Improve performance for cudf::contains when searching for a scalar
#11202
Conversation
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
ttnghia
added
3 - Ready for Review
Contributor
Author
|
Here is the benchmark results, comparing the performance after vs before this PR: |
Contributor
Author
|
Here is the original benchmark results: Before:After: |
This comment was marked as off-topic.
This comment was marked as off-topic.
vuule
reviewed
Jul 6, 2022
Contributor
vuule
left a comment
vuule
left a comment
There was a problem hiding this comment.
Looks good, just a few suggestions.
cpp/benchmarks/search/search.cpp
Outdated
Comment on lines
91
to
92
| BINARY_SEARCH_BENCHMARK_DEFINE(Column_AllValid, false) | ||
| BINARY_SEARCH_BENCHMARK_DEFINE(Column_HasNulls, true) |
Contributor
There was a problem hiding this comment.
Not sure if intentional, but AFAICT we don't need separate definitions here. Validity can be another parameter, something like (if CreateRange is available in the version that we use)
->ArgsProduct({
benchmark::CreateRange(100000, 100000000, 10),
{0,1}
})
Contributor
Author
There was a problem hiding this comment.
Oh I didn't know that there is such way. I'm trying that...
Contributor
Author
There was a problem hiding this comment.
The current Google benchmark version used in cudf doesn't support CreateRange. I've created another PR to update it: #11209
Contributor
Author
There was a problem hiding this comment.
@vuule It seems that we can't upgrade Google benchmark, so unfortunately your suggestion here can't be worked on.
Contributor
There was a problem hiding this comment.
I think we should look at this as an opportunity to convert this benchmark to nvbench. The reason not to allow something like #11209 is to encourage us to switch (which is also important for adjacent initiatives like building dashboards around better benchmark tracking).
bdice
requested changes
Jul 6, 2022
bdice
reviewed
Jul 6, 2022
karthikeyann
reviewed
Jul 7, 2022
karthikeyann
approved these changes
Jul 7, 2022
bdice
reviewed
Jul 8, 2022
bdice
approved these changes
Jul 8, 2022
Contributor
bdice
left a comment
bdice
left a comment
There was a problem hiding this comment.
One suggestion, otherwise LGTM!
Contributor
Author
|
@gpucibot merge |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
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.
The current implementation of
cudf::contains(column_view, scalar)usesthrust::findandthrust::any_of(which also callsthrust::find_ifunder the hood). These thrust APIs were known to have performance regression (NVIDIA/cccl#720).This PR replaces
thrust::findandthrust::any_ofincudf::containsbythrust::count_if, which improves performance significantly.Benchmarks show that the run time can be reduced as much as 80% after modification, or up to 5X speedup.
Closes #3806.