[DRAFT]: add support for some Tegra NVDEC/NVENC/VIC monitoring, generic GPU usage, and EMC usage #1322
Conversation
added mem clock and controller load (if current permissions permit reading them), gpu clock, gpu load, and cpu temp
theofficialgman
force-pushed
the
master-l4t
branch
from
ace2895
to
3669a6c
Compare
| gpu_core_clock | ||
| gpu_nvdec_clock | ||
| gpu_nvenc_clock | ||
| gpu_vic_clock |
There was a problem hiding this comment.
can't these just fit under the gpu_core_clock param?
There was a problem hiding this comment.
wdym? like enable all of them if only the gpu_core_clock param is enabled? I thought the granularity would be nice since not all users will want to waste space showing all of them if they only want one.
There was a problem hiding this comment.
I might not be understand the structure correct.
Can all 3-4 of these be monitored at the same time?
There was a problem hiding this comment.
these are different hardware units. their clocks are different and independent
There was a problem hiding this comment.
So I fully understand, one device will only have one of these?
In that case I don't see a point in the granularity
There was a problem hiding this comment.
you do not understand. no
these are different hardware engines on the same SOC. nvdec (nvidia decode engine), nvenc (nvidia encode engine), and vic (nvidia video image compositor) are all present on the same SOC.
example during chromium hardware accelerated video decode and playback:
There was a problem hiding this comment.
Thank you for clarifying. It seems fine to have different params for this.
As for how it's displayed, this seems a bit clunky but I also don't know a better way to do it
also add loop for gpu.0/devfreq and exit on first entry
I am opening this as a draft as it is not mergeable in its current state.
This PR adds support for some of the Nvidia Tegra system of chips that have NVDEC/NVENC/VIC stats available via
/sys/kernel/debuginterfaces. These are the same stats as used in tegrastats. They require that the current user be able to read from this directory (not the default) so I am looking for advise on implementing some sort of functionality that would allow for mangohud to read these files without needing to modify permissions on every reboot of that directory.This PR also adds generic support for reading the
CPU-thermtemperature sensors (also as used on the Tegra chips and potentially other SOCs)Finally, this PR also adds support for reading the GPU load from
/sys/devices/gpu.0/loadand a hardware specific frequency from/sys/devices/gpu.0/devfreq/57000000.gpu/cur_freq(a loop could be added to find the required folder, similar to https://github.com/hakandundar34coding/system-monitoring-center/blob/e0c6246d38af487a378adf486eed61485eea7566/src/Gpu.py#L657-L663).As previously explained, this PR is not mergeable in its current state and I am looking for feedback on its contents and where each of these functionality can be best placed. I can split up the PR into separate PRs if desired based on feedback.
closes #864