xrender-sync-fence = true


# Window type settings
wintypes:
{
  tooltip = { fade = false; shadow = true; opacity = 0.99; focus = true; };
  #toolbar = { fade = false; shadow = true; opacity = 0.99; focus = true; };
  #normal = { fade = false; shadow = true; opacity = 0.99; focus = true; };
  menu = { fade = false; shadow = true; opacity = 0.99; focus = true; };
  #dialog = { fade = false; shadow = true; opacity = 0.99; focus = true; };
  popup_menu = { fade = false; shadow = true; opacity = 0.99; focus = true; };
};

Expected behavior

xcompmgr like performance

Current Behavior

slower & stuttering

Other details

I noticed a lot of slowdowns and stuttering while dragging floating windows and resizing them.
I tried glx and xrender backend with enabling and disabling some glx options.
It was worse with vsync enabled which i disabled because i can use the gpu drivers vsync.
I also got a similar setup with a nvidia gtx 970, nvidia-lts, ForceCompositionPipeline=On, 1080p 240hz using xrender backend which has the same problems (testing xcompmgr on it later).

I experience stuttering while watching twitch and dragging a transparent window over it.
When dragging a transparent terminal over another terminal it lags behind the curser. (glx only)
Resizing certain floating windows stutters a lot.
I also notice a little stutter while dragging in general (like 60hz window movement on 120hz monitor)
Only on nvidia: moving a chrome window with animations/videos in it is stuttering from time to time.

Today i tried xcompmgr and i got no stuttering at all. Overall im not noticing any slowdowns.
config used: xcompmgr -c -C -t-5 -l-5 -r3.3 -o.33

[Hippo0o]

On nvidia i get similar results.
xcompmgr gives me shadows and transparency while it isn't slower than no compositor.
picom is noticebly slower in some cases compared to no compositor.

[mighty9245]

so you basically have single rx580 handling two 2k 120hz monitors? i'm not sure about rx580 performance but that sounds like a problem. what's your overall performance? what if you leave only one monitor?

can you please compare performance between --experimental-backends --backend glx, --backend glx, --experimental-backends --backend xrender, --backend xrender with your config? which one gives you the best feeling?

[Hippo0o]

Its 1 ultrawide monitor. I can't test the rx580 setup for the next few days since its my pc at work.
But i can say that i testet all of those 4 options and got same results for my above mentioned problems. Especially the resizing lag/stuttering/freezing of said floating window didn't change.
While randomly testing --backend xrender --benchmark 1000 and --backend glx --benchmark 1000 the time difference was ~2sec vs ~1sec. idk if that helps.
I will upload a video of my 2x 1080p 240h,60hz monitor nvidia setup showing an example of what i mean.

[Hippo0o]

Setup: nvidia gtx 970, nvidia-driver 440.64 (main problems equal to above mentioned setup)
Monitor settings

xrandr --output DVI-D-0 --mode 1920x1080 --pos 0x0 --rotate left --output DP-0 --primary --mode 1920x1080 --rate 240 --pos 1080x350 --rotate normal 
nvidia-settings --assign CurrentMetaMode="DP-0: 1920x1080_240 +1080+350 {ForceCompositionPipeline=On}, DVI-D-0: nvidia-auto-select +0+0 {rotation=left, ForceCompositionPipeline=On}"

https://www.youtube.com/watch?v=PYXiw7ivUa8 testing picom and xcompmgr.
Idk why obs stutters with xcompmgr tho.
https://www.youtube.com/watch?v=Z3Uh0wj8gtA i made a cheap phone recording to prove xcompmgr being smooth compared to picom.

[yshui]

This could be a bug rather than a performance problem. Unfortunately I don't have a high refresh rate monitor to reproduce this problem.

Can you turn on trace logging and attach a log file? I will have a look, but don't hold your breath.

[Hippo0o]

logs.tar.gz
i did:
strace picom &> picom.strace
apitrace trace picom
picom --log-level=debug &> picom.log
just dragging a floating glxgears

[yshui]

Can you do --log-level=trace?

[Hippo0o]

https://pastebin.com/2b45rcBZ

[Hippo0o]

i did try 1 monitor 60hz but results were the same.

[mighty9245]

please, upload your log to pastebin/hastebin/whateverbin and edit your comment to include link to log instead the log itself.

[yshui]

@Hippo0o I added some more logs, can you fetch the latest next branch, and attach the --log-level=trace log again?

[Hippo0o]

using picom-git aur:

$ picom --diagnostics                                                                     
**Version:** vgit-33a5c

### Extensions:

* Shape: Yes
* XRandR: Yes
* Present: Present

### Misc:

* Use Overlay: Yes* Config file used: None

### Drivers (inaccurate):

NVIDIA  

the lag is better but till there

[yshui]

@Hippo0o you are 1 commit behind. can you rebuild and get the trace log again?

[Hippo0o]

picom.log
its performing a lot better

[yshui]

@Hippo0o

its performing a lot better

How much better? Is this still a problem? If it's fine with you, can you upload a new video?

[Hippo0o]

@yshui idk why but its as worse as before again with newer version
https://www.youtube.com/watch?v=OfDWSq8d8qU
i will reboot and try again i guess.

[Hippo0o]

@yshui i found out why
i have to install the picom-git package from aur to have better performance with newest build from git?

[Hippo0o]

when i remove /etc/xdg/picom.conf which isnt included in the picom-git package it works better

[yshui]

Can you upload a video of the "good" version?

