[REQ] add ability to display VRAM Usage

[NicTanghe]

I'm currently trying to troubleshoot something that is probably related to VRAM usage.
It would b very handy if HTOP could also display this info.

[BenBE]

What do you mean by this exactly? Is there an other tool that does this already or can you point to some documentation for this feature?

[NicTanghe]

Right now I`m using,
nvidia-smi -l --query-gpu=timestamp,temperature.gpu,memory.used,memory.free --format=csv
(for an nvidia Gpu)
Output:

nvidia-smi -l --query-gpu=timestamp,temperature.gpu,memory.used,memory.free --format=csv
timestamp, temperature.gpu, memory.used [MiB], memory.free [MiB]
2022/10/04 14:44:54.814, 46, 487 MiB, 7472 MiB
2022/10/04 14:44:59.816, 46, 487 MiB, 7472 MiB
2022/10/04 14:45:04.816, 46, 487 MiB, 7472 MiB

https://nvidia.custhelp.com/app/answers/detail/a_id/3751/~/useful-nvidia-smi-queries

[BenBE]

Given that this is highly dependent on the vendor of the GPU used we decided to not support this in htop directly, as otherwise we'd have to implement this for all vendors or get nagged to do it for the others we didn't do yet.

One planned feature that could help in this regard is #526. Another references worth taking a look at regarding GPU meters is e.g. #406 (GPU utilization).

[NicTanghe]

I Wil close this as it is a duplicate of 406:

Since there are only 2 or soon 3 vendors, it shouldn’t b that hard to just check what vendor it is and then run it`s part of the code.

Closed as its a duplicate of #406