umio-yasuno / amdgpu_top Goto Github PK

have you thought about adding some way to configure the GPU inside of amdgpu_top? not sure how hard it would be in the TUI, but should be doable in the GUI

I'd be happy to lend a hand :-)

Since version 0.3, trying to open the GUI does nothing.
Does not print any error message, the cmdline tool just exits with code 0.

Hello!

For the programs that i have installed I prefer to have icons too, besides their names and descriptions, so that they are more user friendly, easier / faster to recognize and I can pin them to the start menu's Favorites section, the desktop or the task bar for quick and fast access.

Unfortunately for this program, the TUI and the simple-TUI doesn't have any icon / entry in the start menu.

I understand that it's not common to create icons / desktop entries for command line interfaces, but at the moment I cannot make up my mind which one i prefer to view stats (the GUI or the TUI) and it would be really nice to be able to find both of them in the start menu and from there to be able to either pin them to the desktop or the task bar or just click to open them.

I know that htop monitor, which is a TUI too, has an icon / desktop entry in the menu and can be opened from there with a click.

While opening its icon is also visible in the task bar.

And I assume it can be used as an example on how to create a desktop entry for a programs that runs in the terminal and probably this is its desktop entry file:

https://github.com/htop-dev/htop/blob/main/htop.desktop

At the moment I can click on the Htop icon in the start menu, or on the desktop or on the task bar and open it in the terminal.

For this program I have to open the terminal, type "amdgpu_top" and click enter.

It would be nice to be able to open the TUI version of this program as simple as I can do it with Htop.

I installed this program (v0.1.10 ) on Debian 12 + KDE Plasma 5.27.5 (on Wayland session)

So, what do you think, is this a good idea, to have another entry, maybe with a text like "AMDGPU TOP (TUI)" and an icon?

For me, the icon could even be the same as long as the name have this GUI / TUI distinction.

As for the simple TUI version, I have no idea if I would want an icon / desktop entry for that too, as I don't see myself using it much, but I don't think I would mind having and entry for that too in the menu, hopefully closed to the other two.

Thank you!

Hi there,

In version 0.3.1-1, running "amdgpu_top --dump" outputs list of variables in gpu_metrics_v1_3 at the bottom of the dump.

Some of these provided useful information not included elsewhere in the dump, such as 'average_gfx_activity'. This information could be used in scripting etc.

In 0.4.0-1, running "amdgpu_top --dump" no longer produces this output and is therefore less useful as load information is only available when starting amdgpu_top interactively.

Thanks,
Matthew Hayward.

So I noticed when I was playing a video on youtube, the pci sent measurement increased steadily up to 1440p (around 200MB/s) and then reset to almost zero at 4k resolution

Second image, I swap from 1440p youtube to 4k

When building with cargo install --locked --path ., an error launched with

undefined reference to `amdgpu_device_get_fd'

I saw amdgpu_device_get_fd was push to drm 18 months ago, but still not found in my local OS.

ps. I'm running 20.04.6, kernel version 5.15.0-72-generic

Should this value be in mW?

Tested software:

• Firefox 113.0.1 with media.ffmpeg.vaapi.enabled in about:config set to true
• mpv 0.35.1 media player with following config options in mpv.conf

profile=gpu-hq
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
hwdec=vaapi

System info:

Host: b650mds3h Kernel: 6.3.2-arch1-1 arch: x86_64 bits: 64
    compiler: gcc v: 13.1.1 Desktop: KDE Plasma v: 5.27.5 tk: Qt v: 5.15.9
    wm: kwin_wayland dm: SDDM Distro: EndeavourOS base: Arch Linux
Machine:
  Type: Desktop System: Gigabyte product: B650M DS3H v: N/A
    serial: <superuser required>
  Mobo: Gigabyte model: B650M DS3H v: x.x serial: <superuser required>
    UEFI: American Megatrends LLC. v: F6 date: 05/17/2023
Graphics:
  Device-1: AMD Navi 31 [Radeon RX 7900 XT/7900 XTX] vendor: Sapphire
    driver: amdgpu v: kernel arch: RDNA-3 pcie: speed: 16 GT/s lanes: 16 ports:
    active: DP-2 empty: DP-1,HDMI-A-1,HDMI-A-2 bus-ID: 03:00.0
    chip-ID: 1002:744c
  Device-2: AMD Raphael vendor: Gigabyte driver: amdgpu v: kernel
    arch: RDNA-2 pcie: speed: 16 GT/s lanes: 16 ports: active: none empty: DP-3,
    DP-4, DP-5, HDMI-A-3 bus-ID: 12:00.0 chip-ID: 1002:164e temp: 37.0 C
  Display: wayland server: X.org v: 1.21.1.8 with: Xwayland v: 23.1.1
    compositor: kwin_wayland driver: X: loaded: amdgpu
    unloaded: modesetting,radeon alternate: fbdev,vesa dri: radeonsi
    gpu: amdgpu,amdgpu display-ID: 0
  Monitor-1: DP-2 res: 2560x1440 size: N/A
  API: OpenGL v: 4.6 Mesa 23.0.3 renderer: AMD Radeon RX 7900 XTX (gfx1100
    LLVM 15.0.7 DRM 3.52 6.3.2-arch1-1) direct-render: Yes

Screenshots that shows the issue:

amdgpu-top version: 0.1.8

Awesome project! It would be super useful if this utility also listed the utilisation of the video decode and encode chips.

When running in a toolbox / podman container on fedora 38 Silverblue I get a panic.

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: NoGlutinConfigs(ConfigTemplate { color_buffer_type: Rgb { r_size: 8, g_size: 8, b_size: 8 }, alpha_size: 8, depth_size: 0, stencil_size: 0, num_samples: None, min_swap_interval: None, max_swap_interval: None, config_surface_types: WINDOW, api: None, transparency: false, single_buffering: false, stereoscopy: None, float_pixels: false, max_pbuffer_width: None, hardware_accelerated: None, max_pbuffer_height: None, native_window: None }, Error { raw_code: None, raw_os_message: None, kind: NotFound })', crates/amdgpu_top_gui/src/lib.rs:234:7

works fine if I run the gui version directly: not in a container.
The non-gui does work in a container, but as expected, does not have the complete view.

I do not think it should necessarily work in the container. Just that the programme should handle the Err and exit cleanly.

This is a really slick piece of s/w. Thank you!

The RPM package is generated by cargo-generate-rpm.

I have not tested it.

working branch: https://github.com/Umio-Yasuno/amdgpu_top/tree/rpm

https://github.com/Umio-Yasuno/amdgpu_top/releases/download/v0.1.9/amdgpu_top-0.1.9-1.x86_64.rpm

Hello,

I sometimes encounter a problem where running amdgpu_top is not possible. When it happens I also can't read the temperature from KDE System Monitor and the performance seems to degrade

Here is the error message I get :

amdgpu_device_initialize: DRM version is 0.0.0 but this driver is only compatible with 3.x.x.
Failed to DeviceHandle::init
Error: DevicePath { render: "/dev/dri/renderD128", card: "/dev/dri/card0", pci: None }
Fallback: list: [
    DevicePath {
        render: "/dev/dri/renderD129",
        card: "/dev/dri/card1",
        pci: Some(
            "0000:04:00.0",
        ),
    },
]
Fallback: to: DevicePath { render: "/dev/dri/renderD129", card: "/dev/dri/card1", pci: Some("0000:04:00.0") }
thread 'main' panicked at 'byte index 18446744073709551613 is out of bounds of ``', /home/coelacanth/.cargo/registry/src/index.crates.io-6f17d22bba15001f/libdrm_amdgpu_sys-0.1.6/amdgpu/device_handle.rs:347:41
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

My device is a AMD ATI 04:00.0 Renoir

The current PCIe Link Speed is showing the maximum value instead of the current value.