[Hippo0o]

removing

"_GTK_FRAME_EXTENTS@:c",
"_NET_WM_STATE@:32a *= '_NET_WM_STATE_HIDDEN'"

from the shadow-exclude list makes the freezes go away but the window still lags behind the cursor.
Im sure tho that it is the performance increase i mentioned

[Hippo0o]

@yshui https://youtu.be/R-LoozLQRNg?t=41

[Hippo0o]

i did another comparision to xcompmgr https://www.youtube.com/watch?v=3YxxiQUsHwU

[yshui]

@Hippo0o is there any difference between the xrender and the glx backends? And does toggling --experimental-backends make any difference?

[Hippo0o]

@yshui for all the problems i described there is no difference between those options.
I also tested the same scenarios on my intel(modesetting) gpu laptop and there were no such problems at all. It runs really well.

[WillPower3309]

Could potentially be an ultra-wide issue, I have similar (albeit less) stuttering on my 2560 x 1080 60 hz panel.

[TheToto]

I have the same problem on a simple 1920*1080 screen, amd cpu/gpu
Dragging is OK but resizing is really slow.

[Elserjo]

I also can confirm problem with laggy windows on 1920*1080 and nvidia 1070 on picom 8.1 (and it was on prev versions). It happens when i open some another window with OpenGL context, it could be mpv or Firefox (with toggled on open gl compositing)
options as xrender or experimental-backends doesn't give any effect.

https://github.com/yshui/picom/issues/366

[66RING]

I have the same problem on a simple 1920*1080 screen, amd cpu/gpu
Dragging is OK but resizing is really slow.

same

[billksun]

Could it be the VRR blacklist?

https://www.reddit.com/r/archlinux/comments/dxqj4z/psa_compton_is_now_picom/f8rvxlo?utm_source=share&utm_medium=web2x

[Sushisource]

@billksun At least in my case, I have it properly blacklisted and still experiencing super slow resizing and slow dragging.

[TheToto]

Wow, I tested again picom today and it is as fast as xcompmgr now.

[Mange]

I have 3 1440p screens, two at 75 Hz and one at 144 Hz, and when picom is running I can move around windows at around 20 fps. If I kill the process it will allow me to move around windows at the correct framerate.

I use Nvidia RTX2070 (with official drivers) + X11 + Awesome WM + picom fork with kawase blue and rounded corners.

Picom config

backend = "glx";
dbe = false;
detect-client-opacity = true;
detect-rounded-corners = true;
glx-no-stencil = true;
glx-swap-method = 2;
refresh-rate = 0;
vsync = true;

# Shadow
shadow = false; # Disabled for rounded corners for now
shadow-radius = 7;
shadow-offset-x = -7;
shadow-offset-y = -7;
shadow-opacity = 0.8;
shadow-red = 0.0;
shadow-green = 0.0;
shadow-blue = 0.0;

shadow-exclude = [
  "name = 'Notification'",
  "class_g ?= 'Notify-osd'",
  "_GTK_FRAME_EXTENTS@:c",
  "class_g ?= 'i3-frame'",
  "class_i *= 'polybar'",
  "class_i *= 'tray'",
  "class_g *= 'screenshot'",
  "class_g = 'slop'",
  "class_g = 'Conky'",
  "class_g = 'GLava'",
  "class_g = 'Firefox' && argb",
  "class_g = 'Popup'",
  "n:a:Conky",
  "class_g ?= 'maim'",
  "fullscreen",
  "width >= 3440"
];

# For i3 tabs support for flashfocus
opacity-rule = [
  "0:_NET_WM_STATE@:32a *= '_NET_WM_STATE_HIDDEN'"
];

# Blur background
blur-background = true; # Blur background of semi-transparent / ARGB windows.
blur-background-frame = true; # Blur background of windows when the window frame is not opaque. Implies --blur-background.
blur-background-fixed = true; # Use fixed blur strength instead of adjusting according to window opacity.
blur-method = "kawase"; # requires picom fork
blur-strength = 8; # requires picom fork

blur-background-exclude = [
    "class_i *= 'tray'",
    "class_g = 'Firefox' && argb",
    "class_g = 'conky'",
    "class_g = 'GLava'",
    "class_g ?= 'maim'",
    "class_g *= 'screenshot'",
    "class_g ?= 'slop'"
];

# Fading
# Fading just makes things feel slower IMO...
fading = false;
fade-exclude = [
    "width >= 3440",
    "class_g *= 'screenshot'",
    "class_g = 'slop'",
    "fullscreen"
];

# Rounded corners
corner-radius = 10.0;
round-borders = 1;

rounded-corners-exclude = [
  "class_g = 'awesome'",
  "class_g = 'Polybar'",
  "class_g = 'Firefox' && argb"
];

# Window type settings
wintypes:
{
  tooltip = { fade = true; shadow = true; opacity = 0.85; focus = true; };
  dock = { shadow = false; }
  dnd = { shadow = false; }
  popup_menu = { opacity = 0.9; }
  dropdown_menu = { opacity = 0.9; }
};

--experimental-backends does not do any difference here, although --experimental-backends fixes other display issues like flickering windows. Same config works well on my laptop with Intel / 1080p 60Hz.

If I disable blurred backgrounds AND rounded corners, then the problem is fixed. Just one of them does not, although blur and no rounded corners have a slightly better framerate than having both enabled. Could it be shader related? I've tried putting the card in "Maximum performance mode" in the Nvidia X Server Settings app.

Is there anything I should try to change in order to find a cause?