DRAFT: Vectorize ForUtil encoding for the 9.0 codec (same format)

gradle/generation/forUtil.gradle

@@ -20,12 +20,12 @@
 configure(project(":lucene:core")) {
 
   task generateForUtilInternal() {
-    description "Regenerate gen_ForUtil.py"
+    description "Regenerate gen_DefaultForUtil90.py"
     group "generation"
 
-    def genDir = file("src/java/org/apache/lucene/codecs/lucene90")
-    def genScript = file("${genDir}/gen_ForUtil.py")
-    def genOutput = file("${genDir}/ForUtil.java")
+    def genDir = file("src/java/org/apache/lucene/internal/vectorization")
+    def genScript = file("${genDir}/gen_DefaultForUtil90.py")
+    def genOutput = file("${genDir}/DefaultForUtil90.java")
 
     inputs.file genScript
     outputs.file genOutput
@@ -40,6 +40,28 @@ configure(project(":lucene:core")) {
   }
 
   regenerate.dependsOn wrapWithPersistentChecksums(generateForUtilInternal, [ andThenTasks: ["spotlessJava", "spotlessJavaApply"] ])
+
+  task generatePanamaForUtilInternal() {
+    description "Regenerate gen_PanamaForUtil.py"
+    group "generation"
+
+    def genDir = file("src/java20/org/apache/lucene/internal/vectorization")
+    def genScript = file("${genDir}/gen_PanamaForUtil90.py")
+    def genOutput = file("${genDir}/PanamaForUtil90.java")
+
+    inputs.file genScript
+    outputs.file genOutput
+
+    doLast {
+      quietExec {
+        workingDir genDir
+        executable project.externalTool("python3")
+        args = [ '-B', genScript ]
+      }
+    }
+  }
+
+  regenerate.dependsOn wrapWithPersistentChecksums(generatePanamaForUtilInternal, [ andThenTasks: ["spotlessJava", "spotlessJavaApply"] ])
 }
 
 configure(project(":lucene:backward-codecs")) {

lucene/core/src/generated/checksums/generateForUtil.json

@@ -1,4 +1,4 @@
 {
-    "lucene/core/src/java/org/apache/lucene/codecs/lucene90/ForUtil.java": "f2091e2b7284b70c740052a9b0ee389e67eb48a9",
-    "lucene/core/src/java/org/apache/lucene/codecs/lucene90/gen_ForUtil.py": "7a137c58f88d68247be4368122780fa9cc8dce3e"
+    "lucene/core/src/java/org/apache/lucene/internal/vectorization/DefaultForUtil90.java": "8760243bd670376792b272f0a5cc54acb79c3ed6",
+    "lucene/core/src/java/org/apache/lucene/internal/vectorization/gen_DefaultForUtil90.py": "71cef263b135816680b29dd97a90961402aa005c"
 }

lucene/core/src/generated/checksums/generatePanamaForUtil.json

@@ -0,0 +1,4 @@
+{
+    "lucene/core/src/java20/org/apache/lucene/internal/vectorization/PanamaForUtil90.java": "5db61312b5f550c0feeade7a6cdff257ad7dbc8d",
+    "lucene/core/src/java20/org/apache/lucene/internal/vectorization/gen_PanamaForUtil90.py": "0d412b484d6ddb269bdb91f0203dad19cd998061"
+}

lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90PostingsFormat.java

@@ -31,6 +31,7 @@
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
 import org.apache.lucene.index.TermState;
+import org.apache.lucene.internal.vectorization.ForUtil90;
 import org.apache.lucene.store.DataOutput;
 import org.apache.lucene.util.IOUtils;
 import org.apache.lucene.util.packed.PackedInts;
@@ -356,7 +357,7 @@ public final class Lucene90PostingsFormat extends PostingsFormat {
   public static final String PAY_EXTENSION = "pay";
 
   /** Size of blocks. */
-  public static final int BLOCK_SIZE = ForUtil.BLOCK_SIZE;
+  public static final int BLOCK_SIZE = ForUtil90.BLOCK_SIZE;
 
