gsmiller commented on PR #13657:
URL: https://github.com/apache/lucene/pull/13657#issuecomment-2290359914
As one more thought, all I can think is that these impacted tasks are
possibly taking the `leafSlices.length == 0` condition branch and avoiding the
need to create the collectors list a
gsmiller commented on PR #13657:
URL: https://github.com/apache/lucene/pull/13657#issuecomment-2290353375
I'll also mention that I experimented with inlining the private search
method to avoid having to pass leafSlices and collectors as arguments to the
private method and I see the same eff
gsmiller opened a new pull request, #13657:
URL: https://github.com/apache/lucene/pull/13657
After merging #13568 @mikemccand and @jpountz
[noticed](https://github.com/apache/lucene/pull/13568#issuecomment-2288666763)
some surprising nightly benchmark performance regressions. @epotyom and I
github-actions[bot] commented on PR #13609:
URL: https://github.com/apache/lucene/pull/13609#issuecomment-2290122158
This PR has not had activity in the past 2 weeks, labeling it as stale. If
the PR is waiting for review, notify the d...@lucene.apache.org list. Thank you
for your contributi
gsmiller merged PR #13644:
URL: https://github.com/apache/lucene/pull/13644
--
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.ap
uschindler merged PR #13649:
URL: https://github.com/apache/lucene/pull/13649
--
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.
uschindler closed issue #13638: `gradlew eclipse` no longer works
URL: https://github.com/apache/lucene/issues/13638
--
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 unsubscr
uschindler commented on PR #13649:
URL: https://github.com/apache/lucene/pull/13649#issuecomment-2289976761
Works all fine, merging now. Thanks!
--
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 th
uschindler commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1717587581
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ d
navneet1v commented on PR #13469:
URL: https://github.com/apache/lucene/pull/13469#issuecomment-2289797047
@msokolov and @benwtrent I see that we exposed the FlatVectorsFormat but I
am not seeing entry for this format in this file:
https://github.com/apache/lucene/blob/main/lucene/core/src/
rmuir commented on PR #13572:
URL: https://github.com/apache/lucene/pull/13572#issuecomment-2289745259
sorry @goankur i'm super-behind, vacations and end of summer. i definitely
want to try out the VNNI path but just haven't found the cycles yet.
--
This is an automated message from the A
msokolov merged PR #13655:
URL: https://github.com/apache/lucene/pull/13655
--
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.ap
msokolov commented on PR #13655:
URL: https://github.com/apache/lucene/pull/13655#issuecomment-2289624603
> I think this is sane. Though it is used by incremental merger still, which
is still a viable and good way to merge graphs :)
Oh! well, I have no excuse then :) Thanks for lo
uschindler commented on PR #13649:
URL: https://github.com/apache/lucene/pull/13649#issuecomment-2289610751
Let me try it out later and merge it. Not at 💻.
--
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
uschindler commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1717404931
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ d
uschindler commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1717405562
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ d
uschindler commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1717404931
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ d
dweiss commented on PR #13649:
URL: https://github.com/apache/lucene/pull/13649#issuecomment-2289582426
I think this is ready to merge, if you agree, click on merge any time, Uwe.
Thanks!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on
dweiss commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1717397089
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ defau
gsmiller commented on PR #13644:
URL: https://github.com/apache/lucene/pull/13644#issuecomment-2289362834
Sounds right to me. Thanks @bugmakerr !
--
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
gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289348515
OK I pushed a small change (#13656) to `IndexSearcher` that reverts the
"common" (i.e., non-drill-sideways) code-path back to what it was before this
faceting change. It appears to reve
gsmiller merged PR #13656:
URL: https://github.com/apache/lucene/pull/13656
--
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.ap
gsmiller commented on PR #13656:
URL: https://github.com/apache/lucene/pull/13656#issuecomment-2289336382
Because this change is small (just undoing a small bit of #13568 in
IndexSearcher) and appears to correct the performance regressions the nightly
benchmarks demonstrated, I'm going to m
gsmiller commented on PR #13656:
URL: https://github.com/apache/lucene/pull/13656#issuecomment-2289330332
Also, benchmarks (wikimedium10m) comparing this PR to the tip of main seems
to indicate it reverses the performance impact we observed in the nightly runs:
```
gsmiller commented on PR #13656:
URL: https://github.com/apache/lucene/pull/13656#issuecomment-2289311663
Benchmarks (`wikimedium10m`) comparing this change to
`9b481f76f77b40a5061fb3b70763e4b4a06d31d0` (last commit on main before #13568)
show no performance regression, which is what we're
gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289299245
Thanks @epotyom ! I think I've isolated the change that's causing the
regression and I've put up a PR (#13656) to temporarily fix it. It just reverts
the IndexSearcher changes back to w
gsmiller opened a new pull request, #13656:
URL: https://github.com/apache/lucene/pull/13656
Note that this results in some silly code duplication so I don't think we
ought to merge this as-is. This is more to show the performance regression in
isolation.
--
This is an automated message
epotyom commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289274913
@gsmiller oh you are right, it was wikimedium10**k**. Results for
`wikimediumall` confirm that this PR causes the regression.
```
TaskQPS baseline
gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289239609
@epotyom just to confirm, are you using `wikimedium10k` or `wikimedium10m`?
As another check, I reverted out the changes to IndexSearcher and
DrillSideways/DrillSidewaysQuery (so
epotyom commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289194440
@gsmiller my results for `wikimedium10k` are quite different, but it might
be that something is wrong with my setup, double-checking; I'm also running
tests for `wikimediumall` but they
gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289194124
After looking through the change for possible ways it could impact
non-faceting tasks, the only thing that jumped out to me was the
synchronization of the collector list in `CollectorOw
bugmakerr commented on PR #13644:
URL: https://github.com/apache/lucene/pull/13644#issuecomment-2289159552
@gsmiller I add the entry under 9.12.0 Optimizations, please correct me if
I'm wrong.
