Reduce allocation rate in HNSW concurrent merge (#14011)

The PR removes the allocation of a new `LockedRow` for each locking operation in `HnswLock`. Even though the object was very quickly released, and JIT supports on-stack allocation, it didn't happen in my experiments on OpenJDK 21 - it's easy to avoid the allocation, rather than rely on the JIT.

lucene/core/src/java/org/apache/lucene/util/hnsw/HnswConcurrentMergeBuilder.java

@@ -24,6 +24,7 @@ import java.util.ArrayList;
 import java.util.List;
 import java.util.concurrent.Callable;
 import java.util.concurrent.atomic.AtomicInteger;
+import java.util.concurrent.locks.Lock;
 import org.apache.lucene.search.TaskExecutor;
 import org.apache.lucene.util.BitSet;
 import org.apache.lucene.util.FixedBitSet;
@@ -56,7 +57,7 @@ public class HnswConcurrentMergeBuilder implements HnswBuilder {
     this.taskExecutor = taskExecutor;
     AtomicInteger workProgress = new AtomicInteger(0);
     workers = new ConcurrentMergeWorker[numWorker];
-    hnswLock = new HnswLock(hnsw);
+    hnswLock = new HnswLock();
     for (int i = 0; i < numWorker; i++) {
       workers[i] =
           new ConcurrentMergeWorker(
@@ -221,13 +222,16 @@ public class HnswConcurrentMergeBuilder implements HnswBuilder {
 
     @Override
     void graphSeek(HnswGraph graph, int level, int targetNode) {
-      try (HnswLock.LockedRow rowLock = hnswLock.read(level, targetNode)) {
+      Lock lock = hnswLock.read(level, targetNode);
-        NeighborArray neighborArray = rowLock.row();
+      try {
+        NeighborArray neighborArray = ((OnHeapHnswGraph) graph).getNeighbors(level, targetNode);
         if (nodeBuffer == null || nodeBuffer.length < neighborArray.size()) {
           nodeBuffer = new int[neighborArray.size()];
         }
         size = neighborArray.size();
-        if (size >= 0) System.arraycopy(neighborArray.nodes(), 0, nodeBuffer, 0, size);
+        System.arraycopy(neighborArray.nodes(), 0, nodeBuffer, 0, size);
+      } finally {
+        lock.unlock();
       }
       upto = -1;
     }

lucene/core/src/java/org/apache/lucene/util/hnsw/HnswGraphBuilder.java

@@ -27,6 +27,7 @@ import java.util.Locale;
 import java.util.Objects;
 import java.util.SplittableRandom;
 import java.util.concurrent.TimeUnit;
+import java.util.concurrent.locks.Lock;
 import org.apache.lucene.search.KnnCollector;
 import org.apache.lucene.search.TopDocs;
 import org.apache.lucene.util.FixedBitSet;
@@ -338,9 +339,12 @@ public class HnswGraphBuilder implements HnswBuilder {
       }
       int nbr = candidates.nodes()[i];
       if (hnswLock != null) {
-        try (HnswLock.LockedRow rowLock = hnswLock.write(level, nbr)) {
+        Lock lock = hnswLock.write(level, nbr);
-          NeighborArray nbrsOfNbr = rowLock.row();
+        try {
+          NeighborArray nbrsOfNbr = getGraph().getNeighbors(level, nbr);
           nbrsOfNbr.addAndEnsureDiversity(node, candidates.scores()[i], nbr, scorerSupplier);
+        } finally {
+          lock.unlock();
         }
       } else {
         NeighborArray nbrsOfNbr = hnsw.getNeighbors(level, nbr);

lucene/core/src/java/org/apache/lucene/util/hnsw/HnswLock.java

@@ -17,49 +17,39 @@
 
 package org.apache.lucene.util.hnsw;
 
-import java.io.Closeable;
 import java.util.concurrent.locks.Lock;
 import java.util.concurrent.locks.ReentrantReadWriteLock;
 
 /**
- * Provide (read-and-write) locked access to rows of an OnHeapHnswGraph. For use by
+ * Provide (read-and-write) striped locks for access to nodes of an {@link OnHeapHnswGraph}. For use
- * HnswConcurrentMerger and its HnswGraphBuilders.
+ * by {@link HnswConcurrentMergeBuilder} and its HnswGraphBuilders.
  */
 final class HnswLock {
   private static final int NUM_LOCKS = 512;
   private final ReentrantReadWriteLock[] locks;
-  private final OnHeapHnswGraph graph;
 
-  HnswLock(OnHeapHnswGraph graph) {
+  HnswLock() {
-    this.graph = graph;
     locks = new ReentrantReadWriteLock[NUM_LOCKS];
     for (int i = 0; i < NUM_LOCKS; i++) {
       locks[i] = new ReentrantReadWriteLock();
     }
   }
 
-  LockedRow read(int level, int node) {
+  Lock read(int level, int node) {
     int lockid = hash(level, node) % NUM_LOCKS;
     Lock lock = locks[lockid].readLock();
     lock.lock();
-    return new LockedRow(graph.getNeighbors(level, node), lock);
+    return lock;
   }
 
-  LockedRow write(int level, int node) {
+  Lock write(int level, int node) {
     int lockid = hash(level, node) % NUM_LOCKS;
     Lock lock = locks[lockid].writeLock();
     lock.lock();
-    return new LockedRow(graph.getNeighbors(level, node), lock);
+    return lock;
   }
 
-  record LockedRow(NeighborArray row, Lock lock) implements Closeable {
+  private static int hash(int v1, int v2) {
-    @Override
-    public void close() {
-      lock.unlock();
-    }
-  }
-
-  static int hash(int v1, int v2) {
     return v1 * 31 + v2;
   }
 }