/* * Copyright 2019 Google Inc. All Rights Reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. */ package org.openjdk.bench.java.lang; import org.openjdk.jmh.annotations.Benchmark; import org.openjdk.jmh.annotations.BenchmarkMode; 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 java.util.Arrays; import java.util.concurrent.ThreadLocalRandom; import java.util.concurrent.TimeUnit; /** * Explores the cost of copying the contents of one array to another, as is * commonly seen in collection classes that need to resize their backing * array, like ArrayList. * * We have multiple variations on copying, and we explore the cost of * clearing the old array, which might help for generational GCs. * * Benchmarks the operations in the ancient * JDK-6428387: array clone() much slower than Arrays.copyOf * * 2019 results on x86: * * The "simple" benchmarks below have the same performance, except that * simple_copyLoop is surprisingly 5x slower. The array copying intrinsics * are very effective and a naive loop does not get optimized the same way. * OTOH there is no intrinsic for Arrays.fill but the naive array zeroing loop * *does* get optimized to something a little faster than than arraycopy. * * System.arraycopy and Arrays.fill have such outstanding performance that * one should use them to replace handwritten loops whenever possible. * * This benchmark is great for measuring cache effects, e.g. size=10^6 has 5x * the per-element cost of size=10^3 (See "The Myth of RAM".) * * (cd $(hg root) && for size in 3 16 999 999999; do make test TEST="micro:java.lang.ArrayFiddle" MICRO="FORK=2;WARMUP_ITER=4;ITER=4;OPTIONS=-opi $size -p size=$size" |& perl -ne 'print if /^Benchmark/ .. /^Finished running test/'; done) */ @BenchmarkMode(Mode.AverageTime) @OutputTimeUnit(TimeUnit.NANOSECONDS) @State(Scope.Benchmark) public class ArrayFiddle { @Param("999") public int size; public int largerSize; public Object[] data; public Object[] copy; @Setup public void setup() { largerSize = size + (size >> 1); data = new Object[size]; ThreadLocalRandom rnd = ThreadLocalRandom.current(); for (int i = data.length; i--> 0; ) data[i] = rnd.nextInt(256); copy = data.clone(); } // --- "simple" benchmarks just make an array clone @Benchmark public Object[] simple_clone() { return data.clone(); } @Benchmark public Object[] simple_copyOf() { return Arrays.copyOf(data, data.length); } @Benchmark public Object[] simple_arraycopy() { Object[] out = new Object[data.length]; System.arraycopy(data, 0, out, 0, data.length); return out; } @Benchmark public Object[] simple_copyLoop() { final Object[] data = this.data; int len = data.length; Object[] out = new Object[len]; for (int i = 0; i < len; i++) out[i] = data[i]; return out; } // --- "grow" benchmarks have an output array that is larger private Object[] input_array() { System.arraycopy(data, 0, copy, 0, size); return copy; } @Benchmark public Object[] grow_copyLoop() { Object[] in = input_array(); Object[] out = new Object[largerSize]; for (int i = 0, len = in.length; i < len; i++) out[i] = in[i]; return out; } @Benchmark public Object[] grow_copyZeroLoop() { Object[] in = input_array(); Object[] out = new Object[largerSize]; for (int i = 0, len = in.length; i < len; i++) { out[i] = in[i]; in[i] = null; } return out; } @Benchmark public Object[] grow_arraycopy() { Object[] in = input_array(); Object[] out = new Object[largerSize]; System.arraycopy(in, 0, out, 0, size); return out; } @Benchmark public Object[] grow_arraycopy_fill() { Object[] in = input_array(); Object[] out = new Object[largerSize]; System.arraycopy(in, 0, out, 0, size); Arrays.fill(in, null); return out; } @Benchmark public Object[] grow_arraycopy_zeroLoop() { Object[] in = input_array(); Object[] out = new Object[largerSize]; System.arraycopy(in, 0, out, 0, size); for (int i = 0, len = in.length; i < len; i++) in[i] = null; return out; } @Benchmark public Object[] grow_copyOf() { Object[] in = input_array(); Object[] out = Arrays.copyOf(in, largerSize); return out; } @Benchmark public Object[] grow_copyOf_fill() { Object[] in = input_array(); Object[] out = Arrays.copyOf(in, largerSize); Arrays.fill(in, null); return out; } @Benchmark public Object[] grow_copyOf_zeroLoop() { Object[] in = input_array(); Object[] out = Arrays.copyOf(in, largerSize); for (int i = 0, len = in.length; i < len; i++) in[i] = null; return out; } }