diff options
| author | Kévin Le Gouguec <kevin.legouguec@gmail.com> | 2025-02-02 18:01:51 +0100 |
|---|---|---|
| committer | Kévin Le Gouguec <kevin.legouguec@gmail.com> | 2025-02-02 18:01:51 +0100 |
| commit | c682631f888cba13a8de489d2a435fce24578571 (patch) | |
| tree | 30efacb3ea53025f3a9aa339f6c1b4bbd701f66d /guides/sysadmin/machines | |
| parent | 26e3c9d598bbefbf8862b0aa135dbed6ed0f2089 (diff) | |
| download | memory-leaks-c682631f888cba13a8de489d2a435fce24578571.tar.xz | |
Add a sysfs knob to the pile
Diffstat (limited to 'guides/sysadmin/machines')
| -rw-r--r-- | guides/sysadmin/machines/amdahl30/killing-time.org | 23 |
1 files changed, 18 insertions, 5 deletions
diff --git a/guides/sysadmin/machines/amdahl30/killing-time.org b/guides/sysadmin/machines/amdahl30/killing-time.org index 611b8cb..bbf5bfb 100644 --- a/guides/sysadmin/machines/amdahl30/killing-time.org +++ b/guides/sysadmin/machines/amdahl30/killing-time.org @@ -291,6 +291,19 @@ not overwork it to death? /Software/, logging out & back in frees up some VRAM. It also yields compositing glitches 🤷 - More to the point, /it has no effect on the symptoms in-game/. + +- Figured I would ask ValveSoftware/Proton about the logs; filed + [[https://github.com/ValveSoftware/Proton/issues/8424][#8424]]; got dup'd into [[https://github.com/ValveSoftware/Proton/issues/7805][#7805]], per the "one report per game" policy. + That issue is about a /crash on Alt-Tab/, with an /Nvidia dGPU/; + unsure how lumping our two reports together will help. Had to try 🤷 + +- Found [[https://gitlab.freedesktop.org/drm/amd/-/issues/2516][drm/amd#2516]]; noticed that I have + - =/sys/module/gpu_sched/parameters/sched_policy=: 1 + - =/sys/module/amdgpu/parameters/sched_policy=: 0 + Changed the kernel command-line to set the former to 0, as suggested + in that issue; symptoms persist. No idea what the latter is about, + nor how it differs from the former. I can find [[https://docs.kernel.org/gpu/amdgpu/module-parameters.html#sched-policy-int][the docs for amdgpu]] + but nor for gpu_sched. * This is insane Selected subset of moving parts; "testability" considering ease of clean reverts: @@ -309,11 +322,11 @@ clean reverts: That's looking at software packages as individual blackboxes; config-wise, worth noting: -| Part | Testability | -|-----------------+--------------------| -| AMD pstate | 😬 UEFI roundtrip | -| sysfs | OK | -| Plasma Renderer | OK | +| Part | Testability | +|-----------------+---------------------------------------------------| +| AMD pstate | 😬 UEFI roundtrip | +| sysfs | OK; worst case: reboot & edit kernel command-line | +| Plasma Renderer | OK | Let's throw in: |