[FEA] Hash function refactoring #13706
Labels
0 - Backlog
In queue waiting for assignment
feature request
New feature or request
libcudf
Affects libcudf (C++/CUDA) code.
Comments
bdice
added
feature request
This was referenced Jul 17, 2023
GregoryKimball
added
0 - Backlog
|
|
@davidwendt proposed removing unsanitized nulls from hashing tests. I agree with this idea. I refactored MD5's tests in 0188115 and will do the same for SHA in PR #14391, but additional work is needed for other hashing algorithms to remove unsanitized nulls from the tests. |
|
We also need to standardize null behavior for hashing algorithms. See #10451. |
3 tasks
rapids-bot bot
pushed a commit
that referenced
this issue
Feb 20, 2024
The `cudf::hashing::spark_murmurhash3_x86_32()` function was moved to the Spark plugin since it had common code with the Spark implementation of `xxhash_64` (also implemented in the plugin). This change deprecates the API and the generic `cudf::hashing::hash()` function to be removed in a follow-on release. Reference hash cleanup issue: #13706 Authors: - David Wendt (https://github.com/davidwendt) Approvers: - Bradley Dice (https://github.com/bdice) - Karthikeyan (https://github.com/karthikeyann) URL: #15074
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Following up from #13681 and #13612, there are some tasks I think can be done to clean up hashing code. I am opening this issue to be a tracker for the work we've deferred from other PRs.
getblock32can be reused by multiple hash functions. https://github.com/rapidsai/cudf/pull/13612/files#r1265587501Also consider optimizing this to use
memcpyorif constexpron key type perhaps.rotate_bits_leftshould not use CUDA intrinsics (it doesn't affect the PTX/SASS compared to a naive shift-based implementation), thereby making themconstexpr-friendly and possible to put into a sharedhppheader. Separate MurmurHash32 from hash_functions.cuh #13681 (comment)@briefof hashing APIs should mention the algorithm name. https://github.com/rapidsai/cudf/pull/13612/files#r1267270076The text was updated successfully, but these errors were encountered: