Issues

• Move repo to the ASF

  mikemccand/luceneutil#624 · Filter by author dsmiley opened on Oct 9, 2026

• Add simple native dot-product code

  mikemccand/luceneutil#620 · Filter by author mikemccand opened on Oct 5, 2026

• Body field has been indexed TWICE per document since Nov 2021

  mikemccand/luceneutil#616 · Filter by author neoremind opened on Sep 4, 2026
  7

• Add a mode to compute the recall ceiling due to quantization alone

  mikemccand/luceneutil#609 · Filter by author xande opened on Aug 13, 2026
  1

• Regenerate main wiki benchmark corpus from Cohere v3 export?

  mikemccand/luceneutil#605 · Filter by author mikemccand opened on Aug 3, 2026

• Can/should we add each JVM iteration to the nightly charts

  mikemccand/luceneutil#598 · Filter by author mikemccand opened on Jul 20, 2026
  2

• Split wikinightly.tasks to manage growing nightly runtime

  mikemccand/luceneutil#594 · Filter by author sgup432 opened on Jul 17, 2026

• Ensure knnPerfTest.py uses new query vectors for warmup vs real test

  mikemccand/luceneutil#591 · Filter by author mikemccand opened on Jul 7, 2026
  1

• Add utility to measure vector search CPU cost sweeping across varying inputs like topK, vector count, dimension, segment count while preserving recall

  mikemccand/luceneutil#590 · Filter by author mikemccand opened on Jun 30, 2026
  1

• Upgrade to Cohere v4 Matryoshka Wikipedia embeddings?

  mikemccand/luceneutil#584 · Filter by author mikemccand opened on Jun 8, 2026
  1

• Add "partitipation ratio" measure (maybe just "condition") for vectors so we can see how helpful preconditioning would be

  mikemccand/luceneutil#580 · Filter by author mikemccand opened on May 11, 2026
  1

• Why is runNightlyKnn.py using inefficient non-optimized non-bulk Panama APIs?

  mikemccand/luceneutil#579 · Filter by author mikemccand opened on May 10, 2026

• Implement NativeVectorizationProvider using BLAS library

  mikemccand/luceneutil#573 · Filter by author mikemccand opened on May 1, 2026

• Should nightly benchmarks swap in a performant (native JMA) NativeVectorizationProvider?

  mikemccand/luceneutil#572 · Filter by author mikemccand opened on May 1, 2026
  1

• Understand hot CPUTime methods from nightly vectors index

  mikemccand/luceneutil#571 · Filter by author mikemccand opened on May 1, 2026
  1

• Record more JFR events like C2 compile / decompile, gc/heap, etc.

  mikemccand/luceneutil#570 · Filter by author mikemccand opened on May 1, 2026

• Expose SIMD CPU counters over time (-I interval option to perf)

  mikemccand/luceneutil#569 · Filter by author mikemccand opened on Apr 24, 2026

• Enable measuring recall loss due solely to quantization

  mikemccand/luceneutil#567 · Filter by author mikemccand opened on Apr 17, 2026
  2

• Expose re-ranking in knnPerfTest.py

  mikemccand/luceneutil#565 · Filter by author mikemccand opened on Apr 13, 2026

• Improve "make tidy" to also fixup the java sources?

  mikemccand/luceneutil#564 · Filter by author mikemccand opened on Apr 10, 2026
  2

• Add score-weighted recall metric

  mikemccand/luceneutil#563 · Filter by author mikemccand opened on Apr 10, 2026

• CPU JFR profiler is missing nearly everything?

  mikemccand/luceneutil#562 · Filter by author mikemccand opened on Apr 2, 2026
  3

• Can we measure / aggregate something about the HNSW graph during indexing to infer its "quality"?

  mikemccand/luceneutil#561 · Filter by author mikemccand opened on Mar 31, 2026
  1

• Can we measure correlation of a filter and HNSW connectedness?

  mikemccand/luceneutil#560 · Filter by author mikemccand opened on Mar 30, 2026

• Add HNSW graph profiling

  mikemccand/luceneutil#559 · Filter by author mikemccand opened on Mar 30, 2026