   /**
    * Expert: The maximum number of skip levels. Smaller values result in slightly smaller indexes,
@@ -449,12 +450,12 @@ public static final class IntBlockTermState extends BlockTermState {
     public long payStartFP;
     /**
      * file offset for the start of the skip list, relative to docStartFP, if there are more than
-     * {@link ForUtil#BLOCK_SIZE} docs; otherwise -1
+     * {@link ForUtil90#BLOCK_SIZE} docs; otherwise -1
      */
     public long skipOffset;
     /**
      * file offset for the last position in the last block, if there are more than {@link
-     * ForUtil#BLOCK_SIZE} positions; otherwise -1
+     * ForUtil90#BLOCK_SIZE} positions; otherwise -1
      */
     public long lastPosBlockOffset;
     /**

lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90PostingsReader.java

@@ -16,14 +16,14 @@
  */
 package org.apache.lucene.codecs.lucene90;
 
-import static org.apache.lucene.codecs.lucene90.ForUtil.BLOCK_SIZE;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.DOC_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.MAX_SKIP_LEVELS;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.PAY_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.POS_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.TERMS_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.VERSION_CURRENT;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.VERSION_START;
+import static org.apache.lucene.internal.vectorization.ForUtil90.BLOCK_SIZE;
 
 import java.io.IOException;
 import java.util.Arrays;
@@ -39,6 +39,7 @@
 import org.apache.lucene.index.PostingsEnum;
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SlowImpactsEnum;
+import org.apache.lucene.internal.vectorization.ForUtil90;
 import org.apache.lucene.store.DataInput;
 import org.apache.lucene.store.IndexInput;
 import org.apache.lucene.util.ArrayUtil;
@@ -306,7 +307,7 @@ public ImpactsEnum impacts(FieldInfo fieldInfo, BlockTermState state, int flags)
 
   final class BlockDocsEnum extends PostingsEnum {
 
-    final PForUtil pforUtil = new PForUtil(new ForUtil());
+    final PForUtil pforUtil = new PForUtil();
 
     private final long[] docBuffer = new long[BLOCK_SIZE + 1];
     private final long[] freqBuffer = new long[BLOCK_SIZE];
@@ -394,7 +395,7 @@ public PostingsEnum reset(IntBlockTermState termState, int flags) throws IOExcep
       this.needsFreq = PostingsEnum.featureRequested(flags, PostingsEnum.FREQS);
       this.isFreqsRead = true;
       if (indexHasFreq == false || needsFreq == false) {
-        for (int i = 0; i < ForUtil.BLOCK_SIZE; ++i) {
+        for (int i = 0; i < ForUtil90.BLOCK_SIZE; ++i) {
           freqBuffer[i] = 1;
         }
       }
@@ -563,7 +564,7 @@ public long cost() {
   // Also handles payloads + offsets
   final class EverythingEnum extends PostingsEnum {
 
-    final PForUtil pforUtil = new PForUtil(new ForUtil());
+    final PForUtil pforUtil = new PForUtil();
 
     private final long[] docBuffer = new long[BLOCK_SIZE + 1];
     private final long[] freqBuffer = new long[BLOCK_SIZE + 1];
@@ -1047,7 +1048,7 @@ public long cost() {
 
   final class BlockImpactsDocsEnum extends ImpactsEnum {
 
-    final PForUtil pforUtil = new PForUtil(new ForUtil());
+    final PForUtil pforUtil = new PForUtil();
 
     private final long[] docBuffer = new long[BLOCK_SIZE + 1];
     private final long[] freqBuffer = new long[BLOCK_SIZE];
@@ -1240,7 +1241,7 @@ public long cost() {
 
   final class BlockImpactsPostingsEnum extends ImpactsEnum {
 
-    final PForUtil pforUtil = new PForUtil(new ForUtil());
+    final PForUtil pforUtil = new PForUtil();
 
     private final long[] docBuffer = new long[BLOCK_SIZE];
     private final long[] freqBuffer = new long[BLOCK_SIZE];
@@ -1532,7 +1533,7 @@ public long cost() {
 
   final class BlockImpactsEverythingEnum extends ImpactsEnum {
 
-    final PForUtil pforUtil = new PForUtil(new ForUtil());
+    final PForUtil pforUtil = new PForUtil();
 
     private final long[] docBuffer = new long[BLOCK_SIZE];
     private final long[] freqBuffer = new long[BLOCK_SIZE];
@@ -1696,7 +1697,7 @@ public BlockImpactsEverythingEnum(FieldInfo fieldInfo, IntBlockTermState termSta
           docFreq);
 
       if (indexHasFreq == false) {
-        for (int i = 0; i < ForUtil.BLOCK_SIZE; ++i) {
+        for (int i = 0; i < ForUtil90.BLOCK_SIZE; ++i) {
           freqBuffer[i] = 1;
         }
       }

lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90PostingsWriter.java

@@ -16,13 +16,13 @@
  */
 package org.apache.lucene.codecs.lucene90;
 
