Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Use algorithm benchmarks to improve performance #1224

Closed
maxkfranz opened this issue Jan 20, 2016 · 1 comment
Closed

Use algorithm benchmarks to improve performance #1224

maxkfranz opened this issue Jan 20, 2016 · 1 comment
Labels
performance Improving speed and memory usage priority-2-medium
Milestone
3.3.0

Comments

@maxkfranz
Copy link
Member

Ref : #936
@maxkfranz maxkfranz added this to the 2.7.0 milestone Jan 20, 2016
@maxkfranz maxkfranz added the performance Improving speed and memory usage label Jan 20, 2016
@maxkfranz maxkfranz modified the milestones: future, 3.2.0 Dec 7, 2016
@d2fong d2fong modified the milestones: 3.2.0, 3.3.0 Jun 7, 2017
@maxkfranz maxkfranz mentioned this issue Mar 5, 2018
Closed
maxkfranz added a commit that referenced this issue Jul 4, 2018
@maxkfranz
Refactor Bellman-Ford implementation
075db77 
Ref:
- Use algorithm benchmarks to improve performance #1224
- Bellman-Ford returning negative weight cycles #2123
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Improve performance of A*
d70acfc 
- Replace `var` with `let`
- Use a heap for the node queue
- Cheaper check for in open set & helper functions
- Reduce id() calls and use a map for faster closed set lookup
- Build resultant path without recursion
- Improves performance by ~2.1x on GAL

Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Clean up betweenness centrality
4641478 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Simplify root resolution in degree centrality
931f1e8 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Clean up closeness centrality
665714e 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Use eles.byGroup() and unmerge the loops to reduce iterations and n…
2cfb5f2 
…umber of intermediary collections in Bellman-Ford

Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Tweaks to Page Rank
47e77e7 
- Replace `var` with `let`
- Use `eles.byGroup()` to avoid double iteration for getting nodes and edges lists
- Avoid redundant ID calls
- Avoid `filter()` call by skipping loops inline
- Use multiplication instead of `Math.pow()` for squaring
- Improvement of ~1.07x

Ref #1224
maxkfranz added a commit that referenced this issue Jul 24, 2018
@maxkfranz
Update BFS/DFS with small optimisations
bf0a51b 
- Change `var` to `let`
- Avoid double iteration to create separate nodes and edges lists -- re-usable `eles.byGroup()`
- Reduce number of `filter()` calls
- Reduce number of `ele.id()` calls
- Use an in-place merge to construct the path rather than an intermediate array
- Improves performance by ~1.3x on GAL

Ref #1224
maxkfranz referenced this issue Jul 24, 2018
@maxkfranz
Clean up A* : Lint & use util.defaults()
b742225
maxkfranz added a commit that referenced this issue Jul 25, 2018
@maxkfranz
Clean up eles.dijkstra()
8719c4b 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 26, 2018
@maxkfranz
Add internal eles.indexOf(ele) function
1be861d 
- Gets index of element in collection
- Useful for algorithms that use matrices

Ref #1224
maxkfranz added a commit that referenced this issue Jul 26, 2018
@maxkfranz
Clean up Floyd-Warshall
a45a219 
- Use `util.defaults()`
- Change `var` to `let`
- Reduce overall code length -- less ~50 lines
- Remove `cy.getElementById()` calls
- Use `eles.merge()` to build resultant path rather than an intermediate array -- less memory usage
- No significant change in performance

Ref #1224
maxkfranz added a commit that referenced this issue Jul 26, 2018
@maxkfranz
Use a 1-dimensional array for Floyd-Warshall rather than multi-dimens…
bbca3a5 
…ional arrays

- Reduces array allocations
- Array allocations specify fixed N*N size
- Increases performance on GAL benchmark by ~1.8x

Ref #1224
maxkfranz added a commit that referenced this issue Jul 26, 2018
@maxkfranz
Use util.defaults() in Bellman-Ford
83a59ff 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 27, 2018
@maxkfranz
Add internal eles.indexOfId(id) function for where the ID is alread…
1861b55 
…y read, for performance

Ref #1224
maxkfranz added a commit that referenced this issue Jul 27, 2018
@maxkfranz
Use eles.spawn() over cy.collection() in internal `eles.byGroup()…
727a926 
…` for performance

Ref #1224
maxkfranz added a commit that referenced this issue Jul 27, 2018
@maxkfranz
Add vector normalisation functions to math
01411de 
Ref #1224
maxkfranz added a commit that referenced this issue Jul 27, 2018
@maxkfranz
Optimisations for Page Rank
8c39abb 
- Use `util.defaults()`
- Use general normalisation function in `math`
- Use single dimensional array for N*N matrix
- Declare array sizes on init and do not change array lengths
- Re-use arrays that represent the vectors of the main loop of the alg. -- better memory usage & less GC
- Improves benchmark on GAL graph by ~1.5x

Ref #1224
maxkfranz added a commit that referenced this issue Jul 27, 2018
@maxkfranz
Optimisations for Kruskal
fbb2699 
- Change `var` to `let`
- `anySame()` changes to `has()`
- `findSet()` only handles indices rather than creating new objects
- Use `merge()` to modify collections in-place rather than creating many collections

Red #1224
maxkfranz referenced this issue Jul 30, 2018
@maxkfranz
Optimisations & improvements for Karger-Stein
f97ba5f 
- Add an error message if the algorithm is run on a graph containing multiple components.
- Minimise array creations and use fixed-size allocations where possible.
- Modify returned collections in-place rather than creating intermediate arrays.
- Fix misnamed file (ES modules refactoring)
- Benchmark for Karger-Stein on GAL ~2.3x faster than v3.2.
- Clean up benchmarks.
maxkfranz referenced this issue Jul 30, 2018
@maxkfranz
Clean up unmerge filtering in algorithms
560019b 
- Add `eles.unmergeAt()` to remove at a particular index
- Add `eles.unmergeBy()` to remove by a particular condition
- Misc. cleanup
@maxkfranz
Copy link
Member Author

Here are the latest benchmarks run on the GAL-Filtered graph, with the code as of today.

Algorithm : comparison multiplier of v3.3 speed / v3.2 speed

  • A* : ~3.3x
  • Bellman-Ford : ~1.3x
  • Betweenness centrality : ~1.2x
  • BFS/DFS : ~1.8x
  • Closeness centrality : ~1.5x
  • Degree centrality : ~1.4x
  • Dijkstra : ~2.0x
  • Floyd-Warshall : ~1.75x
  • Karger-Stein : ~2.5x
  • Kruskal : ~2.8x
  • Page Rank : ~1.5x

Computer specs:

MacBook Pro (Retina, 13-inch, Early 2015)
Mac OS 10.12.6
3.1 GHz Intel Core i7
16 GB 1867 MHz DDR3

May be of interest to : @gbader @jvwong @d2fong

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
performance Improving speed and memory usage priority-2-medium
Projects
None yet
Milestone
3.3.0
Development

No branches or pull requests
2 participants
@maxkfranz @d2fong