Remove memory access flags from cuda_async_memory_resource #1754
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Outdated suggestions cannot be applied.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 #1753
It is a follow up from #1743
I would like for rapidsai/cudf#17553 to merge first, that way I don't break the build.
I've learned that I was using
cudaMemPoolSetAccess
incorrectly. This API should only be used from apeer
device, not from the device that created the pool. This is the reason why callingcudaMemPoolSetAccess
with none throws an error as documented here #1753. I have tested that I can still export the fabric handles and import them using UCX in a peer device with the default access that pool owner device gets (read+write is the default). Note that this read+write default access cannot be revoked from the owner, as it wouldn't make sense to have memory that nobody has access to, but peers can callcudaMemPoolSetAccess
to gain read+write access or to stop accessing (none) a peer's pool memory.Checklist