Skip to content

Parallelize Pairwise Duplicate Comparison #5

Description

@tim-dickey

Goal

Improve scalability by parallelizing pairwise similarity comparisons for large file sets while maintaining deterministic output ordering.

Scope

  • Strategy: partition signatures into blocks/chunks; compute pairwise within + across blocks using process pool.
  • Deterministic ordering: gather results then sort globally before output.
  • Config flag: --compare-workers (distinct from signature --workers).
  • Heuristic: enable automatically if N > threshold (e.g., >1500 files) unless overridden.
  • Measure overhead vs benefit; update benchmarking to include comparison phase metrics.

Acceptance Criteria

  • Parallel mode yields identical result set to serial (order after final sort).
  • Throughput improvement documented for large synthetic dataset (benchmarks updated).
  • Tests ensure parity (serial vs parallel) and stable ordering.

Non-Goals

  • Implement LSH pruning (separate future issue).

Future

  • Adaptive threshold based on average shingle set size.

Activity

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions