Skip to content

Ensure max number of SM are allocated by mig-parted#85

Open
JunAr7112 wants to merge 1 commit intoNVIDIA:mainfrom
JunAr7112:update_sm_loop
Open

Ensure max number of SM are allocated by mig-parted#85
JunAr7112 wants to merge 1 commit intoNVIDIA:mainfrom
JunAr7112:update_sm_loop

Conversation

@JunAr7112
Copy link

No description provided.

Copy link

@rajathagasthya rajathagasthya left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This fix looks like a band-aid, not a comprehensive solution as noted in NVIDIA/mig-parted#158.

} else {
pInfo := p.GetInfo()
existingInfo := existing.GetInfo()
if pInfo.GIProfileID < existingInfo.GIProfileID {

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this correct? This prefers lowest profile IDs while changes are made to iterate from the higher IDs first. Doesn't the higher profile IDs have more SMs?

Copy link
Author

@JunAr7112 JunAr7112 Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wanted to preserve the lowest profile preference for the devicelib function as it is not directly relevant to the changes needed to ensure max number of SM are allocated per device. This will also ensure that our current mig-parted tests will pass.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Removed this change. Not necessary

Signed-off-by: Arjun <agadiyar@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants