perf!: Reduce Seedfinder heap allocation #905
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
Codecov Report
@@ Coverage Diff @@
## main #905 +/- ##
=======================================
Coverage 48.66% 48.66%
=======================================
Files 331 331
Lines 17129 17131 +2
Branches 8094 8092 -2
=======================================
+ Hits 8336 8337 +1
Misses 3087 3087
- Partials 5706 5707 +1
Continue to review full report at Codecov.
|
| namespace ActsExamples { | ||
|
|
||
| /// A proto track is a collection of hits identified by their indices. | ||
| using ProtoTrack = std::vector<Index>; | ||
| using ProtoTrack = boost::container::small_vector<Index, 10>; |
There was a problem hiding this comment.
this uses 3x more space than necessary for 3 SP, right?
There was a problem hiding this comment.
Yeah true, I just went with a low-ish number. Let's see what I get with 3.
There was a problem hiding this comment.
In my testing, reducing to 3 appears to actually slow things down by about 3%, which I can't explain. I suggest we keep it at 10.
Added. |
|
Clang seems to dislike the template template. |
|
Should compile now. |
robertlangenberg
approved these changes
Aug 10, 2021
stephenswat
added a commit
to stephenswat/acts
that referenced
this pull request
Aug 19, 2021
This is just a tiny little improvement on acts-project#905. I noticed that we are currently using an output iterator function to write some results to a vector, and then copying that vector over to append it to another, existing vector. I think it would be (very marginally) quicker and neater to avoid this extra copy and just have the function take the insertion iterator of the vector we are planning to write to in the end.
stephenswat
added a commit
to stephenswat/acts
that referenced
this pull request
Aug 19, 2021
This is just a tiny little improvement on acts-project#905. I noticed that we are currently using an output iterator function to write some results to a vector, and then copying that vector over to append it to another, existing vector. I think it would be (very marginally) quicker and neater to avoid this extra copy and just have the function take the insertion iterator of the vector we are planning to write to in the end.
paulgessinger
pushed a commit
that referenced
this pull request
Aug 20, 2021
This is just a tiny little improvement on #905. I noticed that we are currently using an output iterator function to write some results to a vector, and then copying that vector over to append it to another, existing vector. I think it would be (very marginally) quicker and neater to avoid this extra copy and just have the function take the insertion iterator of the vector we are planning to write to in the end.
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.
The Seed finder does a large number of heap allocations for various different things. With this PR, I'm reducing the number of heap allocations considerably with little change to the public interface.
On my machine, the runtime of the
SeedingAlgorithmon 10000 events with the generic detector drops from about 3s to 2s, for about a 33% speedup.The number of heap allocations is reduced from 220 million to just over 300000 so about a reduction of >99%.
I checked that the estimated parameter ROOT files are identical by using our
compareRootFiles.Cscript.BREAKING CHANGE: The interface of a number of methods changes:
SeedFilter::filterSeeds_2SpFixed(takes an output iterator, is now void)SeedFilter::filterSeeds_1SpFixed(takes an output iterator, is now void)Seedfinder::createSeedsForGroup(takes an output iterator, is now void, takes a state object as a reference)Grid::collect(returnsboost::container::small_vector<size_t, N>whereN = 3^DIMof the grid)