On my system, I have a PCIe Gen 4.0 graphics card and CPU (Radeon RX 6800 and Ryzen 7 3700X), but my motherboard only supports PCIe Gen 3.0, so the graphics card is working at Gen 3.0 speeds (8.0 GT/s), as seen on dmesg:

[    0.584108] pci 0000:0b:00.0: 126.016 Gb/s available PCIe bandwidth, limited by 8.0 GT/s PCIe x16 link at 0000:00:03.1 (capable of 252.048 Gb/s with 16.0 GT/s PCIe x16 link)

But the value on amdgpu_top is shown as:
PCIe Link Speed => Gen4x16 (Max. Gen4x16

I have read on this issue here that it's possible to get the current link speed by checking the card's pp_dpm_pcie

On my system, running cat /sys/bus/pci/devices/0000:0d:00.0/pp_dpm_pcie returns:

0: 2.5GT/s, x1 310Mhz 
1: 8.0GT/s, x16 619Mhz *

It should be displayed as mV in this case

Hello!

For the programs that i have installed I prefer to have icons too, besides their names and descriptions, so that they are more user friendly, easier / faster to recognize and I can pin them to the start menu's Favorites section, the desktop or the task bar for quick access.

Unfortunately for this program, the GUI doesn't have any icon when displayed on both the start menu and on the desktop and it has one on the task bar when it's opened, which looks like a white "W" on a yellow filled circle shape background, which I assume it's some kind of Wayland default when there's no icon or something is wrong as I've seen this icon a few times before for other programs

I tested this program (v0.1.10 ) on Debian 12 + KDE Plasma 5.27.5 (on Wayland session)

Can you please add an icon for it?

For me even a simple one like just with the letters "AT" or "AGT" (from AmdgpuTop or AmdGpuTop) would suffice.
And if you make it, could you make it in the SVG format so that it scales properly to any resolution?

Thank you!

I have 3 displays running through MST over USB C (Steam Deck). Doing raw calculations this would be 24.87 Gbps before compression.

DisplayStream Compression is most definitely being used to lower this requirement.

Are there any metrics available to see in real-time what the actual bandwidth is?

release amdgpu_top as a crate on https://crates.io/ so people having rust installed already can just run cargo install amdgpu-top

Hi,

Having build issues unfortunately...

Compiling amdgpu_top_tui v0.6.1 (/Builds/Arch/.../amdgpu_top/crates/amdgpu_top_tui)
error[E0308]: mismatched types
--> crates/amdgpu_top_tui/src/view/gpu_metrics.rs:122:14
|
122 | (metrics.get_average_gfx_power(), "mW"),
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected Option<u16>, found Option<u32>
|
= note: expected enum Option<u16>
found enum Option<u32>

error[E0308]: mismatched types
--> crates/amdgpu_top_tui/src/view/gpu_metrics.rs:226:31
|
226 | let v = check_metrics_val(gpu_metrics.get_average_socket_power());
| ----------------- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected Option<u16>, found Option<u32>
| |
| arguments to this function are incorrect
|
= note: expected enum Option<u16>
found enum Option<u32>
note: function defined here
--> /Builds/Arch/.../amdgpu_top/crates/libamdgpu_top/src/stat/gpu_metrics_util.rs:3:8
|
3 | pub fn check_metrics_val(val: Option) -> String {
| ^^^^^^^^^^^^^^^^^

Any help would be greatly appreciated.

Kind regards,

Hello!

I prefer that most of the programs on my computer open as maximized because I can see all the content in their window better and because then I can quickly close the program when I need to just by going to the top-right corner and click as there will be fore sure the close button when a window is maximized.

To prepare that, on when the program open for the first time, which will be in windowed mode, like pretty much all the programs, I double click on the title bar or press the maximize button to make it maximized and continue to use that program.

When I close it, it's maximized state is remembered and the next time I open it, its window is maximized.

Pretty much all the programs do this, except this one.

I tested this program (v0.1.10 ) on Debian 12 + KDE Plasma 5.27.5 (on Wayland session)

I could not test if the windowed (non-maximized) state is saved / remember as the program always opens in the windowed mode and without the maximized state being remembered I have no way of knowing if the program opens in windowed mode because that one it can remember it or because that's the default one

I tested if the state is remembered when the program was snapped to an edge (half-screen) but that is not remembered either.

Also, I don't know how this remembering of the window state work, if it's KDE that remembers the state for all programs or the programs themselves are remembering their own window state for the next time they are opened.

What I know is that all KDE / Qt-based programs remember their maximized state and also some of the non KDE / Qt-based ones like Firefox.

Thank you!

The PCIe Bandwidth Usage indicator always reports zero, even when under heavy GPU load.

My system is as follows:
CPU: Ryzen 7 3700X
GPU: Radeon RX 6800
Motherboard: Gigabyte B450 AORUS PRO

If any more information is needed, I'll be glad to help!

GPU: AMD R7 250

AMDGPU_TOP version: amdgpu_top 0.5.0

Kernel: Linux vince-MS-7681 6.2.0-39-generic #⁠40~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Nov 16 10:53:04 UTC 2 x86_64 x86_64 x86_64 GNU/Linux

os-release

PRETTY_NAME="Ubuntu 22.04.3 LTS"
NAME="Ubuntu"
VERSION_ID="22.04"
VERSION="22.04.3 LTS (Jammy Jellyfish)"
VERSION_CODENAME=jammy
ID=ubuntu
ID_LIKE=debian
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
UBUNTU_CODENAME=jammy

lspci

01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Oland XT [Radeon HD 8670 / R5 340X OEM / R7 250/350/350X OEM]

Logs

$ RUST_BACKTRACE=1 RUST_BACKTRACE=full amdgpu_top
There are no the AMD GPU devices found.
thread 'main' panicked at src/main.rs:18:13:
explicit panic
stack backtrace:
   0:     0x55adf7bcf5cf - <unknown>
   1:     0x55adf78cec1c - <unknown>
   2:     0x55adf7b9ab5d - <unknown>
   3:     0x55adf7bd0efe - <unknown>
   4:     0x55adf7bd0ae3 - <unknown>
   5:     0x55adf7bd1af9 - <unknown>
   6:     0x55adf7bd15f8 - <unknown>
   7:     0x55adf7bd1586 - <unknown>
   8:     0x55adf7bd1571 - <unknown>
   9:     0x55adf78cce44 - <unknown>
  10:     0x55adf78ce852 - <unknown>
  11:     0x55adf77ee5e9 - <unknown>
  12:     0x55adf77e2ee3 - <unknown>
  13:     0x55adf77eff4d - <unknown>
  14:     0x7ff8fa629d90 - __libc_start_call_main
                               at ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16
  15:     0x7ff8fa629e40 - __libc_start_main_impl
                               at ./csu/../csu/libc-start.c:392:3
  16:     0x55adf77e2e11 - <unknown>
  17:                0x0 - <unknown>

Great tool, love it. Found your reference in radeontop issue thread.

I like how radeontop reports clocks as ratio & percent of maximums. I think it would be cool in this tool, too. Most stats already report that way, but this could done for a few more stats, too. Just a suggestion.

Cool tool. I like radentop, but that is even better.

Sorting by pid I think not only is pointless, it is kind of broken. If you sort by pid, only 6 processes are shown, which basically hides what is really going on.

RDD Process is not even a correct name. It is one of the firefox processes: pid: 1676379 cmdline: /usr/lib/firefox-esr/firefox-esr -contentproc -parentBuildID 20230309232621 -prefsLen 36817 -prefMapSize 219046 -appDir /usr/lib/firefox-esr/browser 1639246 true rdd, but yes, /proc/1676379/comm shows "RDD Process"

Additional suggestion, instead of opening every fdinfo file and trying to parse it for drm info, another option is to go to fd directory and read symlinks, then only use the one that points to /dev/dri/renderD128

Best Linux AMD info tool I have ever used. I love it man. Good job.

Hi! I see a lot of great features on master such as allowing -d to be combined with -J, and the -s flag for JSON mode, though these don't appear to be on the most recent version (at least installed using cargo)? I was wondering if you planned to cut a new release soon?

Thanks 🙂

As the title says, if you set the RUSTFLAGS environment variable to -C opt-level=3 -C target-cpu=znver3 (or native when on a 3rd gen Ryzen cpu). Removing either argument will let it compile.

Note: I am on version 1.70.0 of the toolchain. The nightly version will not work because tinyvec_macros uses the feature proc_macro_span_shrink which does not exist anymore.

This should probably be reported to the upstream rust compiler though, but I thought this should still be reported here.

Error message

Hello!

After asking other projects to add their builds to Debian's repository:

OpenSnitch
evilsocket/opensnitch#798

KDiskMark
JonMagon/KDiskMark#110

I would like to ask this one too as it seems great to me and I would like to have it directly into Debian's repository for quick and easy install / reinstall and upgrade!

From what I've heard the queue for accepting new packages is quite long, so it's better to start the process sooner than later.

I understand that you don't have the resources to add support for all Linux distributions, but Debian is one of the biggest and most popular distributions out there:

Ranked 4th on DistroWatch and GamingOnLinux:

https://distrowatch.com/

https://www.gamingonlinux.com/users/statistics/#LinuxDistributionsCombined-top

And it has one of the largest repository of packages (63127 on stable and 69473 on unstable)
https://pkgs.org/

So, can you please try to put this there too?

Thank you!

BTW, I asked the GPU-viewer project the same thing as I would love to see both projects in Debian's repository:

arunsivaramanneo/GPU-Viewer#74

Maybe you can help each other with this task, even though I assume you can also ask the developers of the other two for help.

Sunshine is an open source gamestreaming solution. As such it encodes the screen contents in realtime using hardware or software encoders. In my case I'm using sunshine's VA-API support (which is based on its bundled/inbuilt ffmpeg) to stream either h264 or h265. I've double-checked that a) a stream is running b) VA-API is used (as opposed to x264 software encoding). Yet, amdgpu_top fails to detect this encoder usage.

It's noteworthy that somewhow sunshine isn't showing up either when I issue the command lsof /dev/dri/renderD128. Currently my sunshine host is explicitely configured to use VA-API. This is confirmed by the current logs stating the following:

[2023:04:12:14:34:56]: Info: // Testing for available encoders, this may generate errors. You can safely ignore those errors. //
[2023:04:12:14:34:56]: Info: 
[2023:04:12:14:34:56]: Info: // Ignore any errors mentioned above, they are not relevant. //
[2023:04:12:14:34:56]: Info: 
[2023:04:12:14:34:56]: Info: Found encoder vaapi: [h264_vaapi, hevc_vaapi]

Otherwise x264 would also be listed. And sunshine's cpu usage is also in line with hardware encoding. It just seems "hidden" somewhow.

https://github.com/Umio-Yasuno/amdgpu_top/tree/cpu_usage

Hi,
Just tried to compile your code on Ubuntu 22.04.2. Never tried a RUST/CARGO compile before, but made sure your defined dependencies are setup before compiling, however it fails as below:

I have never used, coded or compiled a rust program before, so I am at a total loss where to start. Any ideas?
All the best
Aidan

