Back to index...

	/*
	* Copyright (c) 2015, 2017 Oracle and/or its affiliates. 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. Oracle designates this
	* particular file as subject to the "Classpath" exception as provided
	* by Oracle in the LICENSE file that accompanied this code.
	*
	* 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 jdk.internal.util;

	import jdk.internal.HotSpotIntrinsicCandidate;
	import jdk.internal.misc.Unsafe;

	/**
	* Utility methods to find a mismatch between two primitive arrays.
	*
	* <p>Array equality and lexicographical comparison can be built on top of
	* array mismatch functionality.
	*
	* <p>The mismatch method implementation, {@link #vectorizedMismatch}, leverages
	* vector-based techniques to access and compare the contents of two arrays.
	* The Java implementation uses {@code Unsafe.getLongUnaligned} to access the
	* content of an array, thus access is supported on platforms that do not
	* support unaligned access. For a byte[] array, 8 bytes (64 bits) can be
	* accessed and compared as a unit rather than individually, which increases
	* the performance when the method is compiled by the HotSpot VM. On supported
	* platforms the mismatch implementation is intrinsified to leverage SIMD
	* instructions. So for a byte[] array, 16 bytes (128 bits), 32 bytes
	* (256 bits), and perhaps in the future even 64 bytes (512 bits), platform
	* permitting, can be accessed and compared as a unit, which further increases
	* the performance over the Java implementation.
	*
	* <p>None of the mismatch methods perform array bounds checks. It is the
	* responsibility of the caller (direct or otherwise) to perform such checks
	* before calling this method.
	*/
	public class ArraysSupport {
	static final Unsafe U = Unsafe.getUnsafe();

	private static final boolean BIG_ENDIAN = U.isBigEndian();

	public static final int LOG2_ARRAY_BOOLEAN_INDEX_SCALE = exactLog2(Unsafe.ARRAY_BOOLEAN_INDEX_SCALE);
	public static final int LOG2_ARRAY_BYTE_INDEX_SCALE = exactLog2(Unsafe.ARRAY_BYTE_INDEX_SCALE);
	public static final int LOG2_ARRAY_CHAR_INDEX_SCALE = exactLog2(Unsafe.ARRAY_CHAR_INDEX_SCALE);
	public static final int LOG2_ARRAY_SHORT_INDEX_SCALE = exactLog2(Unsafe.ARRAY_SHORT_INDEX_SCALE);
	public static final int LOG2_ARRAY_INT_INDEX_SCALE = exactLog2(Unsafe.ARRAY_INT_INDEX_SCALE);
	public static final int LOG2_ARRAY_LONG_INDEX_SCALE = exactLog2(Unsafe.ARRAY_LONG_INDEX_SCALE);
	public static final int LOG2_ARRAY_FLOAT_INDEX_SCALE = exactLog2(Unsafe.ARRAY_FLOAT_INDEX_SCALE);
	public static final int LOG2_ARRAY_DOUBLE_INDEX_SCALE = exactLog2(Unsafe.ARRAY_DOUBLE_INDEX_SCALE);

	private static final int LOG2_BYTE_BIT_SIZE = exactLog2(Byte.SIZE);

	private static int exactLog2(int scale) {
	if ((scale & (scale - 1)) != 0)
	throw new Error("data type scale not a power of two");
	return Integer.numberOfTrailingZeros(scale);
	}

	private ArraysSupport() {}

	/**
	* Find the relative index of the first mismatching pair of elements in two
	* primitive arrays of the same component type. Pairs of elements will be
	* tested in order relative to given offsets into both arrays.
	*
	* <p>This method does not perform type checks or bounds checks. It is the
	* responsibility of the caller to perform such checks before calling this
	* method.
	*
	* <p>The given offsets, in bytes, need not be aligned according to the
	* given log<sub>2</sub> size the array elements. More specifically, an
	* offset modulus the size need not be zero.
	*
	* @param a the first array to be tested for mismatch, or {@code null} for
	* direct memory access
	* @param aOffset the relative offset, in bytes, from the base address of
	* the first array to test from, otherwise if the first array is
	* {@code null}, an absolute address pointing to the first element to test.
	* @param b the second array to be tested for mismatch, or {@code null} for
	* direct memory access
	* @param bOffset the relative offset, in bytes, from the base address of
	* the second array to test from, otherwise if the second array is
	* {@code null}, an absolute address pointing to the first element to test.
	* @param length the number of array elements to test
	* @param log2ArrayIndexScale log<sub>2</sub> of the array index scale, that
	* corresponds to the size, in bytes, of an array element.
	* @return if a mismatch is found a relative index, between 0 (inclusive)
	* and {@code length} (exclusive), of the first mismatching pair of elements
	* in the two arrays. Otherwise, if a mismatch is not found the bitwise
	* compliment of the number of remaining pairs of elements to be checked in
	* the tail of the two arrays.
	*/
	@HotSpotIntrinsicCandidate
	public static int vectorizedMismatch(Object a, long aOffset,
	Object b, long bOffset,
	int length,
	int log2ArrayIndexScale) {
	// assert a.getClass().isArray();
	// assert b.getClass().isArray();
	// assert 0 <= length <= sizeOf(a)
	// assert 0 <= length <= sizeOf(b)
	// assert 0 <= log2ArrayIndexScale <= 3

