fix(api/libnvml): fix removal for process info v3 APIs on the upstream 535.98 driver - Githubissues

XuehaiPan / nvitop

An interactive NVIDIA-GPU process viewer and beyond, the one-stop solution for GPU process management.

https://nvitop.readthedocs.io

Apache License 2.0

4.56k stars 144 forks source link

fix(api/libnvml): fix removal for process info v3 APIs on the upstream 535.98 driver #89

Closed XuehaiPan closed 1 year ago

XuehaiPan commented 1 year ago

Issue Type

Bug fix

Description

Add an extra handler to check the removal of the process info v3 APIs.

Copied from https://github.com/XuehaiPan/nvitop/issues/88#issuecomment-1680052592:

Version change:

495.46 -> 510.39.01: https://github.com/NVIDIA/nvidia-settings/commit/b2f0e7f437c42d92ed58120ec8d880f5f4b90d60

Add process info v3 APIs but use v2 nvmlProcessInfo_st struct type
default:
- nvmlDeviceGetComputeRunningProcesses -> nvmlDeviceGetComputeRunningProcesses_v3
- nvmlProcessInfo_st -> nvmlProcessInfo_v2_st

530.41.03 -> 535.43.02: https://github.com/NVIDIA/nvidia-settings/commit/39c3e28e84f3ffb034abaf1ae92dbb570c207d05

Process info v3 APIs use v3 nvmlProcessInfo_st struct type without a version bump
default:
- nvmlDeviceGetComputeRunningProcesses -> nvmlDeviceGetComputeRunningProcesses_v3
- nvmlProcessInfo_st -> nvmlProcessInfo_v3_st

535.86.05 -> 535.98: https://github.com/NVIDIA/nvidia-settings/commit/0cb3beffa0cb8a1f8cb405291b11a1e2eb7a4786

Remove process info v3 APIs and v3 nvmlProcessInfo_st struct type
default:
- nvmlDeviceGetComputeRunningProcesses -> nvmlDeviceGetComputeRunningProcesses_v2
- nvmlProcessInfo_st -> nvmlProcessInfo_v2_st

UPDATE:

535.98 -> 535.104.05: https://github.com/NVIDIA/nvidia-settings/commit/74cae7fa6a3da595a1bd87918ef0a67bb4326925

Re-add process info v3 APIs but use v2 nvmlProcessInfo_st struct type
default:
- nvmlDeviceGetComputeRunningProcesses -> nvmlDeviceGetComputeRunningProcesses_v3
- nvmlProcessInfo_st -> nvmlProcessInfo_v2_st

Motivation and Context

Fixes #88