Repository navigation
Conversation
An imported PinnedMemoryResource reported numa_id 0 whenever the driver could not report the pool's location (a CUDA 12 build or a driver before 13.2). Unpickling a resource now carries the exporter's node, and PinnedMemoryResource.from_allocation_handle accepts an optional numa_id for a raw allocation handle. An import whose node is unknown still reports 0 and does not pass it on when pickled. from_allocation_handle raises ValueError for a negative numa_id, and for one that differs from the pool's known node, whether the driver reported it or the resource was already registered in the process. Closes NVIDIA#3508. Signed-off-by: Ralf Juengling <rjuengling@nvidia.com>
Contributor
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
closes #3508
An imported
PinnedMemoryResourcereportednuma_id == 0when the driver cannot report the pool's location (a CUDA 12 build or a driver before 13.2); #3516 fixed the case where it can. Now the exporter's node travels with the pickle, andPinnedMemoryResource.from_allocation_handletakes an optionalnuma_idfor raw handles. An import whose node is still unknown reports 0 as before and sends no node when re-pickled. A negativenuma_id, or one that differs from the pool's known node, raisesValueError.Tested with a CUDA 13.5 build; the no-location-query paths are covered by patching
driver_versionin the child process. A CUDA 12 build and a multi-node NUMA host were not tested.Checklist