....
....
Compiling libamdgpu_top v0.1.7 (/usr/src/amdgpu_top/crates/libamdgpu_top)
error[E0432]: unresolved import std::os::fd::IntoRawFd
--> crates/libamdgpu_top/src/device_path.rs:49:17
|
49 | use std::os::fd::IntoRawFd;
| ^^^^^^^^^^^^^^^^^^^^^^ no IntoRawFd in os::fd

error[E0603]: module fd is private
--> crates/libamdgpu_top/src/device_path.rs:49:26
|
49 | use std::os::fd::IntoRawFd;
| ^^ private module
|
note: the module fd is defined here

error[E0599]: no method named into_raw_fd found for struct File in the current scope
--> crates/libamdgpu_top/src/device_path.rs:55:34
|
55 | DeviceHandle::init(f.into_raw_fd()).map_err(|v| anyhow!(v))
| ^^^^^^^^^^^ method not found in File
|
= help: items from traits can only be used if the trait is in scope
help: the following trait is implemented but not in scope; perhaps add a use for it:
|
1 | use std::os::unix::io::IntoRawFd;
|

Some errors have detailed explanations: E0432, E0599, E0603.
For more information about an error, try rustc --explain E0432.
error: could not compile libamdgpu_top due to 3 previous errors
warning: build failed, waiting for other jobs to finish...
error: failed to compile amdgpu_top v0.1.7 (/usr/src/amdgpu_top), intermediate artifacts can be found at /usr/src/amdgpu_top/target

• Implement in GUI mode only.
  • SMI mode and TUI mode are adjusted to 80 columns.
• I'm going to use fluent-rs cargo-18n.

Hello

Thank you for such a great monitoring tool.

I have installed amdgpu_top from git and run LuxMark v3.1 (GPU benchmark with the default scene - Luxball) to benchmark OpenCL performance with rocm-opencl package v2.0.0-40502-sles152.164 from openSUSE Tumbleweed (20231022), Linux 6.5.8.

I was expecting the compute load on the fdinfo chart to go to 100% during the benchmark, but nothing happened. I have an AMD Radeon RX 6800 XT Midnight card.

Same with GravityMark 1.81 Vulkan/OpenGL computer shaders benchmark, but GFX load does go up to 100% in this case.

Also, is there any chance amdgpu_top could show the decoder currently in use (AVC, HEVC, VP9 ...) ?

     │HBM Temp (C) [  655,  655,  655,  655,]                                               │ 

RX 6900 XT does not even have HBM.

I could git out Fury X and test it later there.

thread 'main' panicked at crates/libamdgpu_top/src/device_path.rs:87:57:
byte index 19 is out of bounds of ``
stack backtrace:
   0:     0x55f80965ec2c - <unknown>
   1:     0x55f809358ba0 - <unknown>
   2:     0x55f80962a3c2 - <unknown>
   3:     0x55f809660aee - <unknown>
   4:     0x55f809660270 - <unknown>
   5:     0x55f809661481 - <unknown>
   6:     0x55f809660eac - <unknown>
   7:     0x55f809660e06 - <unknown>
   8:     0x55f809660df1 - <unknown>
   9:     0x55f809356dc4 - <unknown>
  10:     0x55f8093600df - <unknown>
  11:     0x55f80935ff06 - <unknown>
  12:     0x55f809511578 - <unknown>
  13:     0x55f80950fd9a - <unknown>
  14:     0x55f809334325 - <unknown>
  15:     0x55f80932e8ec - <unknown>
  16:     0x55f809273c2a - <unknown>
  17:     0x55f809265f23 - <unknown>
  18:     0x55f809266339 - <unknown>
  19:     0x7fb245fedcd0 - <unknown>
  20:     0x7fb245fedd8a - __libc_start_main
  21:     0x55f809265e51 - <unknown>
  22:                0x0 - <unknown>

works fine on v0.5.0

Linux 5.4.265-1-MANJARO (x64-bit) | X11 | AMD A4-6210 | AMD Radeon R3 / AMD Radeon R5 M330 with forced AMDGPU drivers

Hello!

I want to be able to make the device info panel (the left one) thinner, so to reduce its width.
Or hide it completely as once that I read that info i don't need it anymore, nothing is changing there or is changing rarely and I don't need it all the time.

I noticed this problem when I wanted to see if the video hardware acceleration is working on VLC and how it's affecting the stats.

So I put VLC to play a video and snapped it to the left half of my screen and this program to the right half of my screen.

