jpountz commented on PR #11860:
URL: https://github.com/apache/lucene/pull/11860#issuecomment-1382689973
For reference, there seems to be a 6-7% QPS drop on nightly benchmarks
associated with this change.
https://people.apache.org/~mikemccand/lucenebench/VectorSearch.html I think
it's fine
jpountz commented on PR #12079:
URL: https://github.com/apache/lucene/pull/12079#issuecomment-1382690674
The last data point at
https://people.apache.org/~mikemccand/lucenebench/sparseResults.html#tot_merge_times
has a drop for overall merging that I expect to be mostly contributed by this
benwtrent commented on PR #11860:
URL: https://github.com/apache/lucene/pull/11860#issuecomment-1382728572
This for sure has to do with reading for the memory offsets and then reading
the neighbors.
I can dig into this a little bit next week unless somebody else has a really
good id
rmuir closed issue #11381: remove non-NRT replication support [LUCENE-10345]
URL: https://github.com/apache/lucene/issues/11381
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
rmuir merged PR #12038:
URL: https://github.com/apache/lucene/pull/12038
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apach
rmuir merged PR #12056:
URL: https://github.com/apache/lucene/pull/12056
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apach
rmuir opened a new pull request, #12086:
URL: https://github.com/apache/lucene/pull/12086
Went thru the new checks as usual. Now that `Finalize` has our bugfix, I
enabled it.
Closes #12057
--
This is an automated message from the Apache Git Service.
To respond to the message, pleas
rmuir closed issue #12057: ban finalizers in the build somehow (worst-case: use
error-prone)
URL: https://github.com/apache/lucene/issues/12057
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the sp
rmuir merged PR #12086:
URL: https://github.com/apache/lucene/pull/12086
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apach
rmuir opened a new pull request, #12087:
URL: https://github.com/apache/lucene/pull/12087
Clean up this query a bit, and move it around to support:
* NumericDocValuesField.newSlowSetQuery()
* SortedNumericDocValuesField.newSlowSetQuery()
This complements the existing docvalue
rmuir commented on issue #12028:
URL: https://github.com/apache/lucene/issues/12028#issuecomment-1382953573
I don't think it is good to degrade to `BooleanQuery` when using points or
doc-values, it will only hurt performance.
Let's add `NumericDocValuesField.newSlowSetQuery()` and
`S
rmuir commented on PR #12087:
URL: https://github.com/apache/lucene/pull/12087#issuecomment-1382954253
intended as followups:
* look into PointRangeQuery and implement necessary estimation for
IndexOrDocValuesQuery to do the right thing
* Add newSetQuery() to IntField/LongField/DoubleF
rmuir commented on PR #12087:
URL: https://github.com/apache/lucene/pull/12087#issuecomment-1383064386
Here's my benchmarks with attached java program:
[NumSetBenchmark.java.txt](https://github.com/apache/lucene/files/10419558/NumSetBenchmark.java.txt)
* `main` uses `IntPoint.newSetQuery`
rmuir commented on PR #12087:
URL: https://github.com/apache/lucene/pull/12087#issuecomment-1383064967
the benchmark above uses queries such as `"la|21,22,23",// 2226 hits`
in this case we form a boolean query of TermQuery:"la" AND admin2code in
(21,22,23). The admin2 codes are ty
14 matches
Mail list logo