picom backends: xrender less laggy than glx?

Hi! :) This is not really a bug report, but rather a question out of curiosity. I wonder why with my current configuration and setup, the xrender backend is less laggy than the glx backend. Both prevent tearing, but when I move a floating window around, it lags much more with the glx backend than with the xrender backend. Note: I am referring to the "experimental" backends! I would be really interested to learn why this could be the case or if I did something wrong in my config. I'm a long-term user of compton/picom, but most of the time I used the intel Xorg driver with the TearFree option and disabled vsync in the compositor. But since a few months, I finally switched to the modesetting driver because there has been a bug in the intel driver that affected me (and because the modesetting driver is said to be more performant anyway).

Platform

Fedora 34 (pre-release), kernel 5.11.14-300.fc34.x86_64

GPU, drivers, and screen setup

Intel Corporation WhiskeyLake-U GT2 [UHD Graphics 620], modesetting driver for Xorg, external 1920x1080 monitor connected to laptop via DisplayPort (over USB-C). vainfo: Driver version: Intel i965 driver for Intel(R) Coffee Lake - 2.4.1 glxinfo -B:

name of display: :0
display: :0  screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Intel (0x8086)
    Device: Mesa Intel(R) UHD Graphics 620 (WHL GT2) (0x3ea0)
    Version: 21.0.2
    Accelerated: yes
    Video memory: 3072MB
    Unified memory: yes
    Preferred profile: core (0x1)
    Max core profile version: 4.6
    Max compat profile version: 4.6
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 3.2
OpenGL vendor string: Intel
OpenGL renderer string: Mesa Intel(R) UHD Graphics 620 (WHL GT2)
OpenGL core profile version string: 4.6 (Core Profile) Mesa 21.0.2
OpenGL core profile shading language version string: 4.60
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

OpenGL version string: 4.6 (Compatibility Profile) Mesa 21.0.2
OpenGL shading language version string: 4.60
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile

OpenGL ES profile version string: OpenGL ES 3.2 Mesa 21.0.2
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20

Environment

i3 :)

picom version

picom --diagnostics:

**Version:** v8.2

### Extensions:

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

### Misc:

* Use Overlay: No (Another compositor is already running)
* Config file used: /home/ro/.config/picom.conf

### Drivers (inaccurate):

modesetting

Configuration:

grep '^[^#]' .config/picom.conf:

shadow = false;
fading = false
backend = "xrender";
vsync = true;
mark-wmwin-focused = true;
mark-ovredir-focused = true;
detect-rounded-corners = true;
detect-client-opacity = true;
refresh-rate = 0;
detect-transient = true;
detect-client-leader = true;
glx-no-stencil = true
glx-no-rebind-pixmap = true
use-damage = true;
log-level = "warn";

Thank you very much! :heart:

Apr 20 '21 14:04 ro-i

I am on nvidia (proprietary driver) and also noticed that glx tends to be more laggy, but in other way - gpu accelerated terminals (alacritty, kitty) take ~1.5x more time to start compared to xrender:

Tests

With picom --experimental-backends --backend=glx --no-vsync --config=/dev/null

~ took 3s ❯ hyperfine 'alacritty -e false'
Benchmark #1: alacritty -e false
  Time (mean ± σ):     188.1 ms ±  11.5 ms    [User: 98.6 ms, System: 46.6 ms]
  Range (min … max):   171.0 ms … 213.9 ms    15 runs

Alacritty takes about ~190ms to launch.

Same with legacy backend: picom --backend=glx --no-vsync --config=/dev/null

~ took 3s ❯ hyperfine 'alacritty -e false'
Benchmark #1: alacritty -e false
  Time (mean ± σ):     184.4 ms ±  21.1 ms    [User: 94.0 ms, System: 43.2 ms]
  Range (min … max):   145.4 ms … 214.4 ms    17 runs

Now with xrender:

Experimental: picom --experimental-backends --backend=xrender --no-vsync --config=/dev/null

~ ❯ hyperfine 'alacritty -e false'
Benchmark #1: alacritty -e false
  Time (mean ± σ):     124.1 ms ±  10.8 ms    [User: 84.2 ms, System: 37.2 ms]
  Range (min … max):   114.0 ms … 148.6 ms    19 runs

Legacy: picom --backend=xrender --no-vsync --config=/dev/null

~ took 2s ❯ hyperfine 'alacritty -e false'
Benchmark #1: alacritty -e false
  Time (mean ± σ):     126.2 ms ±   7.9 ms    [User: 87.9 ms, System: 38.4 ms]
  Range (min … max):   114.2 ms … 143.4 ms    23 runs

Notice how it takes only ~125ms on average now.

