Lucene: Status=Open

[Save this search]

Author relation

Updated ago

Comment count

Reaction count

Review Requested

User

Last comment user

Type

Filters: Status (Open), Issue type, Author relation, Created, Updated, Updated ago, Comment count, Reaction count, Review Requested, Mentioned, Reviewed, Commented, User, Last comment user, Draft, Component, Type, Labels, Commits?, Reporter, Assignee

#16691 PR: Optimize DocValues bulk decoding using Panama widening casts (#16677)

2.1 hours ago 9 comments 0 votes 0 watches CH-Abhinav, github-actions[bot], rmuir

Description This PR introduces SIMD bulk decoding for DocValues (8-bit, 16-bit, and 32-bit ... Currently, default scalar loops handle these widening. This PR tries to use Panama Vector API's shape-widening operations (convertShape with ...

github-actions[bot] 3.6 days ago:	This PR has not had activity in the past 2 weeks, labeling it as stale. If the PR is waiting for review, notify the dev@lucene.apache.org list. Thank you for your contribution!
CH-Abhinav 19 days ago:	Can anyone approve changes or suggest some feedback on this PR please 🥺?

#16889 PR: ConcurrentMergeScheduler can spawn more merge threads than maxMergeCount

2.3 hours ago 15 comments 0 votes 0 watches github-actions[bot], goingforstudying-ctrl, vsop-479

DescriptionI was looking at #13593 and the race aoli-al described is still there in main: ... So a merge can become pending right between the stall check and getNextMerge(), and merge() happily ... This makes the cap airtight in two spots: - A merge thread asked to schedule more work while we're ...

goingforstudying-ctrl 2.3 hours ago:	Done in 6c8b1c5. maybeStall now returns false inside the cap loop when the merge queue is empty, ... The flushing thread no longer stalls while merge threads sit on the latch.
vsop-479 12.8 hours ago:	Producers should also return false when the maxMergeCount is reached and there are no pending ... There is no work to schedule, so we should stop scheduling rather than stall the producer. My suggestion was to keep both checks inside the loop: ` while (mergeThreadCount() >= ...

#16790 PR: Vectorize 3 loops in OptimizedScalarQuantizer

2.6 hours ago 55 comments 0 votes 0 watches github-actions[bot], kaivalnp, slow-J

I was looking into HNSW indexing performance (also https://github.com/apache/lucene/pull/16631 ) ... OptimizedScalarQuantizer.scalarQuantize centers the vector, then runs a coordinate descent to pick ... Each descent iteration walks the whole vector twice, once for the loss and once for the next step, ...

slow-J 2.8 hours ago:	Good catch, I'll remove it!
slow-J 2.8 hours ago:	Sure, I'll use reduceLanes then!

#16896 PR: fix long overflow in GCD-encoded DocValues range transform

4 hours ago 5 comments 0 votes 0 watches costin, github-actions[bot], kkewwei

Description In NumericDocValues, GCD encoding stores(v - min) / gcd as an unsigned long, andmin ... When querying the range[-1, 1] ,saturatingShiftLower andsaturatingShiftUpper computev - delta . Sincedelta < 0 , the computation overflows, causing the encoded range to become[Long.MAX_VALUE, ...

kkewwei 6.1 days ago:	@costin Would you mind take another look in your spare time?
kkewwei 12.1 days ago:	@costin Very thank you for the detailed review and insightful feedback. I’ve refined the boundary checks to allow some cases withbitsPerValue == 64 to safely ... Please see the comments for details.

#16835 PR: Speed up rangeIntoBitSet for table-encoded numeric and sorted numeric DocValues

4 hours ago 18 comments 0 votes 0 watches HoustonPutman, github-actions[bot], kkewwei, romseygeek, sgup432

Description Inspired by the gcd/min and transformTableBounds, this change usestransformTableBounds ... Numeric and sorted numeric DocValues can then evaluaterangeIntoBitSet directly on packed ordinals, ... .

kkewwei 5.1 days ago:	@HoustonPutman Would you mind take another look in your spare time? Very thank you.
kkewwei 14.7 days ago:	Yes, it was confusing to read. Added a comment to clarify it.

#17006 PR: Count soft deletes that aren't written yet in IndexWriter's doc stats

7.1 hours ago 1 comments 0 votes 0 watches diwu-sf, github-actions[bot], shbhar

DescriptionWith soft deletes, IndexWriter#getDocStats().numDocs and IndexWriterhasDeletions() can ... A soft delete or soft update resolved by an auto flush isn't counted. numDocs is too high, and ... A hard delete in the same spot is counted. 2.

shbhar 3.4 days ago:

Thanks for creating this bug fix PR @diwu-sf, in case the repro, diagrams, tests or suggested fixes ...

#16996 PR: Fix soft delete counts of a writer opened from an NRT reader

7.1 hours ago 0 comments 0 votes 0 watches diwu-sf, github-actions[bot]

Description A writer can be opened on an NRT reader's commit with IndexWriterConfigsetIndexCommit( ... With soft deletes, the new writer gets its delete counts wrong. With assertions on, opening it fails with softDeleteCount doesn't match 1 !

«
1
2
3
4
5
»

2,734 issues

[23.0 msec search, 24.0 msec total]