Thanks to visit codestin.com
Credit goes to Github.com

Skip to content

feat: optimize SVD ?#167

Draft
lpatiny wants to merge 1 commit intomainfrom
speed
Draft

feat: optimize SVD ?#167
lpatiny wants to merge 1 commit intomainfrom
speed

Conversation

@lpatiny
Copy link
Member

@lpatiny lpatiny commented Sep 29, 2023 •
edited
Loading

The order of processing loops may be critical in order to optimize speed.

The implemented algorithms EVD, SVD, etc may assume a specific row / col or col / row order. The goal being that in the inner loops the memory is accessed sequentially.

By simply inverting the get / set in matrix (just using square matrices not to check everything)
we observe major time difference (in ms)

The best improvement was observe for a SVD of 1000x1000 elements (x3 !) but the inversion
can also lead to worse results like for LuDecomposition (/2 ...).

Inverted

┌─────────┬──────┬────────────────────────────┬──────────────────────┬─────────────────────────┐
│ (index) │ size │ SingularValueDecomposition │   LuDecomposition    │ EigenvalueDecomposition │
├─────────┼──────┼────────────────────────────┼──────────────────────┼─────────────────────────┤
│    0    │  10  │    0.047006314899820094    │ 0.006285565016191445 │   0.04497034566020066   │
│    1    │ 100  │     9.556330035184766      │  0.5338207964597829  │    11.88394686695906    │
│    2    │ 500  │     964.0846458325783      │  57.523048850654185  │   1427.5238334983587    │
│    3    │ 1000 │     7546.173541992903      │  563.8283982227246   │   15070.775082990527    │
│    4    │ 1500 │     25270.795874997973     │  2161.1955276678004  │    39320.5441249907     │
└─────────┴──────┴────────────────────────────┴──────────────────────┴─────────────────────────┘

Non inverted

┌─────────┬──────┬────────────────────────────┬───────────────────────┬─────────────────────────┐
│ (index) │ size │ SingularValueDecomposition │    LuDecomposition    │ EigenvalueDecomposition │
├─────────┼──────┼────────────────────────────┼───────────────────────┼─────────────────────────┤
│    0    │  10  │    0.028381821972091822    │ 0.0047396303146609135 │  0.029047660718630176   │
│    1    │ 100  │     7.579296062081497      │  0.39713138298076744  │   11.749581896539393    │
│    2    │ 500  │     1620.0299582481384     │  38.316282882820815   │   1338.5159477517009    │
│    3    │ 1000 │     22116.20337499678      │   307.1329264702166   │   19482.321583002806    │
│    4    │ 1500 │     44339.38079200685      │  1027.2492165982724   │    41522.77608400583    │
└─────────┴──────┴────────────────────────────┴───────────────────────┴─────────────────────────┘

This is just at the stage of reflection that could lead to some speed improvement.

@codecov
Copy link

codecov bot commented Sep 29, 2023 •
edited
Loading

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.28%. Comparing base (bde1c9a) to head (e82b2ba).
⚠️ Report is 19 commits behind head on main.

Additional details and impacted files 
@@            Coverage Diff             @@
##             main     #167      +/-   ##
==========================================
+ Coverage   71.27%   71.28%   +0.01%     
==========================================
  Files          34       34              
  Lines        5295     5297       +2     
  Branches      849      849              
==========================================
+ Hits         3774     3776       +2     
  Misses       1516     1516              
  Partials        5        5

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@lpatiny