feature: Support AMDGPU Data Collection #1641

yretenai · 2024-12-02T00:11:59Z

Description

Adds GPU metrics gathering ~~via amdgpu_top's libamdgpu_top crate~~ on Linux.

Some notes: the library queries /proc/{pid}/fdinfo, which can probably be parsed without amdgpu_top's libraries. Intel Arc apparently also uses this fdinfo but I cannot confirm.

Testing

If relevant, please state how this was tested. All changes must be tested to work:

If this is a code change, please also indicate which platforms were tested:

Windows
macOS
Linux

Checklist

If relevant, ensure the following have been met:

Areas your change affects have been linted using rustfmt (cargo fmt)
The change has been tested and doesn't appear to cause any unintended breakage
Documentation has been added/updated if needed (README.md, help menu, doc pages, etc.)
The pull request passes the provided CI pipeline
There are no merge conflicts
If relevant, new tests were added (don't worry too much about coverage)

codecov · 2024-12-02T00:17:22Z

Codecov Report

Attention: Patch coverage is 6.62983% with 338 lines in your changes missing coverage. Please review.

Project coverage is 41.37%. Comparing base (1fe17dd) to head (95eab0f).
Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
src/data_collection/amd.rs	3.33%	319 Missing ⚠️
src/data_collection.rs	40.62%	19 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1641      +/-   ##
==========================================
- Coverage   42.03%   41.37%   -0.67%     
==========================================
  Files         116      118       +2     
  Lines       17625    17926     +301     
==========================================
+ Hits         7409     7417       +8     
- Misses      10216    10509     +293

Flag	Coverage Δ
macos-14	`37.38% <0.00%> (-0.08%)`	⬇️
ubuntu-latest	`43.09% <6.62%> (-0.73%)`	⬇️
windows-2019	`37.31% <0.00%> (-0.08%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

yretenai · 2024-12-02T00:20:35Z

CI failing due to lacking libdrm/libdrm_amdgpu libraries. What's the best way forward here?

ClementTsang · 2024-12-02T02:04:35Z

Can the libraries be installed?

yretenai · 2024-12-02T04:25:06Z

libdrm should exist in most linux distros via their package managers, i'm atm investigating a dependency-free solution by parsing /proc/*/fdinfo/*

yretenai · 2024-12-04T03:56:25Z

Most recent commit parses AMD GPU metrics via procfs (for per-process utilization and video memory usage) and sysfs (for overall AMDGPU memory usage and temperature sensors) and as such doesn't rely on any libraries.

However the code is significantly more complex.

ClementTsang

Left a few comments for now, mostly looks good though.

src/data_collection.rs

ClementTsang · 2024-12-06T16:02:30Z

src/data_collection/amd.rs

+}
+
+// needs previous state for usage calculation
+static PROC_DATA: LazyLock<Mutex<HashMap<PathBuf, HashMap<u32, AMDGPUProc>>>> =


non-blocking: This doesn't need to be changed for this PR, but a mutex around this seems a little overkill.

It's the only way I could get a mutable reference to PROC_DATA without making the static itself mutable, and thus requiring unsafe code to access.

src/data_collection/amd.rs

ClementTsang · 2024-12-06T16:06:12Z

src/data_collection/amd.rs

+    // get vram memory info from sysfs
+    let vram_total_path = device_path.join("mem_info_vram_total");
+    let vram_used_path = device_path.join("mem_info_vram_used");


Would you happen to know whether any of these checks in this file wake up the GPU if it is currently sleeping? We had an issue with temperature checks in https://github.com/ClementTsang/bottom/blob/main/src/data_collection/temperature/linux.rs#L226 where we were waking up devices (mainly GPUs) when checking their temperature, for example.

I do not, but I can ask around.

ClementTsang · 2024-12-06T16:09:37Z

src/data_collection/amd.rs

+    })
+}
+
+pub fn get_amd_temp(device_path: &Path) -> Option<Vec<AMDGPUTemperature>> {


Does this differ from https://github.com/ClementTsang/bottom/blob/main/src/data_collection/temperature/linux.rs#L226?

Besides having better names, no. They end up symlinking to the same hwmon endpoint

Hm, guess it's fine to leave it in for now but I might integrate this into that code path in a later PR.

yretenai · 2024-12-06T21:29:38Z

Left a few comments for now, mostly looks good though.

I will make the necessary changes tomorrow.

yretenai · 2024-12-07T18:00:03Z

Force pushes were rewording commit messages.

Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

gpu: fix clippy issues Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

…ead of current memory usage gpu: requested syntax changes Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

yretenai · 2024-12-07T18:03:53Z

I accidentally reset the signature of the 4th commit from HEAD, which I just fixed by resetting the entire branch, apologies! History should be preserved now.

ClementTsang · 2024-12-17T07:22:13Z

Sorry for the delay, was on vacation - will take a look at this again in a sec.

ClementTsang

LGTM, thanks for doing this!

ClementTsang · 2024-12-20T07:46:34Z

@all-contributors please add @yretenai for code.

allcontributors · 2024-12-20T07:46:43Z

@ClementTsang

I've put up a pull request to add @yretenai! 🎉

This MR contains the following updates: | Package | Update | Change | |---|---|---| | [ClementTsang/bottom](https://github.com/ClementTsang/bottom) | minor | `0.10.2` -> `0.11.0` | MR created with the help of [el-capitano/tools/renovate-bot](https://gitlab.com/el-capitano/tools/renovate-bot). **Proposed changes to behavior should be submitted there as MRs.** --- ### Release Notes <details> <summary>ClementTsang/bottom (ClementTsang/bottom)</summary> ### [`v0.11.0`](https://github.com/ClementTsang/bottom/blob/HEAD/CHANGELOG.md#0110---2025-08-05) [Compare Source](ClementTsang/bottom@0.10.2...0.11.0) ##### Features - [#1625](ClementTsang/bottom#1625): Add the ability to configure the disk widget's table columns. - [#1641](ClementTsang/bottom#1641) + [#1692](ClementTsang/bottom#1692): Support AMD GPU data collection on Linux. - [#1642](ClementTsang/bottom#1642): Support changing the widget borders. - [#1717](ClementTsang/bottom#1717): Support delete key (fn + delete on macOS) to kill processes. - [#1306](ClementTsang/bottom#1306): Support using left/right key to collapse/expand process trees respectively. - [#1767](ClementTsang/bottom#1767): Add a virtual memory column for processes. - [#1770](ClementTsang/bottom#1770) (originally [#1627](ClementTsang/bottom#1627)): Add option to have process tree entries be collapsed by default. ##### Bug Fixes - [#1551](ClementTsang/bottom#1551): Fix missing parent section names in default config. - [#1552](ClementTsang/bottom#1552): Fix typo in default config. - [#1565](ClementTsang/bottom#1565): Fix issue where CPU usage in basic mode looks weird if core count isn't divisible by four. - [#1578](ClementTsang/bottom#1578): Fix missing selected text background colour in `default-light` theme. - [#1593](ClementTsang/bottom#1593): Fix using `"none"` for chart legend position in configs. - [#1594](ClementTsang/bottom#1594): Fix incorrect default config definitions for chart legends. - [#1596](ClementTsang/bottom#1596): Fix support for nilfs2 file system. - [#1660](ClementTsang/bottom#1660): Fix properly cleaning up the terminal if the program is terminated due to an `Err` bubbling to the top. - [#1663](ClementTsang/bottom#1663): Fix network graphs using log scaling having broken lines when a point was 0. - [#1667](ClementTsang/bottom#1667): Fix for ARC/SWAP not being hidden in basic mode after refactor. - [#1683](ClementTsang/bottom#1683): Fix graph lines potentially showing up behind legends. - [#1701](ClementTsang/bottom#1701): Fix process kill dialog occasionally causing panics. - [#1755](ClementTsang/bottom#1755): Fix missing stats/incorrect mount name for certain entries in the disk widget. - [#1759](ClementTsang/bottom#1759): Fix increment for data tables if the change is greater than the number of entries left. ##### Changes - [#1559](ClementTsang/bottom#1559): Rename `--enable_gpu` to `--disable_gpu`, and make GPU features enabled by default. - [#1570](ClementTsang/bottom#1570): Consider `$XDG_CONFIG_HOME` on macOS when looking for a default config path in a backwards-compatible fashion. - [#1686](ClementTsang/bottom#1686): Allow hyphenated arguments to work as well (e.g. `--autohide-time`). - [#1701](ClementTsang/bottom#1701): Redesign process kill dialog. - [#1706](ClementTsang/bottom#1706): Disable mouse capture when `disable_click` is set. - [#1769](ClementTsang/bottom#1769): Change how we calculate swap usage in Windows. ##### Other - [#1655](ClementTsang/bottom#1655): Better handle NVIDIA GPUs on Linux with only libnvidia-ml.so.1. - [#1658](ClementTsang/bottom#1658): Make it possible to override completion/manpage generation output directory via env. - [#1663](ClementTsang/bottom#1663): Rework how data is stored internally, reducing memory usage a bit. - [#1749](ClementTsang/bottom#1749): Fix invalid desktop file values. </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever MR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this MR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this MR, check this box --- This MR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate).

