[stdlib] Optimize some Set operations #21300
Closed
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
|
@swift-ci benchmark |
|
@swift-ci test |
|
Build failed |
Build comment file:Performance: -O
Code size: -O
Performance: -Osize
Code size: -Osize
Performance: -Onone
How to read the dataThe tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.If you see any unexpected regressions, you should consider fixing the Noise: Sometimes the performance results (not code size!) contain false Hardware Overview |
|
The benchmark suit doesn't cover all these operations. Here are some highlights from local testing:
|
|
Similar charts for operations returning sets:
|
|
@swift-ci please test |
This comment has been minimized.
This comment has been minimized.
|
Build failed |
jrose-apple
reviewed
Dec 13, 2018
|
Swift(linux-x86_64).SIL.parse_stdlib.sil (from Swift(linux-x86_64)): Hm... 🤨 |
|
@swift-ci please test linux platform |
|
Build failed |
lorentey
force-pushed
the
set-on-fire
branch
from
76f1ac5
to
64f62f7
Compare
|
@swift-ci please test |
|
@swift-ci benchmark |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Build comment file:Performance: -O
Code size: -O
Performance: -Osize
Code size: -Osize
Performance: -Onone
How to read the dataThe tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.If you see any unexpected regressions, you should consider fixing the Noise: Sometimes the performance results (not code size!) contain false Hardware Overview |
|
Build failed |
Use a temporary bitset to speed up the Sequence variant by roughly a factor of 3.
Use a temporary bitset to speed up the Sequence variant by roughly a factor of 3.
In the Sequence variant, don’t convert the argument to a Set — just check if self contains all elements. This makes isSuperset(of:) roughly 4 times faster.
- Use a temporary bitset to speed up the Sequence variant by roughly a factor of 4. - Fix a logic error causing the a == b case for the set variant to be O(n) instead of O(1).
- In the sequence case, don’t convert the input to a set; simply test its elements one by one. This leads to a cool 4x speedup. - In the set case, prefer to iterate over the smaller set. The leads to a speedup proportional to the ratio of the counts between the two sets if the caller used an unlucky ordering.
Use a temporary bitset to avoid hashing elements more than once, and to prevent rehashings during the creation of the result set. This leads to a speedup of about 0-4x, depending on the number of elements removed.
This works the same way as Set.subtracting<S>(_:), and has similar performance benefits.
Use a temporary bitset to speed up the Sequence variant by roughly a factor of ~4-6, and the set-set variant by a factor of ~1-4, depending on the ratio of overlapping elements.
This was referenced May 7, 2019
|
@swift-ci please benchmark |
Performance: -O
Code size: -O
Performance: -Osize
Code size: -Osize
Performance: -Onone
How to read the dataThe tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.If you see any unexpected regressions, you should consider fixing the Noise: Sometimes the performance results (not code size!) contain false Hardware Overview |
|
@swift-ci please test |
|
Build failed |
Performance: -O
Code size: -O
Performance: -Osize
Code size: -Osize
Performance: -Onone
How to read the dataThe tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.If you see any unexpected regressions, you should consider fixing the Noise: Sometimes the performance results (not code size!) contain false Hardware Overview |
|
The runtimes of the extended Similarly, there is a small issue with too low runtime (even before these changes) in a case of |
|
Closing in favor of #40012. |
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 temporary bitmaps to considerably speed up some high-level Set operations.
Set.isSubset<S>(of:)Set.isStrictSubset<S>(of:)Set.isSuperset<S>(of:)Set.isStrictSuperset<S>(of:)Set.isDisjoint<S>(with:)Set.subtracting(_:)Set.filter(_:)Set.intersection<S>(_:)Set.intersection(_:)Benchmark results depend on the size of the sets and the element type. The numbers above were measured with integer keys across a range of counts from 1 to 1 million items.