	int log2ValuesPerWidth = LOG2_ARRAY_LONG_INDEX_SCALE - log2ArrayIndexScale;
	int wi = 0;
	for (; wi < length >> log2ValuesPerWidth; wi++) {
	long bi = ((long) wi) << LOG2_ARRAY_LONG_INDEX_SCALE;
	long av = U.getLongUnaligned(a, aOffset + bi);
	long bv = U.getLongUnaligned(b, bOffset + bi);
	if (av != bv) {
	long x = av ^ bv;
	int o = BIG_ENDIAN
	? Long.numberOfLeadingZeros(x) >> (LOG2_BYTE_BIT_SIZE + log2ArrayIndexScale)
	: Long.numberOfTrailingZeros(x) >> (LOG2_BYTE_BIT_SIZE + log2ArrayIndexScale);
	return (wi << log2ValuesPerWidth) + o;
	}
	}

	// Calculate the tail of remaining elements to check
	int tail = length - (wi << log2ValuesPerWidth);

	if (log2ArrayIndexScale < LOG2_ARRAY_INT_INDEX_SCALE) {
	int wordTail = 1 << (LOG2_ARRAY_INT_INDEX_SCALE - log2ArrayIndexScale);
	// Handle 4 bytes or 2 chars in the tail using int width
	if (tail >= wordTail) {
	long bi = ((long) wi) << LOG2_ARRAY_LONG_INDEX_SCALE;
	int av = U.getIntUnaligned(a, aOffset + bi);
	int bv = U.getIntUnaligned(b, bOffset + bi);
	if (av != bv) {
	int x = av ^ bv;
	int o = BIG_ENDIAN
	? Integer.numberOfLeadingZeros(x) >> (LOG2_BYTE_BIT_SIZE + log2ArrayIndexScale)
	: Integer.numberOfTrailingZeros(x) >> (LOG2_BYTE_BIT_SIZE + log2ArrayIndexScale);
	return (wi << log2ValuesPerWidth) + o;
	}
	tail -= wordTail;
	}
	return ~tail;
	}
	else {
	return ~tail;
	}
	}

	// Booleans
	// Each boolean element takes up one byte

	public static int mismatch(boolean[] a,
	boolean[] b,
	int length) {
	int i = 0;
	if (length > 7) {
	if (a[0] != b[0])
	return 0;
	i = vectorizedMismatch(
	a, Unsafe.ARRAY_BOOLEAN_BASE_OFFSET,
	b, Unsafe.ARRAY_BOOLEAN_BASE_OFFSET,
	length, LOG2_ARRAY_BOOLEAN_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[i] != b[i])
	return i;
	}
	return -1;
	}

	public static int mismatch(boolean[] a, int aFromIndex,
	boolean[] b, int bFromIndex,
	int length) {
	int i = 0;
	if (length > 7) {
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_BOOLEAN_BASE_OFFSET + aFromIndex;
	int bOffset = Unsafe.ARRAY_BOOLEAN_BASE_OFFSET + bFromIndex;
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_BOOLEAN_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[aFromIndex + i] != b[bFromIndex + i])
	return i;
	}
	return -1;
	}


	// Bytes

	/**
	* Find the index of a mismatch between two arrays.
	*
	* <p>This method does not perform bounds checks. It is the responsibility
	* of the caller to perform such bounds checks before calling this method.
	*
	* @param a the first array to be tested for a mismatch
	* @param b the second array to be tested for a mismatch
	* @param length the number of bytes from each array to check
	* @return the index of a mismatch between the two arrays, otherwise -1 if
	* no mismatch. The index will be within the range of (inclusive) 0 to
	* (exclusive) the smaller of the two array lengths.
	*/
	public static int mismatch(byte[] a,
	byte[] b,
	int length) {
	// ISSUE: defer to index receiving methods if performance is good
	// assert length <= a.length
	// assert length <= b.length

	int i = 0;
	if (length > 7) {
	if (a[0] != b[0])
	return 0;
	i = vectorizedMismatch(
	a, Unsafe.ARRAY_BYTE_BASE_OFFSET,
	b, Unsafe.ARRAY_BYTE_BASE_OFFSET,
	length, LOG2_ARRAY_BYTE_INDEX_SCALE);
	if (i >= 0)
	return i;
	// Align to tail
	i = length - ~i;
	// assert i >= 0 && i <= 7;
	}
	// Tail < 8 bytes
	for (; i < length; i++) {
	if (a[i] != b[i])
	return i;
	}
	return -1;
	}

	/**
	* Find the relative index of a mismatch between two arrays starting from
	* given indexes.
	*
	* <p>This method does not perform bounds checks. It is the responsibility
	* of the caller to perform such bounds checks before calling this method.
	*
	* @param a the first array to be tested for a mismatch
	* @param aFromIndex the index of the first element (inclusive) in the first
	* array to be compared
	* @param b the second array to be tested for a mismatch
	* @param bFromIndex the index of the first element (inclusive) in the
	* second array to be compared
	* @param length the number of bytes from each array to check
	* @return the relative index of a mismatch between the two arrays,
	* otherwise -1 if no mismatch. The index will be within the range of
	* (inclusive) 0 to (exclusive) the smaller of the two array bounds.
	*/
	public static int mismatch(byte[] a, int aFromIndex,
	byte[] b, int bFromIndex,
	int length) {
	// assert 0 <= aFromIndex < a.length
	// assert 0 <= aFromIndex + length <= a.length
	// assert 0 <= bFromIndex < b.length
	// assert 0 <= bFromIndex + length <= b.length
	// assert length >= 0

	int i = 0;
	if (length > 7) {
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_BYTE_BASE_OFFSET + aFromIndex;
	int bOffset = Unsafe.ARRAY_BYTE_BASE_OFFSET + bFromIndex;
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_BYTE_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[aFromIndex + i] != b[bFromIndex + i])
	return i;
	}
	return -1;
	}


	// Chars

	public static int mismatch(char[] a,
	char[] b,
	int length) {
	int i = 0;
	if (length > 3) {
	if (a[0] != b[0])
	return 0;
	i = vectorizedMismatch(
	a, Unsafe.ARRAY_CHAR_BASE_OFFSET,
	b, Unsafe.ARRAY_CHAR_BASE_OFFSET,
	length, LOG2_ARRAY_CHAR_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[i] != b[i])
	return i;
	}
	return -1;
	}

	public static int mismatch(char[] a, int aFromIndex,
	char[] b, int bFromIndex,
	int length) {
	int i = 0;
	if (length > 3) {
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_CHAR_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_CHAR_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_CHAR_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_CHAR_INDEX_SCALE);
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_CHAR_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[aFromIndex + i] != b[bFromIndex + i])
	return i;
	}
	return -1;
	}


	// Shorts

	public static int mismatch(short[] a,
	short[] b,
	int length) {
	int i = 0;
	if (length > 3) {
	if (a[0] != b[0])
	return 0;
	i = vectorizedMismatch(
	a, Unsafe.ARRAY_SHORT_BASE_OFFSET,
	b, Unsafe.ARRAY_SHORT_BASE_OFFSET,
	length, LOG2_ARRAY_SHORT_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[i] != b[i])
	return i;
	}
	return -1;
	}

	public static int mismatch(short[] a, int aFromIndex,
	short[] b, int bFromIndex,
	int length) {
	int i = 0;
	if (length > 3) {
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_SHORT_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_SHORT_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_SHORT_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_SHORT_INDEX_SCALE);
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_SHORT_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[aFromIndex + i] != b[bFromIndex + i])
	return i;
	}
	return -1;
	}


	// Ints

	public static int mismatch(int[] a,
	int[] b,
	int length) {
	int i = 0;
	if (length > 1) {
	if (a[0] != b[0])
	return 0;
	i = vectorizedMismatch(
	a, Unsafe.ARRAY_INT_BASE_OFFSET,
	b, Unsafe.ARRAY_INT_BASE_OFFSET,
	length, LOG2_ARRAY_INT_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[i] != b[i])
	return i;
	}
	return -1;
	}

	public static int mismatch(int[] a, int aFromIndex,
	int[] b, int bFromIndex,
	int length) {
	int i = 0;
	if (length > 1) {
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_INT_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_INT_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_INT_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_INT_INDEX_SCALE);
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_INT_INDEX_SCALE);
	if (i >= 0)
	return i;
	i = length - ~i;
	}
	for (; i < length; i++) {
	if (a[aFromIndex + i] != b[bFromIndex + i])
	return i;
	}
	return -1;
	}