Hyperfine.

Jun 20 '21 10:06 MahouShoujoMivutilde

Probably related: #641

Jun 20 '21 18:06 yshui

Probably related: #641

This, specifically there have been changes since v8.2 that affect floating window performance. (I noticed this very clearly while doing my testing).

@ro-i If you try the current git branch (next), I think you will observe much less floating window lag.

@MahouShoujoMivutilde Which version of picom are you using? The release version (8.2) or the current branch? Also, since you're on NVIDIA, I think you'll benefit from both the current branch and flag testing I'm doing.

Jun 25 '21 23:06 kwand

@kwand i am on latest next, yes.

Driver version 460.56, from nvidia-all, kernel is linux-tkg 5.10.y lts with muqss and not much else (i found this combination of versions avoids random freezes (which "been fixed", but in reality just became harder to trigger, but still happen sometimes on latest drivers)).

Also my gpu is quite old - gtx 970.

So i tried your patch, but sadly it seems there is no significant difference.

Tests

no #641, e09679c:

~ ❯ ./test.sh
glx
Benchmark #1: alacritty -e false
  Time (mean ± σ):     183.6 ms ±  12.6 ms    [User: 89.3 ms, System: 40.0 ms]
  Range (min … max):   166.8 ms … 221.7 ms    17 runs

xrender
Benchmark #1: alacritty -e false
  Time (mean ± σ):     121.4 ms ±   9.7 ms    [User: 81.3 ms, System: 35.1 ms]
  Range (min … max):   113.7 ms … 147.4 ms    22 runs

e09679c with #641:


~ took 2s ❯ ./test.sh
glx
Benchmark #1: alacritty -e false
  Time (mean ± σ):     181.2 ms ±  14.2 ms    [User: 86.1 ms, System: 37.8 ms]
  Range (min … max):   162.5 ms … 203.6 ms    16 runs

xrender
Benchmark #1: alacritty -e false
  Time (mean ± σ):     121.5 ms ±   9.0 ms    [User: 84.0 ms, System: 39.4 ms]
  Range (min … max):   113.0 ms … 146.2 ms    23 runs

test.sh

#!/usr/bin/env sh

pid=/tmp/picom.pid

run() {
    echo "$1"
    picom -b --write-pid-path="$pid" --experimental-backends --no-vsync --backend="$1" --config=/dev/null
    sleep 0.5s
    hyperfine 'alacritty -e false'
    kill $(cat $pid)
    sleep 0.5s
}

run glx

run xrender

(I made sure patch applied, tests are on new picom, etc)

Also, you guys been talking about triple buffering, so i think it's may be worth noting that i have it enabled in xorg config:

/etc/X11/xorg.conf.d/20-nvidia.conf

Section "Screen"
    Identifier     "Screen0"

    # for vsync without compositor
    Option         "metamodes" "nvidia-auto-select +0+0 { ForceCompositionPipeline=On, ForceFullCompositionPipeline = On }"
    Option         "TripleBuffer" "On"
    Option         "AllowIndirectGLXProtocol" "Off"

    # Option         "ConnectToAcpid" "Off"

EndSection

Jun 26 '21 15:06 MahouShoujoMivutilde

Also, you guys been talking about triple buffering, so i think it's may be worth noting that i have it enabled in xorg config:

I've been talking about disabling triple buffering. The fact that you have it enabled is might conflict with the patch, which tries to turn it off for picom.

There's no real reason to have triple buffering enabled for a compositor, unless you want to prioritize smoothness at the cost of responsiveness.

Though, I will note the difference with my patch is only slight. There is another method that I tried (it's mentioned in the first post of the PR), which may or may not benefit you, but I've abandoned it now since it introduces a bunch of old glitches and will require a major rewrite of the code, for just 5-10ms less latency. Don't get me wrong that this amount of latency reduction is still significant - it's just that the rewrite is many times more significant that I can't really justify it (at least for my implementation)

Also, I don't really believe your GPU is old enough to be the problem here. I think it's likely it is not even boosting while rendering picom, which should make it perform almost on par with newer GPUs.

Jun 26 '21 20:06 kwand

Option         "metamodes" "nvidia-auto-select +0+0 { ForceCompositionPipeline=On, ForceFullCompositionPipeline = On }"

Also, I just noticed that you have this line in your xorg config. Is this turned on? (i.e. ForceCompositionPipeline and ForceFullCompositionPipeline? I don't really understand the syntax of this config file unfortunately)

If it is, I would turn it off while testing - I've noticed increased lag when it's on and it's a bit superfluous as picom is already trying to do the same thing. The comment you have above seems to say the same thing: "for vsync without compositor"

I still can't quite explain why glX takes 1.5x longer though - even in theory. My best guess (without evidence) is that glX may be more graphically demanding than xrender? If that is the case, I would try re-performing the test with "Prefer Maximum Performance" enabled, as shown here:

This should force the GPU to run at its maximum boost clock.

