Repository navigation
Parallelize bounded CPU FPFH descriptor phases - #145
Merged
Merged
Conversation
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
FPFH computed every SPFH row and weighted descriptor on the caller despite independent point work in each phase. Share one caller-built KD-tree across bounded scoped CPU workers, writing disjoint contiguous slices directly into the existing descriptor arrays. Join all SPFH workers before weighting reads their rows. Reuse existing point staging; small/one-worker clouds stay serial.
Each point keeps the unsorted radius-neighbor order, Darboux/bin/normalization math and 33-bin f32 accumulation order. Matching, seeded RANSAC and validation retain their existing execution. Public signatures and dependencies are unchanged. Scratch remains neighborhood-dependent per worker; concurrent calls have their own budgets.
Validation:
Implementation and full measurement note.
Based on merged PR #144 (main f83ed25). Merge is gated on the frozen source head, all 114 named checks succeeding, no pending/failed checks and a clean merge state.