You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@lucene.apache.org by jp...@apache.org on 2024/12/23 08:40:03 UTC
(lucene) branch main updated: Make inlining of FixedBitSet#get more predictable when checking live docs. (#14077)
This is an automated email from the ASF dual-hosted git repository.
jpountz pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/lucene.git
The following commit(s) were added to refs/heads/main by this push:
new b0585f11de1 Make inlining of FixedBitSet#get more predictable when checking live docs. (#14077)
b0585f11de1 is described below
commit b0585f11de1aca1a5d730408793b268daf2f9a90
Author: Adrien Grand <jp...@gmail.com>
AuthorDate: Mon Dec 23 09:39:59 2024 +0100
Make inlining of FixedBitSet#get more predictable when checking live docs. (#14077)
This helps make calls sites of `Bits#get` bimorphic at most when checking live
docs. This helps because calls to `FixedBitSet#get` can then be inlined when
live docs are stored in a `FixedBitSet`. Another reason why this is important
is because these calls could be subject to auto-vectorizing when applied to an
array of doc IDs, which cannot apply if the `FixedBitSet#get` call is not
inlined.
While live docs are stored in a `FixedBitSet` by default, some features like
document-level security sometimes create wrappers.
---
.../org/apache/lucene/search/IndexSearcher.java | 4 ++-
.../java/org/apache/lucene/search/ScorerUtil.java | 38 ++++++++++++++++++++++
2 files changed, 41 insertions(+), 1 deletion(-)
diff --git a/lucene/core/src/java/org/apache/lucene/search/IndexSearcher.java b/lucene/core/src/java/org/apache/lucene/search/IndexSearcher.java
index 6e8bbf81966..e3074a96d88 100644
--- a/lucene/core/src/java/org/apache/lucene/search/IndexSearcher.java
+++ b/lucene/core/src/java/org/apache/lucene/search/IndexSearcher.java
@@ -845,7 +845,9 @@ public class IndexSearcher {
scorer = new TimeLimitingBulkScorer(scorer, queryTimeout);
}
try {
- scorer.score(leafCollector, ctx.reader().getLiveDocs(), minDocId, maxDocId);
+ // Optimize for the case when live docs are stored in a FixedBitSet.
+ Bits acceptDocs = ScorerUtil.likelyFixedBitSet(ctx.reader().getLiveDocs());
+ scorer.score(leafCollector, acceptDocs, minDocId, maxDocId);
} catch (
@SuppressWarnings("unused")
CollectionTerminatedException e) {
diff --git a/lucene/core/src/java/org/apache/lucene/search/ScorerUtil.java b/lucene/core/src/java/org/apache/lucene/search/ScorerUtil.java
index 1154dc1b154..2081f9ac717 100644
--- a/lucene/core/src/java/org/apache/lucene/search/ScorerUtil.java
+++ b/lucene/core/src/java/org/apache/lucene/search/ScorerUtil.java
@@ -30,7 +30,9 @@ import org.apache.lucene.index.PostingsEnum;
import org.apache.lucene.index.TermsEnum;
import org.apache.lucene.store.ByteBuffersDirectory;
import org.apache.lucene.store.Directory;
+import org.apache.lucene.util.Bits;
import org.apache.lucene.util.BytesRef;
+import org.apache.lucene.util.FixedBitSet;
import org.apache.lucene.util.PriorityQueue;
/** Util class for Scorer related methods */
@@ -108,4 +110,40 @@ class ScorerUtil {
}
return scorable;
}
+
+ /**
+ * Optimize {@link Bits} representing the set of accepted documents for the case when it is likely
+ * implemented via a {@link FixedBitSet}. This helps make calls to {@link Bits#get(int)}
+ * inlinable, which in-turn helps speed up query evaluation. This is especially helpful as
+ * inlining will sometimes enable auto-vectorizing shifts and masks that are done in {@link
+ * FixedBitSet#get(int)}.
+ */
+ static Bits likelyFixedBitSet(Bits acceptDocs) {
+ if (acceptDocs instanceof FixedBitSet) {
+ return acceptDocs;
+ } else if (acceptDocs != null) {
+ return new FilterBits(acceptDocs);
+ } else {
+ return null;
+ }
+ }
+
+ private static class FilterBits implements Bits {
+
+ private final Bits in;
+
+ FilterBits(Bits in) {
+ this.in = in;
+ }
+
+ @Override
+ public boolean get(int index) {
+ return in.get(index);
+ }
+
+ @Override
+ public int length() {
+ return in.length();
+ }
+ }
}