ClementTsang self-assigned this Dec 2, 2024

ClementTsang changed the title ~~Support AMDGPU Data Collection~~ feature: Support AMDGPU Data Collection Dec 6, 2024

ClementTsang reviewed Dec 6, 2024

View reviewed changes

yretenai and others added 3 commits December 7, 2024 18:01

gpu: support amdgpu tracking

dee4b40

Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

gpu: dependency-free amdgpu parsing

0f69df5

gpu: fix clippy issues Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

gpu: change memory usage percentage to be scaled to total memory inst…

95eab0f

…ead of current memory usage gpu: requested syntax changes Co-authored-by: lvxnull2 <184518908+lvxnull2@users.noreply.github.com>

jamartin9 mentioned this pull request Dec 10, 2024

GPU and VRAM load #298

Open

ClementTsang approved these changes Dec 20, 2024

View reviewed changes

allcontributors bot mentioned this pull request Dec 20, 2024

docs: add yretenai as a contributor for code #1646

Merged

ClementTsang merged commit 479276b into ClementTsang:main Dec 20, 2024
37 checks passed

ClementTsang added a commit that referenced this pull request Dec 20, 2024

docs: update changelog for #1641

c366afc

ClementTsang added a commit that referenced this pull request Dec 20, 2024

docs: update changelog for #1641 (#1647)

35662fc

BrewTestBot mentioned this pull request Aug 6, 2025

bottom 0.11.0 Homebrew/homebrew-core#232509

Merged

feature: Support AMDGPU Data Collection #1641

feature: Support AMDGPU Data Collection #1641

Uh oh!

Conversation

yretenai commented Dec 2, 2024 • edited by ClementTsang Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Testing

Checklist

Uh oh!

codecov bot commented Dec 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

yretenai commented Dec 2, 2024

Uh oh!

ClementTsang commented Dec 2, 2024

Uh oh!

yretenai commented Dec 2, 2024

Uh oh!

yretenai commented Dec 4, 2024

Uh oh!

ClementTsang left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ClementTsang Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

yretenai Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ClementTsang Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

yretenai Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

ClementTsang Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

yretenai Dec 6, 2024

Choose a reason for hiding this comment

Uh oh!

ClementTsang Dec 20, 2024

Choose a reason for hiding this comment

Uh oh!

yretenai commented Dec 6, 2024

Uh oh!

yretenai commented Dec 7, 2024

Uh oh!

yretenai commented Dec 7, 2024

Uh oh!

ClementTsang commented Dec 17, 2024

Uh oh!

ClementTsang left a comment

Choose a reason for hiding this comment

Uh oh!

ClementTsang commented Dec 20, 2024

Uh oh!

allcontributors bot commented Dec 20, 2024

Uh oh!

Uh oh!

Uh oh!

yretenai commented Dec 2, 2024 •

edited by ClementTsang

Loading

codecov bot commented Dec 2, 2024 •

edited

Loading