Jun 26 '21 21:06 kwand

unless you want to prioritize smoothness at the cost of responsiveness.

Actually, it seems that triple buffering gives me significantly better latency.

No triple buffering, ForceCompositionPipeline=On:

glx
Benchmark #1: alacritty -e false
  Time (mean ± σ):     216.5 ms ±  12.8 ms    [User: 101.4 ms, System: 46.3 ms]
  Range (min … max):   197.7 ms … 231.3 ms    13 runs

xrender
Benchmark #1: alacritty -e false
  Time (mean ± σ):     166.6 ms ±   7.9 ms    [User: 107.7 ms, System: 44.8 ms]
  Range (min … max):   147.7 ms … 181.4 ms    17 runs

No triple buffering and no ForceCompositionPipeline=On

~ took 25s ❯ ./test.sh
glx
Benchmark #1: alacritty -e false
  Time (mean ± σ):     228.1 ms ±  20.0 ms    [User: 129.5 ms, System: 64.3 ms]
  Range (min … max):   181.5 ms … 260.4 ms    13 runs

xrender
Benchmark #1: alacritty -e false
  Time (mean ± σ):     164.8 ms ±   7.3 ms    [User: 100.9 ms, System: 43.2 ms]
  Range (min … max):   147.5 ms … 180.7 ms    19 runs

No ForceCompositionPipeline=on but triple buffering is on

glx
Benchmark #1: alacritty -e false
  Time (mean ± σ):     180.1 ms ±  15.1 ms    [User: 103.0 ms, System: 48.3 ms]
  Range (min … max):   152.8 ms … 200.6 ms    18 runs

xrender
Benchmark #1: alacritty -e false
  Time (mean ± σ):     121.3 ms ±  10.4 ms    [User: 85.3 ms, System: 37.7 ms]
  Range (min … max):   112.4 ms … 148.2 ms    23 runs

* when i write "no something", what i mean is that i just commenting this thing out from config above. Should not make any difference, according to defaults.

Though, I will note the difference with my patch is only slight

Doesn't hurt to try it anyway. And thank you for trying to come up with improvements :+1:

it's just that the rewrite is many times more significant that I can't really justify it

Yeah, makes sense.

Also, I just noticed that you have this line in your xorg config. Is this turned on?

Yes, but it doesn't actually add any significant latency (see above), in fact - the biggest increase in latency regarding xorg config was from disabling TripleBuffer.

comment you have above seems to say the same thing: "for vsync without compositor"

I wrote that myself, and that's why all my tests are with --no-vsync ;)

Theoretically, having vsync on driver level should give less latency compared to picom --vsync, but that i haven't tested yet.

However, at the moment, fastest runs were with ForceCompositionPipeline=On.

I would try re-performing the test with "Prefer Maximum Performance" enabled, as shown here

I just tried that, and results are the same. P state switches to 4 even with default setting fast enough to not matter.

Jun 26 '21 22:06 MahouShoujoMivutilde

I wrote that myself, and that's why all my tests are with --no-vsync ;)

Well, this was a complete oversight of mine. I didn't notice that at all! I can see why my patch probably has no effect then since I believe it would only work if vsync is enabled.

Thank you for doing the testing - I'm currently unable to do much due to time constraints. They're all quite interesting, and unfortunately, I have no idea how to explain the results.

(Actually, it's possible I was mistaken about enabling triple buffering in the xorg conf forces picom to use triple buffering as well. To reiterate, this does not really matter in your case since you disabled vsync, but it's possible you're seeing gains b/c of enabling/disabling triple buffering for alacritty.)

Theoretically, having vsync on driver level should give less latency compared to picom --vsync, but that i haven't tested yet.

Would love to see the results for this as well, whenever you have the time, since my PR mainly improves vsync inside picom.

Jun 27 '21 00:06 kwand

@kwand Okay, i tested whole bunch of things.

Buckle up

"Sync to vblank" and "allow flipping" are from nvidia OpenGL settings. I found by accident that they both were enabled, and after playing around with them it turned out they influence glx latency quite significantly, so i decided to include them as well.

ffcp is ForceCompositionPipeline stuff from Xorg config above, nvidia's driver level vsync.

Patched means #641 applied.

