gf2121 commented on code in PR #14935: URL: https://github.com/apache/lucene/pull/14935#discussion_r2200926950
########## lucene/benchmark-jmh/src/java/org/apache/lucene/benchmark/jmh/BitsetToArrayBenchmark.java: ########## @@ -0,0 +1,379 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.lucene.benchmark.jmh; + +import java.util.SplittableRandom; +import java.util.concurrent.TimeUnit; +import java.util.stream.IntStream; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +import org.openjdk.jmh.annotations.Param; +import org.openjdk.jmh.annotations.Scope; +import org.openjdk.jmh.annotations.Setup; +import org.openjdk.jmh.annotations.State; +import org.openjdk.jmh.annotations.Warmup; + +@BenchmarkMode(Mode.Throughput) +@OutputTimeUnit(TimeUnit.MICROSECONDS) +@State(Scope.Benchmark) +@Warmup(iterations = 3, time = 1) +@Measurement(iterations = 5, time = 1) +@Fork( + value = 1, + jvmArgsAppend = {"-Xmx1g", "-Xms1g", "-XX:+AlwaysPreTouch"}) +public class BitsetToArrayBenchmark { + + private final SplittableRandom R = new SplittableRandom(4314123142L); + + @Param({"5", "10", "20", "30", "40", "50", "60"}) + int bitCount; + + private final int[] scratch = new int[64]; + private long word; + private int[] resultArray; + private int base; + private int offset; + + @Setup(Level.Trial) + public void setup() { + base = R.nextInt(1000); + resultArray = new int[bitCount + Long.SIZE]; + } + + @Setup(Level.Invocation) + public void setupInvocation() { + word = 0L; + while (Long.bitCount(word) < bitCount) { + word |= 1L << R.nextInt(64); + } + offset = R.nextInt(64); + } + + @Benchmark + public int whileLoop() { + return _whileLoop(word, resultArray, offset, base); + } + + @Benchmark + public int forLoop() { + return _forLoop(word, resultArray, offset, base); + } + + @Benchmark + public int forLoopManualUnrolling() { + return _forLoopManualUnrolling(word, resultArray, offset, base); + } + + @Benchmark + public int dense() { + return _dense(word, resultArray, offset, base); + } + + @Benchmark + public int denseBranchLess() { + return _denseBranchLess(word, resultArray, offset, base); + } + + @Benchmark + public int denseBranchLessUnrolling() { + return _denseBranchLessUnrolling(word, resultArray, offset, base); + } + + @Benchmark + public int denseBranchLessVectorized() { + return _denseBranchLessVectorized(word, resultArray, offset, base, scratch); + } + + @Benchmark + public int denseInvert() { + return _denseInvert(word, resultArray, offset, base); + } + + @Benchmark + public int hybrid() { + return _hybrid(word, resultArray, offset, base, scratch); + } + + private static int _whileLoop(long word, int[] resultArray, int offset, int base) { + while (word != 0) { + int bit = Long.numberOfTrailingZeros(word); + resultArray[offset++] = base + bit; + word ^= 1L << bit; + } + return offset; + } + + private static int _forLoop(long word, int[] resultArray, int offset, int base) { + int to = offset + Long.bitCount(word); + for (int i = offset; i < to; i++) { + int bit = Long.numberOfTrailingZeros(word); + resultArray[i] = base + bit; + word ^= 1L << bit; + } + return to; + } + + private static int _forLoopManualUnrolling(long word, int[] resultArray, int offset, int base) { + int to = offset + Long.bitCount(word); + int i = offset; + + for (; i < to - 3; i += 4) { + int ntz = Long.numberOfTrailingZeros(word); + resultArray[i] = base + ntz; + word ^= 1L << ntz; + ntz = Long.numberOfTrailingZeros(word); + resultArray[i] = base + ntz; + word ^= 1L << ntz; + ntz = Long.numberOfTrailingZeros(word); + resultArray[i] = base + ntz; + word ^= 1L << ntz; + ntz = Long.numberOfTrailingZeros(word); + resultArray[i] = base + ntz; + word ^= 1L << ntz; + } + + for (; i < to; i++) { + int ntz = Long.numberOfTrailingZeros(word); + resultArray[i] = base + ntz; + word ^= 1L << ntz; + } + + return to; + } + + private static int _dense(long word, int[] resultArray, int offset, int base) { + for (int i = 0; i < Long.SIZE; i++) { + if ((word & (1L << i)) != 0) { + resultArray[offset++] = base + i; + } + } + return offset; + } + + private static int _denseBranchLess(long word, int[] resultArray, int offset, int base) { + int lWord = (int) word; + int hWord = (int) (word >>> 32); + for (int i = 0; i < Integer.SIZE; i++) { + resultArray[offset] = base + i; + offset += lWord & 1; + lWord >>>= 1; + } + for (int i = Integer.SIZE; i < Long.SIZE; i++) { + resultArray[offset] = base + i; + offset += hWord & 1; + hWord >>>= 1; + } + return offset; Review Comment: Here is the result: (i'll try to check if it is really compiled to COMV after setting up the env on this new machine) ``` Benchmark (bitCount) Mode Cnt Score Error Units BitsetToArrayBenchmark.denseBranchLess 5 thrpt 5 13.582 ± 0.183 ops/us BitsetToArrayBenchmark.denseBranchLess 10 thrpt 5 13.485 ± 0.290 ops/us BitsetToArrayBenchmark.denseBranchLess 20 thrpt 5 13.575 ± 0.094 ops/us BitsetToArrayBenchmark.denseBranchLess 30 thrpt 5 13.570 ± 0.233 ops/us BitsetToArrayBenchmark.denseBranchLess 40 thrpt 5 13.478 ± 0.110 ops/us BitsetToArrayBenchmark.denseBranchLess 50 thrpt 5 13.577 ± 0.075 ops/us BitsetToArrayBenchmark.denseBranchLess 60 thrpt 5 13.558 ± 0.090 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 5 thrpt 5 8.501 ± 0.154 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 10 thrpt 5 6.373 ± 0.065 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 20 thrpt 5 11.544 ± 0.057 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 30 thrpt 5 11.538 ± 0.014 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 40 thrpt 5 11.475 ± 0.284 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 50 thrpt 5 11.546 ± 0.081 ops/us BitsetToArrayBenchmark.denseBranchLessCmov 60 thrpt 5 9.061 ± 0.213 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 5 thrpt 5 15.190 ± 0.505 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 10 thrpt 5 15.236 ± 0.358 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 20 thrpt 5 15.217 ± 0.324 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 30 thrpt 5 15.271 ± 0.217 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 40 thrpt 5 15.267 ± 0.406 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 50 thrpt 5 15.212 ± 0.345 ops/us BitsetToArrayBenchmark.denseBranchLessNoIdentityArray 60 thrpt 5 15.242 ± 0.308 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 5 thrpt 5 15.702 ± 0.483 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 10 thrpt 5 15.778 ± 0.275 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 20 thrpt 5 15.730 ± 0.468 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 30 thrpt 5 15.730 ± 1.251 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 40 thrpt 5 15.748 ± 0.429 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 50 thrpt 5 15.842 ± 0.213 ops/us BitsetToArrayBenchmark.denseBranchLessUnrolling 60 thrpt 5 15.817 ± 0.396 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 5 thrpt 5 14.905 ± 0.129 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 10 thrpt 5 14.922 ± 0.322 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 20 thrpt 5 14.828 ± 0.516 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 30 thrpt 5 14.686 ± 0.133 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 40 thrpt 5 14.881 ± 0.239 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 50 thrpt 5 14.978 ± 0.233 ops/us BitsetToArrayBenchmark.denseBranchLessVectorized 60 thrpt 5 14.991 ± 0.338 ops/us ``` -- 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.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