-import static org.apache.lucene.codecs.lucene90.ForUtil.BLOCK_SIZE;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.DOC_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.MAX_SKIP_LEVELS;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.PAY_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.POS_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.TERMS_CODEC;
 import static org.apache.lucene.codecs.lucene90.Lucene90PostingsFormat.VERSION_CURRENT;
+import static org.apache.lucene.internal.vectorization.ForUtil90.BLOCK_SIZE;
 
 import java.io.IOException;
 import org.apache.lucene.codecs.BlockTermState;
@@ -110,7 +110,7 @@ public Lucene90PostingsWriter(SegmentWriteState state) throws IOException {
     try {
       CodecUtil.writeIndexHeader(
           docOut, DOC_CODEC, VERSION_CURRENT, state.segmentInfo.getId(), state.segmentSuffix);
-      pforUtil = new PForUtil(new ForUtil());
+      pforUtil = new PForUtil();
       if (state.fieldInfos.hasProx()) {
         posDeltaBuffer = new long[BLOCK_SIZE];
         String posFileName =

lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90SkipReader.java

@@ -19,6 +19,7 @@
 import java.io.IOException;
 import java.util.Arrays;
 import org.apache.lucene.codecs.MultiLevelSkipListReader;
+import org.apache.lucene.internal.vectorization.ForUtil90;
 import org.apache.lucene.store.IndexInput;
 
 /**
@@ -67,7 +68,7 @@ public Lucene90SkipReader(
       boolean hasPos,
       boolean hasOffsets,
       boolean hasPayloads) {
-    super(skipStream, maxSkipLevels, ForUtil.BLOCK_SIZE, 8);
+    super(skipStream, maxSkipLevels, ForUtil90.BLOCK_SIZE, 8);
     docPointer = new long[maxSkipLevels];
     if (hasPos) {
       posPointer = new long[maxSkipLevels];
@@ -95,7 +96,7 @@ public Lucene90SkipReader(
    * the last block boundary 2. moving into the vInt block
    */
   protected int trim(int df) {
-    return df % ForUtil.BLOCK_SIZE == 0 ? df - 1 : df;
+    return df % ForUtil90.BLOCK_SIZE == 0 ? df - 1 : df;
   }
 
   public void init(

lucene/core/src/java/org/apache/lucene/codecs/lucene90/PForUtil.java

@@ -18,6 +18,8 @@
 
 import java.io.IOException;
 import java.util.Arrays;
+import org.apache.lucene.internal.vectorization.ForUtil90;
+import org.apache.lucene.internal.vectorization.VectorizationProvider;
 import org.apache.lucene.store.DataInput;
 import org.apache.lucene.store.DataOutput;
 import org.apache.lucene.util.LongHeap;
@@ -26,35 +28,38 @@
 /** Utility class to encode sequences of 128 small positive integers. */
 final class PForUtil {
 
+  private static final VectorizationProvider VPROVIDER = VectorizationProvider.getInstance();
+
   private static final int MAX_EXCEPTIONS = 7;
-  private static final int HALF_BLOCK_SIZE = ForUtil.BLOCK_SIZE / 2;
+  private static final int HALF_BLOCK_SIZE = ForUtil90.BLOCK_SIZE / 2;
 
   // IDENTITY_PLUS_ONE[i] == i + 1
-  private static final long[] IDENTITY_PLUS_ONE = new long[ForUtil.BLOCK_SIZE];
+  private static final long[] IDENTITY_PLUS_ONE = new long[ForUtil90.BLOCK_SIZE];
 
   static {
-    for (int i = 0; i < ForUtil.BLOCK_SIZE; ++i) {
+    for (int i = 0; i < ForUtil90.BLOCK_SIZE; ++i) {
       IDENTITY_PLUS_ONE[i] = i + 1;
     }
   }
 
   static boolean allEqual(long[] l) {
-    for (int i = 1; i < ForUtil.BLOCK_SIZE; ++i) {
+    for (int i = 1; i < ForUtil90.BLOCK_SIZE; ++i) {
       if (l[i] != l[0]) {
         return false;
       }
     }
     return true;
   }
 
