incusd/device/gpu_sriov: Implement NUMA fallback (NPS) #641
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For systems with multiple NUMA nodes per socket (NPS), it's always better to select a GPU on the same socket, even if on a different node than going across socket.
This adds a bit of logic to map the current NUMA nodes to the matching set of CPU sockets and then extract all NUMA nodes from those sockets to build a list of fallback NUMA nodes. Those are less optimal than a direct match but still better than potentially going across socket.
Sponsored-by: ActivePort (https://www.activeport.com.au)