Adjust GPU graph building params

docs/changelog/137074.yaml

@@ -0,0 +1,5 @@
+pr: 137074
+summary: Adjust GPU graph building params
+area: Search
+type: enhancement
+issues: []

x-pack/plugin/gpu/src/main/java/org/elasticsearch/xpack/gpu/GPUPlugin.java

@@ -7,7 +7,6 @@
 package org.elasticsearch.xpack.gpu;
 
 import org.apache.lucene.codecs.KnnVectorsFormat;
-import org.apache.lucene.util.hnsw.HnswGraphBuilder;
 import org.elasticsearch.common.settings.Setting;
 import org.elasticsearch.common.util.FeatureFlag;
 import org.elasticsearch.index.mapper.vectors.DenseVectorFieldMapper;
@@ -94,13 +93,14 @@ private static KnnVectorsFormat getVectorsFormat(
         DenseVectorFieldMapper.DenseVectorIndexOptions indexOptions,
         DenseVectorFieldMapper.VectorSimilarity similarity
     ) {
+        // TODO: cuvs 2025.12 will provide an API for converting HNSW CPU Params to Cagra params; use that instead
         if (indexOptions.getType() == DenseVectorFieldMapper.VectorIndexType.HNSW) {
             DenseVectorFieldMapper.HnswIndexOptions hnswIndexOptions = (DenseVectorFieldMapper.HnswIndexOptions) indexOptions;
             int efConstruction = hnswIndexOptions.efConstruction();
-            if (efConstruction == HnswGraphBuilder.DEFAULT_BEAM_WIDTH) {
-                efConstruction = ES92GpuHnswVectorsFormat.DEFAULT_BEAM_WIDTH; // default value for GPU graph construction is 128
-            }
-            return new ES92GpuHnswVectorsFormat(hnswIndexOptions.m(), efConstruction);
+            int m = hnswIndexOptions.m();
+            int gpuM = 2 + m * 2 / 3;
+            int gpuEfConstruction = m + m * efConstruction / 256;
+            return new ES92GpuHnswVectorsFormat(gpuM, gpuEfConstruction);
         } else if (indexOptions.getType() == DenseVectorFieldMapper.VectorIndexType.INT8_HNSW) {
             if (similarity == DenseVectorFieldMapper.VectorSimilarity.MAX_INNER_PRODUCT) {
                 throw new IllegalArgumentException(
@@ -115,16 +115,10 @@ private static KnnVectorsFormat getVectorsFormat(
             }
             DenseVectorFieldMapper.Int8HnswIndexOptions int8HnswIndexOptions = (DenseVectorFieldMapper.Int8HnswIndexOptions) indexOptions;
             int efConstruction = int8HnswIndexOptions.efConstruction();
-            if (efConstruction == HnswGraphBuilder.DEFAULT_BEAM_WIDTH) {
-                efConstruction = ES92GpuHnswVectorsFormat.DEFAULT_BEAM_WIDTH; // default value for GPU graph construction is 128
-            }
-            return new ES92GpuHnswSQVectorsFormat(
-                int8HnswIndexOptions.m(),
-                efConstruction,
-                int8HnswIndexOptions.confidenceInterval(),
-                7,
-                false
-            );
+            int m = int8HnswIndexOptions.m();
+            int gpuM = 2 + m * 2 / 3;
+            int gpuEfConstruction = m + m * efConstruction / 256;
+            return new ES92GpuHnswSQVectorsFormat(gpuM, gpuEfConstruction, int8HnswIndexOptions.confidenceInterval(), 7, false);
         } else {
             throw new IllegalArgumentException(
                 "GPU vector indexing is not supported on this vector type: [" + indexOptions.getType() + "]"

x-pack/plugin/gpu/src/main/java/org/elasticsearch/xpack/gpu/codec/ES92GpuHnswVectorsFormat.java

@@ -13,6 +13,7 @@
 import org.apache.lucene.codecs.hnsw.FlatVectorScorerUtil;
 import org.apache.lucene.codecs.hnsw.FlatVectorsFormat;
 import org.apache.lucene.codecs.lucene99.Lucene99FlatVectorsFormat;
+import org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsFormat;
 import org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsReader;
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
@@ -36,8 +37,9 @@ public class ES92GpuHnswVectorsFormat extends KnnVectorsFormat {
     static final String LUCENE99_HNSW_VECTOR_INDEX_EXTENSION = "vex";
     static final int LUCENE99_VERSION_CURRENT = VERSION_GROUPVARINT;
 
-    static final int DEFAULT_MAX_CONN = 16; // graph degree
-    public static final int DEFAULT_BEAM_WIDTH = 128; // intermediate graph degree
+    public static final int DEFAULT_MAX_CONN = (2 + Lucene99HnswVectorsFormat.DEFAULT_MAX_CONN * 2 / 3); // graph degree
+    public static final int DEFAULT_BEAM_WIDTH = Lucene99HnswVectorsFormat.DEFAULT_MAX_CONN + Lucene99HnswVectorsFormat.DEFAULT_MAX_CONN
+        * Lucene99HnswVectorsFormat.DEFAULT_BEAM_WIDTH / 256; // intermediate graph degree
     static final int MIN_NUM_VECTORS_FOR_GPU_BUILD = 2;
 
     private static final FlatVectorsFormat flatVectorsFormat = new Lucene99FlatVectorsFormat(

x-pack/plugin/gpu/src/main/java/org/elasticsearch/xpack/gpu/codec/ES92GpuHnswVectorsWriter.java

@@ -332,6 +332,7 @@ private CagraIndex buildGPUIndex(
             .withCagraGraphBuildAlgo(CagraIndexParams.CagraGraphBuildAlgo.NN_DESCENT)
             .withGraphDegree(M)
             .withIntermediateGraphDegree(beamWidth)
+            .withNNDescentNumIterations(5)
             .withMetric(distanceType)
             .build();
 

x-pack/plugin/gpu/src/test/java/org/elasticsearch/xpack/gpu/codec/GPUDenseVectorFieldMapperTests.java

@@ -44,7 +44,7 @@ public void testKnnVectorsFormat() throws IOException {
         // TODO improve test with custom parameters
         KnnVectorsFormat knnVectorsFormat = getKnnVectorsFormat("hnsw");
         String expectedStr = "Lucene99HnswVectorsFormat(name=Lucene99HnswVectorsFormat, "
-            + "maxConn=16, beamWidth=128, flatVectorFormat=Lucene99FlatVectorsFormat)";
+            + "maxConn=12, beamWidth=22, flatVectorFormat=Lucene99FlatVectorsFormat)";
         assertEquals(expectedStr, knnVectorsFormat.toString());
     }
 
@@ -53,7 +53,7 @@ public void testKnnQuantizedHNSWVectorsFormat() throws IOException {
         // TOD improve the test with custom parameters
         KnnVectorsFormat knnVectorsFormat = getKnnVectorsFormat("int8_hnsw");
         String expectedStr = "Lucene99HnswVectorsFormat(name=Lucene99HnswVectorsFormat, "
-            + "maxConn=16, beamWidth=128, flatVectorFormat=ES814ScalarQuantizedVectorsFormat";
+            + "maxConn=12, beamWidth=22, flatVectorFormat=ES814ScalarQuantizedVectorsFormat";
         assertTrue(knnVectorsFormat.toString().startsWith(expectedStr));
     }