--
This is an automated message from the Apache Git Service.
To respond to the message, plea
msokolov opened a new pull request, #13655:
URL: https://github.com/apache/lucene/pull/13655
This utility method is no longer used by the indexing chain, only by some
unit tests, so we missed tracking the change in behavior when merging. Maybe we
ought to change the unit test instead? But i
gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289109976
I was already setup to run benchmarks so I did a quick run with
`wikimedium10m` that isolated this change. The results are a bit surprising.
@epotyom I'll be curious if you reproduce si
mikemccand commented on issue #13403:
URL: https://github.com/apache/lucene/issues/13403#issuecomment-2289068792
I was reading up on PCA and uncovered [this interesting enumeration of the
challenges/prerequisites for
PCA](https://ekamperi.github.io/mathematics/2021/02/23/pca-limitations.htm
gsmiller commented on PR #13644:
URL: https://github.com/apache/lucene/pull/13644#issuecomment-2288982237
@bugmakerr could you add an entry in CHANGES.txt before we merge please?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to Gi
gsmiller commented on PR #13644:
URL: https://github.com/apache/lucene/pull/13644#issuecomment-2288953441
Looks good. Thanks @bugmakerr! (Looks like the check task failed in a
transient way... out of disk space? I ran the checks locally just to make sure
they pass and they do, as expect
epotyom commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2288910722
@mikemccand , @jpountz thanks for heads up! I'm running local benchmarks now
to check if it is related to this change.
--
This is an automated message from the Apache Git Service.
To r
msokolov commented on issue #13653:
URL: https://github.com/apache/lucene/issues/13653#issuecomment-2288831420
I'm suspecting this may have something to do with the connectComponents
changes - I'll dig
--
This is an automated message from the Apache Git Service.
To respond to the message,
mikemccand commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2288769876
[Term](https://home.apache.org/~mikemccand/lucenebench/Term.html) as well.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to Gi
jpountz commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2288723938
[SloppyPhrase](https://home.apache.org/~mikemccand/lucenebench/SloppyPhrase.html)
and [SpanNear](https://home.apache.org/~mikemccand/lucenebench/SpanNear.html)
had a noticeable slowdown
mikemccand commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2288666763
Hmm is it possible the `CollectorOwner` change here somehow led to [this
surprising nightly benchmark performance
drop](https://home.apache.org/~mikemccand/lucenebench/CombinedHighMed
msokolov merged PR #13654:
URL: https://github.com/apache/lucene/pull/13654
--
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.ap
msokolov commented on issue #13640:
URL: https://github.com/apache/lucene/issues/13640#issuecomment-2288628863
No it's not `HnswGraphBuilder` *per se* but the `HnswConcurrentMergeBuilder`
that calls `finish()` and thence `connectComponents`. Since it's part of the
o.a.l.u.hnsw package it's
msokolov commented on issue #13640:
URL: https://github.com/apache/lucene/issues/13640#issuecomment-2288618203
See https://github.com/apache/lucene/pull/13654/
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL ab
msokolov opened a new pull request, #13654:
URL: https://github.com/apache/lucene/pull/13654
another in an ongoing series of patches to address merge stability test
failures. Here we tweak the back-compat Lucene 94/95 writers so they will call
the new connectComponents() method (via `getCom
benwtrent commented on issue #13640:
URL: https://github.com/apache/lucene/issues/13640#issuecomment-2288559628
Not to be the bearer of bad news:
```
./gradlew test --tests TestLucene95HnswVectorsFormat.testMergeStability
-Dtests.seed=3D48DB65BA75CD03 -Dtests.locale=uz-Latn-UZ
-Dtest
benwtrent opened a new issue, #13653:
URL: https://github.com/apache/lucene/issues/13653
### Description
TestHnswFloatVectorGraph.testReadWrite fails due to a mismatched number of
connections.
`java.lang.AssertionError: arcs differ for node 3 expected:<[2, 52, 6]> but
was:<[2
ChrisHegarty commented on PR #13636:
URL: https://github.com/apache/lucene/pull/13636#issuecomment-2288426886
Apologies for not interacting on this PR before, but I was away. Thanks for
fixing the type of the length parameter in MemorySegmentIndexInput. I'll spend
some time studying the cod
uschindler commented on issue #13638:
URL: https://github.com/apache/lucene/issues/13638#issuecomment-2288112074
I have the git version shipped with TortoiseGit (rather new one, can check
later).
Basically the warning was not the most annoying thing, it was more that the
file was rep
uschindler commented on code in PR #13649:
URL: https://github.com/apache/lucene/pull/13649#discussion_r1716483653
##
gradle/validation/dependencies.gradle:
##
@@ -75,6 +75,22 @@ configure(rootProject) {
it.dependsOn(":versionCatalogFormatDeps")
}
+ // correct crlf/ d
dweiss commented on issue #13638:
URL: https://github.com/apache/lucene/issues/13638#issuecomment-2288034567
I compared main with 8f50976c268. When you list the plugins applied to the
root project, it'll give you this:
```
class org.gradle.api.plugins.HelpTasksPlugin$Inject
class or
dweiss commented on issue #13638:
URL: https://github.com/apache/lucene/issues/13638#issuecomment-2288013013
It's the plugin that formats version catalog - versionCatalogFormatDeps task
writes back the file with the default encoding/ platform line endings. I've
added a post-processing step
53 matches
Mail list logo
