Skip to content
GitLab
Projects Groups Snippets
  • /
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Register
  • Sign in
  • S Slam
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
  • Issues 6
    • Issues 6
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 12
    • Merge requests 12
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Environments
    • Releases
  • Packages and registries
    • Packages and registries
    • Container Registry
  • Monitor
    • Monitor
    • Incidents
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Wiki
    • Wiki
  • Snippets
    • Snippets
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • KEU-ComputerVision
  • Slam
  • Merge requests
  • !19

Add multi-threading

  • Review changes

  • Download
  • Email patches
  • Plain diff
Merged Nicolas Cadart requested to merge feat/multihreading into master Feb 24, 2020
  • Overview 17
  • Commits 6
  • Pipelines 5
  • Changes 11

Enable optional multi-threading for SLAM using OpenMP (Ceres multi-threading is also based on OpenMP).

Add OpenMP dependency.

Some steps can be performed in different threads to increase runtime performance:

  • Keypoints extraction can be efficiently parallelized by running independent extraction on each scanLine in different threads.
  • SLAM ICP for-loops on each keypoint can also be parallelized.
  • SLAM Ceres optimization can also be parallized by computing residuals and jacobians on different threads.
  • Keypoints maps update, maps loading or KD-trees constructions can be performed in independent sections.

Tests achieved a speedup of approximately x2 using at most 4 threads, even if these 4 threads are not requested 100% of the time.

Assignee
Assign to
Reviewers
Request review from
Time tracking
Source branch: feat/multihreading