Only trigger TopN rewrite relatively small limits compared to the table size. #17140
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
|
Thanks! |
|
I ran the test again and got the following plot. It looks like we turn off topN maybe a bit too soon? You can see for 5M, 10M and 20M the "with optimization" performance jumps straight to "without optimization". I think we can squeeze a bit more performance out in those cases |
krlmlr
added a commit
to duckdb/duckdb-r
that referenced
this pull request
May 18, 2025
Only trigger TopN rewrite relatively small limits compared to the table size. (duckdb/duckdb#17140)
krlmlr
added a commit
to duckdb/duckdb-r
that referenced
this pull request
May 18, 2025
Only trigger TopN rewrite relatively small limits compared to the table size. (duckdb/duckdb#17140)
krlmlr
added a commit
to duckdb/duckdb-r
that referenced
this pull request
May 19, 2025
Only trigger TopN rewrite relatively small limits compared to the table size. (duckdb/duckdb#17140)
Mytherin
added a commit
that referenced
this pull request
May 28, 2025
Fixes a bug introduced by #17140 The computation was wrong because it did not check `has_estimated_cardinality` before grabbing it. This PR computes the cardinality if it was not yet there. These fields should probably be changed to `optional_idx` in the future so this doesn't happen.
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Outdated suggestions cannot be applied.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.
Fixes https://github.com/duckdblabs/duckdb-internal/issues/4363
Fixes #16527
TopN isn't always faster. The below experiments show that as the table cardinality increases/as the limit increases, TopN performance ends up suffering from nLog(n) operations performed one tuple at a time, and sorting the whole table becomes faster. The plots show without optimization (i.e sorting the whole table) as having relatively constant performance, while TopN performance degrades as the limit increases.
Looking at the plots, we can see that sorting the table is constant regardless of the limit. However, once the limit value is >= (0.007)*(table_cardinality), the topn optimization takes more time that a normal sort.
The same plot with the changes from this PR that disables the TopN optimization once the limit value is >= 0.007 * table_cardinality (and the limit is > 5000).
With this PR, we now observe the following runtimes
CC @lnkuiper (I think you still have a patch you want to add?)