	// Floats

	public static int mismatch(float[] a,
	float[] b,
	int length) {
	return mismatch(a, 0, b, 0, length);
	}

	public static int mismatch(float[] a, int aFromIndex,
	float[] b, int bFromIndex,
	int length) {
	int i = 0;
	if (length > 1) {
	if (Float.floatToRawIntBits(a[aFromIndex]) == Float.floatToRawIntBits(b[bFromIndex])) {
	int aOffset = Unsafe.ARRAY_FLOAT_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_FLOAT_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_FLOAT_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_FLOAT_INDEX_SCALE);
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_FLOAT_INDEX_SCALE);
	}
	// Mismatched
	if (i >= 0) {
	// Check if mismatch is not associated with two NaN values
	if (!Float.isNaN(a[aFromIndex + i]) \|\| !Float.isNaN(b[bFromIndex + i]))
	return i;

	// Mismatch on two different NaN values that are normalized to match
	// Fall back to slow mechanism
	// ISSUE: Consider looping over vectorizedMismatch adjusting ranges
	// However, requires that returned value be relative to input ranges
	i++;
	}
	// Matched
	else {
	i = length - ~i;
	}
	}
	for (; i < length; i++) {
	if (Float.floatToIntBits(a[aFromIndex + i]) != Float.floatToIntBits(b[bFromIndex + i]))
	return i;
	}
	return -1;
	}

	// 64 bit sizes

	// Long

	public static int mismatch(long[] a,
	long[] b,
	int length) {
	if (length == 0) {
	return -1;
	}
	if (a[0] != b[0])
	return 0;
	int i = vectorizedMismatch(
	a, Unsafe.ARRAY_LONG_BASE_OFFSET,
	b, Unsafe.ARRAY_LONG_BASE_OFFSET,
	length, LOG2_ARRAY_LONG_INDEX_SCALE);
	return i >= 0 ? i : -1;
	}

	public static int mismatch(long[] a, int aFromIndex,
	long[] b, int bFromIndex,
	int length) {
	if (length == 0) {
	return -1;
	}
	if (a[aFromIndex] != b[bFromIndex])
	return 0;
	int aOffset = Unsafe.ARRAY_LONG_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_LONG_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_LONG_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_LONG_INDEX_SCALE);
	int i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_LONG_INDEX_SCALE);
	return i >= 0 ? i : -1;
	}


	// Double

	public static int mismatch(double[] a,
	double[] b,
	int length) {
	return mismatch(a, 0, b, 0, length);
	}

	public static int mismatch(double[] a, int aFromIndex,
	double[] b, int bFromIndex,
	int length) {
	if (length == 0) {
	return -1;
	}
	int i = 0;
	if (Double.doubleToRawLongBits(a[aFromIndex]) == Double.doubleToRawLongBits(b[bFromIndex])) {
	int aOffset = Unsafe.ARRAY_DOUBLE_BASE_OFFSET + (aFromIndex << LOG2_ARRAY_DOUBLE_INDEX_SCALE);
	int bOffset = Unsafe.ARRAY_DOUBLE_BASE_OFFSET + (bFromIndex << LOG2_ARRAY_DOUBLE_INDEX_SCALE);
	i = vectorizedMismatch(
	a, aOffset,
	b, bOffset,
	length, LOG2_ARRAY_DOUBLE_INDEX_SCALE);
	}
	if (i >= 0) {
	// Check if mismatch is not associated with two NaN values
	if (!Double.isNaN(a[aFromIndex + i]) \|\| !Double.isNaN(b[bFromIndex + i]))
	return i;

	// Mismatch on two different NaN values that are normalized to match
	// Fall back to slow mechanism
	// ISSUE: Consider looping over vectorizedMismatch adjusting ranges
	// However, requires that returned value be relative to input ranges
	i++;
	for (; i < length; i++) {
	if (Double.doubleToLongBits(a[aFromIndex + i]) != Double.doubleToLongBits(b[bFromIndex + i]))
	return i;
	}
	}

	return -1;
	}
	}

Back to index...