Each row is 100 counted runs + 3 warm up runs.

backend	picom vsync	sync to vblank	allow flipping	triple buffer	ffcp	patched	mean	median	stddev
xrender	True	True	False	True	False	True	113	114	3.56
xrender	True	False	True	True	False	True	113	114	3.46
xrender	False	True	True	True	False	True	113	114	3.41
xrender	True	False	False	True	False	True	113	114	3.77
xrender	True	True	False	True	True	True	114	113	2.93
xrender	True	False	False	True	True	True	114	114	2.62
xrender	False	True	False	True	False	True	114	114	3.34
xrender	False	False	False	True	False	True	114	114	3.6
glx	False	False	False	True	False	True	114	114	2.88
xrender	False	False	True	True	False	True	114	114	5
xrender	True	True	True	True	False	True	114	114	2.94
xrender	True	True	False	True	False	False	114	114	3.17
xrender	True	False	False	True	False	False	114	114	3.26
xrender	True	True	True	True	False	False	114	114	3.4
xrender	False	True	False	True	True	False	114	114	3.28
xrender	True	False	False	True	True	False	114	114	2.4
xrender	False	False	True	True	True	False	114	114	3.14
xrender	False	True	False	True	True	True	115	114	3.71
xrender	True	False	True	True	True	True	115	114	3.23
xrender	False	False	False	True	True	True	115	114	3.54
xrender	False	True	True	True	True	True	115	114	3.28
xrender	False	False	True	True	True	True	115	114	4.08
xrender	True	True	True	True	True	True	115	114	2.99
glx	False	True	False	True	False	True	115	114	4.35
xrender	True	False	True	True	False	False	115	114	3.17
xrender	False	False	False	True	False	False	115	114	2.9
xrender	False	True	True	True	False	False	115	114	3.42
xrender	False	False	True	True	False	False	115	114	3.43
xrender	True	True	False	True	True	False	115	114	2.82
xrender	True	False	True	True	True	False	115	114	3.53
xrender	False	False	False	True	True	False	115	114	3.14
xrender	False	True	True	True	True	False	115	114	3.39
glx	False	False	False	True	True	False	115	115	3.88
xrender	True	True	True	True	True	False	115	114	2.99
glx	False	False	False	True	True	True	116	115	3.83
xrender	False	True	False	True	False	False	116	114	3.63
glx	False	False	False	True	False	False	116	114	4.21
glx	False	True	False	True	True	True	117	115	5.02
glx	False	True	False	True	False	False	118	116	4.97
glx	False	True	False	True	True	False	118	115	5.13
glx	True	False	False	True	False	True	137	137	10.22
glx	True	True	False	True	False	True	138	134	10.57
glx	True	True	False	True	True	False	142	146	9.62
glx	True	False	False	True	True	False	142	147	9.51
glx	True	True	False	True	True	True	144	147	8.18
glx	True	False	False	True	True	True	144	147	8.42
xrender	True	True	False	False	True	True	147	148	9.89
xrender	True	True	True	False	True	True	147	148	8.99
xrender	True	True	False	False	True	False	148	148	10.72
xrender	True	True	True	False	True	False	149	148	9.88
glx	True	True	False	True	False	False	149	148	9.46
glx	True	False	False	True	False	False	149	148	9.92
glx	False	True	False	False	True	False	150	148	11.08
xrender	True	False	True	False	True	True	150	148	9.64
xrender	True	False	False	False	True	True	150	148	9.62
glx	False	False	False	False	True	True	150	148	10.61
xrender	True	True	False	False	False	True	150	148	9.51
glx	False	True	False	False	True	True	151	148	10.77
xrender	False	True	False	False	True	True	151	148	9.63
xrender	False	True	True	False	True	True	151	148	10.3
xrender	True	False	False	False	True	False	152	148	10.12
glx	False	False	False	False	True	False	152	148	9.54