-  private final ForUtil forUtil;
+  private final ForUtil90 forUtil;
   // buffer for reading exception data; each exception uses two bytes (pos + high-order bits of the
   // exception)
   private final byte[] exceptionBuff = new byte[MAX_EXCEPTIONS * 2];
 
-  PForUtil(ForUtil forUtil) {
-    assert ForUtil.BLOCK_SIZE <= 256 : "blocksize must fit in one byte. got " + ForUtil.BLOCK_SIZE;
-    this.forUtil = forUtil;
+  PForUtil() {
+    assert ForUtil90.BLOCK_SIZE <= 256
+        : "blocksize must fit in one byte. got " + ForUtil90.BLOCK_SIZE;
+    this.forUtil = VPROVIDER.newForUtil90();
   }
 
   /** Encode 128 integers from {@code longs} into {@code out}. */
@@ -65,7 +70,7 @@ void encode(long[] longs, DataOutput out) throws IOException {
       top.push(longs[i]);
     }
     long topValue = top.top();
-    for (int i = MAX_EXCEPTIONS + 1; i < ForUtil.BLOCK_SIZE; ++i) {
+    for (int i = MAX_EXCEPTIONS + 1; i < ForUtil90.BLOCK_SIZE; ++i) {
       if (longs[i] > topValue) {
         topValue = top.updateTop(longs[i]);
       }
@@ -90,7 +95,7 @@ void encode(long[] longs, DataOutput out) throws IOException {
     final byte[] exceptions = new byte[numExceptions * 2];
     if (numExceptions > 0) {
       int exceptionCount = 0;
-      for (int i = 0; i < ForUtil.BLOCK_SIZE; ++i) {
+      for (int i = 0; i < ForUtil90.BLOCK_SIZE; ++i) {
         if (longs[i] > maxUnpatchedValue) {
           exceptions[exceptionCount * 2] = (byte) i;
           exceptions[exceptionCount * 2 + 1] = (byte) (longs[i] >>> patchedBitsRequired);
@@ -122,7 +127,7 @@ void decode(DataInput in, long[] longs) throws IOException {
     final int bitsPerValue = token & 0x1f;
     final int numExceptions = token >>> 5;
     if (bitsPerValue == 0) {
-      Arrays.fill(longs, 0, ForUtil.BLOCK_SIZE, in.readVLong());
+      Arrays.fill(longs, 0, ForUtil90.BLOCK_SIZE, in.readVLong());
     } else {
       forUtil.decode(bitsPerValue, in, longs);
     }
@@ -184,9 +189,9 @@ void skip(DataInput in) throws IOException {
    * there are no exceptions to apply.
    */
   private static void prefixSumOfOnes(long[] longs, long base) {
-    System.arraycopy(IDENTITY_PLUS_ONE, 0, longs, 0, ForUtil.BLOCK_SIZE);
+    System.arraycopy(IDENTITY_PLUS_ONE, 0, longs, 0, ForUtil90.BLOCK_SIZE);
     // This loop gets auto-vectorized
-    for (int i = 0; i < ForUtil.BLOCK_SIZE; ++i) {
+    for (int i = 0; i < ForUtil90.BLOCK_SIZE; ++i) {
       longs[i] += base;
     }
   }
@@ -196,7 +201,7 @@ private static void prefixSumOfOnes(long[] longs, long base) {
    * this assumes there are no exceptions to apply.
    */
   private static void prefixSumOf(long[] longs, long base, long val) {
-    for (int i = 0; i < ForUtil.BLOCK_SIZE; i++) {
+    for (int i = 0; i < ForUtil90.BLOCK_SIZE; i++) {
       longs[i] = (i + 1) * val + base;
     }
   }
@@ -232,7 +237,7 @@ private static void prefixSum32(long[] longs, long base) {
     innerPrefixSum32(longs);
     expand32(longs);
     final long l = longs[HALF_BLOCK_SIZE - 1];
-    for (int i = HALF_BLOCK_SIZE; i < ForUtil.BLOCK_SIZE; ++i) {
+    for (int i = HALF_BLOCK_SIZE; i < ForUtil90.BLOCK_SIZE; ++i) {
       longs[i] += l;
     }
   }