Then seen that I have only half screen available to see the stats of this program, I wanted to minimize the width of its left panel so I could see more of it's right panel that has all the sensors and graphs, but I couldn't as the left panel seems to have a minimum width that it will not go under and also there's no way to minimize / hide the left panel completely for the cases when it's not needed.

I tested this program (v0.1.10 ) on Debian 12 + KDE Plasma 5.27.5 (on Wayland session)

Can you please make it possible to change the width of the panels to as much as we want, even that it will mean that some info in that panel will be covered / not-visible or add a button to hide / restore the left panel?

Thank you!

Hi,

As of the weekend, AMDGPU_TOP GUI no longer starts under Wayland (Hyprland/Sway)...

I am on master/main branch for amdgpu_top, seatd, wlroots, Sway and Hyprland, wayland and wayland protocols.

rustc 1.72.0

Trace is as follows:-

interface 'wl_surface' has no event 2
warning: queue 0x7f0b34000ca0 destroyed while proxies still attached:
zwp_primary_selection_offer_v1#4278190082 still attached
zwp_primary_selection_offer_v1#4278190081 still attached
wl_data_offer#4278190080 still attached
zwp_primary_selection_device_v1#38 still attached
zwp_primary_selection_device_manager_v1#21 still attached
wl_data_device#37 still attached
wl_seat#36 still attached
wl_data_device_manager#35 still attached
wl_registry#33 still attached
warning: queue 0x55cdd9de14d0 destroyed while proxies still attached:
xdg_wm_base#26 still attached
wl_output#15 still attached
wp_fractional_scale_manager_v1#14 still attached
xdg_activation_v1#13 still attached
zwp_text_input_manager_v3#12 still attached
zwp_relative_pointer_manager_v1#11 still attached
zwp_pointer_constraints_v1#10 still attached
zxdg_decoration_manager_v1#9 still attached
wl_seat#8 still attached
wp_viewporter#7 still attached
wl_subcompositor#6 still attached
wl_compositor#5 still attached
wl_shm#4 still attached
wl_registry#2 still attached

Probably due to recent commits to 'wlroots' but I could be wrong.

Kind regards,

When running compute processes I find myself monitoring GPU memory usage a lot, and it would be great if the GUI version could have a VRAM plot, much like the fdinfo plot. I think it would be the last set of parameters to get a plot.

╭ zeph  ~/code/nixcfg 0s
╰─▶ /nix/store/hhclh187i2pkll9i2cqghbbizk7nd06j-amdgpu_top-0.1.7/bin/amdgpu_top
No such file or directory (os error 2)
render_path = "/dev/dri/renderD128"
card_path = "/dev/dri/card0"
thread 'main' panicked at 'explicit panic', src/main.rs:92:21
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

I'm trying to package this for nixpkgs and use it on my G14 2022 (AMD+AMD), but not having a lot of luck so far.

Any advice? Setting RUST_BACKTRACE to 1 or full didn't really help.

when running for a long time all the graphs in the gui have massive time stamps in scientific notation that blend into each other

Hi there. Very cool tool 🙂

Is it planned to be able to (if possibly i don't know the technicals) show GTT per process similar to regular vram? And maybe also the ability to show both as a combined value.
This could maybe be useful on systems with shared memory where only a small portion is reported as dedicated VRAM e.g. Steam Deck.

When using a dark theme in the terminal, the forced white background doesn't look great.

Discussed in #20

Hello greetings from Solus Linux! Trying to update the Solus package to latest today and I'm getting an error when I try to build it. Solus builds with the following two commands:

# Setup Phase:
cargo fetch --locked

# Build Phase:
cargo build --frozen --release \
                     --no-default-features \
                     --features "tui gui json"

This gives us a reproducible build as the build will always use the exact versions of dependencies as declared in the cargo.lock. It appears that the cargo.lock hasn't been updated for v0.2.2 as we're getting the following error:

+ cargo fetch --locked
    Updating crates.io index
    Updating git repository `https://github.com/Umio-Yasuno/libdrm-amdgpu-sys-rs`
error: the lock file /home/build/YPKG/root/amdgpu_top/build/amdgpu_top-0.2.2/Cargo.lock needs to be updated but --locked was passed to prevent this
If you want to try to generate the lock file without accessing the network, remove the --locked flag and use --offline instead.

