Learn heuristic to pick fastest select_k algorithm #1523
Merged
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
This uses the select_k dataset from rapidsai#1497 to learn a heuristic of the fastest select_k variant based off the rows/ cols/k of the input. This heuristic is modelled as a DecisionTree, which is automatically exported in C++ code that is compiled into RAFT. This lets us learn a function to pick the fastest select_k method - which requires only a few if statements in C++ code, making it very cheap to evaluate.
benfred
added
improvement
cjnolet
approved these changes
May 17, 2023
| * on different values of rows/cols/k. The decision tree is converted to c++ | ||
| * code, which is cut and paste below. | ||
| * | ||
| * The code to generate is in cpp/scripts/heuristics/select_k, running the |
There was a problem hiding this comment.
Just a tiny nitpick:
Suggested change
| * The code to generate is in cpp/scripts/heuristics/select_k, running the | |
| * NOTE: The code to generate is in cpp/scripts/heuristics/select_k, running the |
|
/merge |
achirkin
reviewed
May 18, 2023
| */ | ||
| inline Algo choose_select_k_algorithm(size_t rows, size_t cols, int k) | ||
| { | ||
| if (k > 134) { |
There was a problem hiding this comment.
I think, we'd better use log_2(k) instead of k when constructing the heuristic, so that all values of k go in powers of two. For all warp-based algorithms, performance for non-powers of two is equal to their rounded-up powers of two (queue capacity parameter).
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Outdated suggestions cannot be applied.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.
This uses the select_k dataset from #1497 to learn a heuristic of the fastest select_k variant based off the rows/ cols/k of the input. This heuristic is modelled as a DecisionTree, which is automatically exported in C++ code that is compiled into RAFT. This lets us learn a function to pick the fastest select_k method - which requires only a few if statements in C++ code, making it very cheap to evaluate.