xrender	False	False	False	False	True	True	152	148	10.22
xrender	True	True	True	False	False	False	152	148	8.93
xrender	False	True	False	False	True	False	153	148	9.55
xrender	False	True	True	False	True	False	153	148	9.68
xrender	True	False	False	False	False	False	153	148	8.19
xrender	True	False	False	False	False	True	153	148	9.62
xrender	True	True	True	False	False	True	153	148	9.51
xrender	True	False	True	False	True	False	154	148	10.67
xrender	False	False	True	False	True	True	154	148	9.42
glx	False	True	False	False	False	False	154	148	8.89
glx	False	True	False	False	False	True	154	148	10.25
xrender	True	False	True	False	False	True	154	148	10.16
xrender	False	False	False	False	True	False	155	148	9.48
xrender	False	False	True	False	True	False	156	163	8.89
xrender	True	True	False	False	False	False	156	156	9.73
xrender	True	False	True	False	False	False	156	151	8.25
xrender	False	True	True	False	False	False	156	162	8.97
glx	False	False	False	False	False	False	156	149	8.47
glx	False	False	False	False	False	True	156	163	9.56
xrender	False	False	True	False	False	True	156	164	9.01
xrender	False	True	False	False	False	False	157	164	8.74
xrender	False	True	False	False	False	True	157	164	8.67
xrender	False	False	False	False	False	True	157	164	8.57
xrender	False	True	True	False	False	True	157	164	8.63
glx	True	True	False	False	True	True	158	164	8.63
glx	True	False	False	False	True	True	158	164	9.89
xrender	False	False	False	False	False	False	158	164	8.25
xrender	False	False	True	False	False	False	158	164	8.26
glx	True	True	False	False	False	True	158	164	8.26
glx	False	False	True	True	True	True	159	164	9.93
glx	True	True	False	False	True	False	159	164	11.03
glx	True	False	False	False	False	True	159	164	10.28
glx	True	False	False	False	True	False	161	164	14.34
glx	True	True	False	False	False	False	161	164	13.07
glx	False	False	True	True	True	False	162	164	10.37
glx	True	False	False	False	False	False	164	164	15.33
glx	True	True	True	True	True	False	168	165	6.59
glx	True	False	True	True	True	True	169	166	6.1
glx	True	True	True	True	True	True	169	166	8.23
glx	False	True	True	True	False	True	169	170	16.2
glx	False	False	True	True	False	True	169	166	8.9
glx	False	True	True	True	True	False	169	165	7.09
glx	False	False	True	True	False	False	170	168	10.32
glx	True	False	True	True	True	False	170	165	8.91
glx	False	True	True	True	True	True	172	168	10.47
glx	False	True	True	True	False	False	175	177	10.42
glx	True	False	True	True	False	True	200	198	5.73
glx	False	False	True	False	True	False	203	198	10.89
glx	False	False	True	False	False	False	205	198	11.78
glx	False	True	True	False	True	False	206	198	14.83
glx	True	True	True	True	False	True	207	213	9.38
glx	True	False	True	True	False	False	209	213	7.82
glx	True	True	True	True	False	False	209	213	7.88
glx	False	True	True	False	False	False	210	214	14.74
glx	False	True	True	False	True	True	211	213	13.85
glx	False	False	True	False	True	True	212	206	16.12
glx	False	False	True	False	False	True	213	214	12.3
glx	False	True	True	False	False	True	215	214	16.15
glx	True	False	True	False	True	False	234	231	14.11
glx	True	True	True	False	True	False	236	232	12.62
glx	True	False	True	False	True	True	236	232	14.75
glx	True	True	True	False	True	True	238	232	14.03
glx	True	True	True	False	False	False	283	281	21.35
glx	True	False	True	False	False	False	284	289	19.28
glx	True	True	True	False	False	True	284	281	22.15
glx	True	False	True	False	False	True	288	290	20.69

For reference, this is how latency looks without picom running

~ ❯ hyperfine -r 100 -w 3 'alacritty --class floatme,floatme -e false'
Benchmark #1: alacritty --class floatme,floatme -e false
  Time (mean ± σ):     114.4 ms ±   2.8 ms    [User: 89.8 ms, System: 39.3 ms]
  Range (min … max):   108.4 ms … 123.5 ms    100 runs

Same table in json

  [
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 113,
      "median": 114,
      "stddev": 3.56
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 113,
      "median": 114,
      "stddev": 3.46
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 113,
      "median": 114,
      "stddev": 3.41
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 113,
      "median": 114,
      "stddev": 3.77
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 114,
      "median": 113,
      "stddev": 2.93
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 2.62
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 3.34
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 3.6
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 2.88
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 5.0
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 114,
      "median": 114,
      "stddev": 2.94
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 3.17
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 3.26
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 3.4
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 3.28
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 2.4
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 114,
      "median": 114,
      "stddev": 3.14
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 3.71
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 3.23
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 3.54
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 3.28
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 4.08
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 2.99
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 115,
      "median": 114,
      "stddev": 4.35
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.17
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 2.9
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.42
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.43
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 2.82
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.53
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.14
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 3.39
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 115,
      "stddev": 3.88
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 115,
      "median": 114,
      "stddev": 2.99
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 116,
      "median": 115,
      "stddev": 3.83
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 116,
      "median": 114,
      "stddev": 3.63
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 116,
      "median": 114,
      "stddev": 4.21
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 117,
      "median": 115,
      "stddev": 5.02
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 118,
      "median": 116,
      "stddev": 4.97
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 118,
      "median": 115,
      "stddev": 5.13
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 137,
      "median": 137,
      "stddev": 10.22
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 138,
      "median": 134,
      "stddev": 10.57
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 142,
      "median": 146,
      "stddev": 9.62
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 142,
      "median": 147,
      "stddev": 9.51
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 144,
      "median": 147,
      "stddev": 8.18
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 144,
      "median": 147,
      "stddev": 8.42
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 147,
      "median": 148,
      "stddev": 9.89
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 147,
      "median": 148,
      "stddev": 8.99
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 148,
      "median": 148,
      "stddev": 10.72
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 149,
      "median": 148,
      "stddev": 9.88
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 149,
      "median": 148,
      "stddev": 9.46
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 149,
      "median": 148,
      "stddev": 9.92
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 150,
      "median": 148,
      "stddev": 11.08
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 150,
      "median": 148,
      "stddev": 9.64
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 150,
      "median": 148,
      "stddev": 9.62
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 150,
      "median": 148,
      "stddev": 10.61
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 150,
      "median": 148,
      "stddev": 9.51
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 151,
      "median": 148,
      "stddev": 10.77
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 151,
      "median": 148,
      "stddev": 9.63
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 151,
      "median": 148,
      "stddev": 10.3
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 152,
      "median": 148,
      "stddev": 10.12
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 152,
      "median": 148,
      "stddev": 9.54
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 152,
      "median": 148,
      "stddev": 10.22
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 152,
      "median": 148,
      "stddev": 8.93
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 153,
      "median": 148,
      "stddev": 9.55
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 153,
      "median": 148,
      "stddev": 9.68
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 153,
      "median": 148,
      "stddev": 8.19
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 153,
      "median": 148,
      "stddev": 9.62
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 153,
      "median": 148,
      "stddev": 9.51
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 154,
      "median": 148,
      "stddev": 10.67
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 154,
      "median": 148,
      "stddev": 9.42
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 154,
      "median": 148,
      "stddev": 8.89
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 154,
      "median": 148,
      "stddev": 10.25
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 154,
      "median": 148,
      "stddev": 10.16
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 155,
      "median": 148,
      "stddev": 9.48
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 156,
      "median": 163,
      "stddev": 8.89
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 156,
      "median": 156,
      "stddev": 9.73
    },
    {
      "backend": "xrender",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 156,
      "median": 151,
      "stddev": 8.25
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 156,
      "median": 162,
      "stddev": 8.97
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 156,
      "median": 149,
      "stddev": 8.47
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 156,
      "median": 163,
      "stddev": 9.56
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 156,
      "median": 164,
      "stddev": 9.01
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 157,
      "median": 164,
      "stddev": 8.74
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 157,
      "median": 164,
      "stddev": 8.67
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 157,
      "median": 164,
      "stddev": 8.57
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 157,
      "median": 164,
      "stddev": 8.63
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 158,
      "median": 164,
      "stddev": 8.63
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 158,
      "median": 164,
      "stddev": 9.89
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 158,
      "median": 164,
      "stddev": 8.25
    },
    {
      "backend": "xrender",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 158,
      "median": 164,
      "stddev": 8.26
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 158,
      "median": 164,
      "stddev": 8.26
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 159,
      "median": 164,
      "stddev": 9.93
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 159,
      "median": 164,
      "stddev": 11.03
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 159,
      "median": 164,
      "stddev": 10.28
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 161,
      "median": 164,
      "stddev": 14.34
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 161,
      "median": 164,
      "stddev": 13.07
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 162,
      "median": 164,
      "stddev": 10.37
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": false,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 164,
      "median": 164,
      "stddev": 15.33
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 168,
      "median": 165,
      "stddev": 6.59
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 169,
      "median": 166,
      "stddev": 6.1
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 169,
      "median": 166,
      "stddev": 8.23
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 169,
      "median": 170,
      "stddev": 16.2
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 169,
      "median": 166,
      "stddev": 8.9
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 169,
      "median": 165,
      "stddev": 7.09
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 170,
      "median": 168,
      "stddev": 10.32
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": false,
      "mean": 170,
      "median": 165,
      "stddev": 8.91
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": true,
      "patched": true,
      "mean": 172,
      "median": 168,
      "stddev": 10.47
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 175,
      "median": 177,
      "stddev": 10.42
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 200,
      "median": 198,
      "stddev": 5.73
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 203,
      "median": 198,
      "stddev": 10.89
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 205,
      "median": 198,
      "stddev": 11.78
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 206,
      "median": 198,
      "stddev": 14.83
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": true,
      "mean": 207,
      "median": 213,
      "stddev": 9.38
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 209,
      "median": 213,
      "stddev": 7.82
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": true,
      "ffcp": false,
      "patched": false,
      "mean": 209,
      "median": 213,
      "stddev": 7.88
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 210,
      "median": 214,
      "stddev": 14.74
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 211,
      "median": 213,
      "stddev": 13.85
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 212,
      "median": 206,
      "stddev": 16.12
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 213,
      "median": 214,
      "stddev": 12.3
    },
    {
      "backend": "glx",
      "picom_vsync": false,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 215,
      "median": 214,
      "stddev": 16.15
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 234,
      "median": 231,
      "stddev": 14.11
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": false,
      "mean": 236,
      "median": 232,
      "stddev": 12.62
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 236,
      "median": 232,
      "stddev": 14.75
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": true,
      "patched": true,
      "mean": 238,
      "median": 232,
      "stddev": 14.03
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 283,
      "median": 281,
      "stddev": 21.35
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": false,
      "mean": 284,
      "median": 289,
      "stddev": 19.28
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": true,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 284,
      "median": 281,
      "stddev": 22.15
    },
    {
      "backend": "glx",
      "picom_vsync": true,
      "sync_to_vblank": false,
      "allow_flipping": true,
      "triple_buffer": false,
      "ffcp": false,
      "patched": true,
      "mean": 288,
      "median": 290,
      "stddev": 20.69
    }
  ]

So what i noticed:

You can make glx (--no-vsync) as fast as xrender while keeping ffcp and triple buffering. Nice!
Xrender --vsync doesn't actually work (there is still tearing), and latency says the same.
ffcp (115ms) is indeed faster than picom glx --vsync (137ms) .
Allow flipping generally slows glx down (40-60ms difference for both picom --vsync and ffcp).
Sync to vblank does not seem to matter a lot.
Your patch is indeed helps for glx --vsync (patched vs unpatched is about 10ms difference!), xrender and ffcp don't care.
Triple buffering generally improves latency.

Note: If you use a tiling window manager, it is important to launch alacritty as floating window to minimize margin of error.

Methodology

To at least partially automate testing i wrote a new script

test2.sh

It assumes running in tmux (to collect hyperfine's output as it was on screen), but it's not strictly necessary.

It runs picom/nvidia with different options and tests them on alacritty via hyperfine (which also saves json for each run).

It can't, however, change Xorg options (ffcp, triple buffering), so that is expected to be done by hand, since they will require Xorg restart.

It also doesn't install patched/unpatched picom for you.

First argument is expected to be a name for results folder of particular Xorg options (name should be (no-3b|3b),(no-ffcp|ffcp),(641|def) for sort-times.py below).

This folder will be created under ~/picom. Terminal output (tmux buffer after run, essentially) will be saved under the same name but with .txt.

  #!/usr/bin/env bash

pid=/tmp/picom.pid

[ -z "$1" ] && echo -e "usage: \ntest2.sh 'results-folder-name'" && exit 1

clear

results="$HOME/picom/$1"

mkdir -p $results

run() {
    opts="$1, $4, SyncToVBlank=$2, AllowFlipping=$3"
    echo "Running with $opts"

    nvidia-settings -a "SyncToVBlank=$2" >/dev/null
    nvidia-settings -a "AllowFlipping=$3" >/dev/null

    picom $4 -b --write-pid-path="$pid" \
        --experimental-backends --backend="$1" \
        --config=/dev/null

    sleep 0.5s

    hyperfine --export-json "$results/$opts.json" \
        -w 3 -r 100 \
        'alacritty --class floatme,floatme -e false'

    kill "$(cat $pid)"

    sleep 0.5s
}

figlet 'Xorg options:'

grep -P '(Triple|ForceComp)' /etc/X11/xorg.conf.d/20-nvidia.conf

echo -e '\n\n'
figlet 'GLX'
echo -e '\n***   --no-vsync   ***\n'

run glx 0 0 --no-vsync
run glx 0 1 --no-vsync
run glx 1 0 --no-vsync
run glx 1 1 --no-vsync

echo -e '\n***     --vsync    ***\n'

run glx 0 0 --vsync
run glx 0 1 --vsync
run glx 1 0 --vsync
run glx 1 1 --vsync


echo -e '\n\n'
figlet 'XRENDER'
echo -e '\n***   --no-vsync   ***\n'

run xrender 0 0 --no-vsync
run xrender 0 1 --no-vsync
run xrender 1 0 --no-vsync
run xrender 1 1 --no-vsync

echo -e '\n***     --vsync    ***\n'

run xrender 0 0 --vsync
run xrender 0 1 --vsync
run xrender 1 0 --vsync
run xrender 1 1 --vsync

tmux capture-pane -pS -1000000 > "$results.txt"

sort-times.py

This script goes through every folder (Xorg options) under set results (set to ~/picom, or run it inside), parses hyperfine's .json files and outputs pretty table and runs.json file, like you seen above.

#!/usr/bin/env python3

import json
import os
from typing import Iterable

# https://pypi.org/project/tabulate/
from tabulate import tabulate

# $results from test2.sh
results = './'


def find(results: str) -> Iterable[str]:
    '''
    Finds every .json file in subdirectories of the `results` directory,
    ignores files in `results`'s root.
    '''
    for dirpath, _, filenames in os.walk(results):
        for fn in filenames:
            if os.path.splitext(fn)[1] == '.json' and dirpath != './':
                yield os.path.join(dirpath, fn)


def parse(fp: str) -> dict:
    '''
    Takes file path to .json result of a run, returns dict with parsed options used / latency
    '''
    run = {}

    name = os.path.splitext(os.path.basename(fp))[0]
    run['backend'] = name.split(',')[0]
    run['picom_vsync'] = name.find('--vsync') != -1
    run['sync_to_vblank'] = name.find('SyncToVBlank=1') != -1
    run['allow_flipping'] = name.find('AllowFlipping=1') != -1

    dir = os.path.basename(os.path.dirname(fp))
    run['triple_buffer'] = dir.find('no-3b') == -1
    run['ffcp'] = dir.find('no-ffcp') == -1
    run['patched'] = dir.find('def') == -1

    with open(fp, 'r', encoding='utf-8') as f:
        timings = json.load(f)

    # to ms
    run['mean'] = round(timings['results'][0]['mean'] * 1000)
    run['median'] = round(timings['results'][0]['median'] * 1000)
    run['stddev'] = round(timings['results'][0]['stddev'] * 1000, 2)

    return run


runs = [parse(fp) for fp in find(results)]

runs = sorted(runs, key=lambda r: r['mean'])

with open('./runs.json', 'w', encoding='utf-8') as ff:
    json.dump(runs, ff, indent=2)

# for markdown set tablefmt='github'
print(tabulate(runs, tablefmt="fancy_grid", headers='keys'))

Here is a notebook selecting rows with various options to get a feel for latency.

picom-data.tar.gz

Also, if you notice that after reboot everything is slow again, remember that

The NVIDIA X driver does not preserve values set with nvidia-settings between runs of the X server.

from nvidia-settings man page.

So don't forget to add nvidia-settings --load-config-only in autostart. I don't yet know how to set allow flipping = 0 and sync to vblack = 0 in Xorg.conf .

Jun 28 '21 13:06 MahouShoujoMivutilde

@MahouShoujoMivutilde Very sorry for the late reply. This is not much of an update, but I just wanted to let you know that I have read your reply and actually switched to using settings that give the lowest latency (as per your results) a month ago.

The results seem to be right as I seem to notice some latency improvement. But it still puzzles me why there's such a huge difference and how we could improve picom's performance (OK, maybe not a "huge difference". 22-23ms seems to be just slightly more than one frame of latency worse than FFCP, assuming your display is 60Hz)

I have yet to run your tests on my own machine though, but I imagine I'll probably discover something once I do (as I'm now using a 160Hz display). I also have access to a laptop running AMD graphics now, so I want to do some investigation in that area to see if this is NVIDIA-specific problem or something inherent in picom.

Sorry, I've just been really short on time lately, though I would really love to fix this problem myself (I get pretty annoyed at the latency difference when I notice how fast everything is when I need to kill picom sometimes, and when comparing picom to Windows).

If anyone has the time, I think this is a pretty high-priority issue that could be looked into. Investigating how KDE handles their compositing v-sync algorithm might also be worthwhile, since apparently they have a superior algorithm than picom's.

I haven't daily-driven Plasma on my main machine yet (running awesome-wm now), so I can't say much about their claims of latency improvement - though I do know that they were quite infamous for terrible latency before that latency-improving update. Since many people are claiming that issue is fixed, I'd imagine it has to have been quite a significant improvement (or maybe they became too used to terrible latency that any improvement looked subjectively better. I don't know)

Jan 20 '22 01:01 kwand

sorry to notify literally any one associated with this thread, but just a question: Is xrender better than GLX backend on nvidia propietary drivers? and what can i do to lower latency for both backends? i have disabled Flipping which made it much smoother,

    Option         "ForceFullCompositionPipeline" "on"
    Option         "AllowIndirectGLXProtocol" "off"
    Option         "TripleBuffer" "on"

for me picom can take up to 20% cpu (i use animations) and gpu takes 70% with 70W average.

Sep 28 '22 14:09 apprehensions

picom picom copied to clipboard

backends: xrender less laggy than glx?

Platform

GPU, drivers, and screen setup

Environment

picom version

Configuration:

picom
picom copied to clipboard