Could you please take a look at this and perhaps issue a 0.2.2.1 with the issue corrected?

Also, just a thought. When I checked out the source and did a cargo update against it I noticed a bunch of third-party dependencies get bumps. I'd recommend setting something up like Renovate or Dependabot, both of which are fairly easy to setup and will help automate keeping your dependencies updated.

Hi, on the screenshots I can see that amdgpu_top is capable of displaying decoder and encoder usage per process, but for some reason it is not working for me. Nvtop shows only decoder usage but labels it as encoder which is also kinda weird.

Instead of leaving the application iconless we could use the generic system monitor icon. Only 1 line would need to be added.

Icon=utilities-system-monitor

This is the same icon file that the KDE System Monitor uses so most icon packs should include it.

When trying to launch new version with --gui, nothing happens. I guess some changes were made which resulted in GUI not working as intended.

Hello, would that be possible to also have the NPU usage ?

I have a 7840U, let me know if you need anything

More info:

• https://videocardz.com/newz/amd-ryzen-8040-neural-processing-unit-monitoring-will-soon-be-available-in-windows-task-manager
• https://community.amd.com/t5/ai/upcoming-windows-task-manager-update-will-add-npu-monitoring-for/ba-p/664382

I installed via deb package:

❯ amdgpu_top -V   
amdgpu_top v0.5.0

But running smi mode doesn't work:

❯ amdgpu_top --smi

Doesn't print anything.

rocm-smi works normally:

========================= ROCm System Management Interface =========================
=================================== Concise Info ===================================
GPU  Temp (DieEdge)  AvgPwr  SCLK  MCLK   Fan  Perf  PwrCap  VRAM%  GPU%
0    51.0c           4.0W    0Mhz  96Mhz  0%   auto  100.0W   11%   0%
====================================================================================
=============================== End of ROCm SMI Log ================================

When scaling is used (in my case 2) under wayland GUI crashes with:

$ RUST_BACKTRACE=1 amdgpu_top --gui
wl_surface@24: error 2: Buffer size (359x465) is not divisible by scale (2)
Protocol error 2 on object wl_surface@24: 
Protocol error 2 on object wl_surface@24: 
Protocol error 2 on object wl_surface@24: 
Failed to set up a graphics context (OpenGL).
winit EventLoopError: Exit Failure: 1
thread 'main' panicked at crates/amdgpu_top_gui/src/lib.rs:274:9:
explicit panic
stack backtrace:
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

--list output:

$ amdgpu_top --list
amdgpu_top v0.7.0

#0:
    AMD Radeon Graphics (0x15BF.0xC4)
    DevicePath { render: "/dev/dri/renderD128", card: "/dev/dri/card1", pci: "0000:c1:00.0" }

Window manager is Sway. If scaling is set to 1 it starts fine.

I download the released AppImage, and failed to launch it, the it raised the following errors:

./amdgpu_top-x86_64.AppImage: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.33' not found (required by ./amdgpu_top-x86_64.AppImage)
./amdgpu_top-x86_64.AppImage: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found (required by ./amdgpu_top-x86_64.AppImage)
./amdgpu_top-x86_64.AppImage: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.34' not found (required by ./amdgpu_top-x86_64.AppImage)

The system info is:

LSB Version:	core-11.1.0ubuntu2-noarch:printing-11.1.0ubuntu2-noarch:security-11.1.0ubuntu2-noarch
Distributor ID:	Ubuntu
Description:	Ubuntu 20.04.5 LTS
Release:	20.04
Codename:	focal

The gcc is:

gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.