[WIP][RFC] Add sparse host buffer source #16252
Draft
Conversation
This file contains 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
5 tasks
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Outdated suggestions cannot be applied.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.
Related to #15919
DISCLAIMER: This is meant to be a rough RFC to illustrate my idea for a temporary mitigation strategy for
NativeFileremoval. I am probably not the right person to take the libcudf changes over the line but I still threw some C++ code together for fun :)Idea: We don't need
NativeFilesupport to get reasonable partial-IO performance from remote storage if we only transfer the necessary byte ranges into host memory with fsspec and libcudf is able to read from the sparse<offset, byte-range>mapping. The fsspec component is covered in #16166, but that PR currently wastes a lot of host-memory by copying the necessary byte ranges into a larger "proxy" byte range that matches the size of the actual file.@vuule @vyasr - Does a "sparse" host buffer source seem like a reasonable approach for the near term?