/* * Copyright 2023 The original authors * * Licensed 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 dev.morling.onebrc; import java.io.File; import java.io.IOException; import java.io.RandomAccessFile; import java.nio.ByteBuffer; import java.nio.ByteOrder; import java.nio.channels.FileChannel; import java.nio.charset.StandardCharsets; import java.nio.file.Files; import java.nio.file.Path; import java.nio.file.StandardOpenOption; import java.util.ArrayList; import java.util.Arrays; import java.util.List; import java.util.TreeMap; import java.util.stream.Collectors; /** * Changelog: * * Initial submission: 62000 ms * Chunked reader: 16000 ms * Optimized parser: 13000 ms * Branchless methods: 11000 ms * Adding memory mapped files: 6500 ms (based on bjhara's submission) * Skipping string creation: 4700 ms * Custom hashmap... 4200 ms * Added SWAR token checks: 3900 ms * Skipped String creation: 3500 ms (idea from kgonia) * Improved String skip: 3250 ms * Segmenting files: 3150 ms (based on spullara's code) * Not using SWAR for EOL: 2850 ms * Inlining hash calculation: 2450 ms * * Best performing JVM on MacBook M2 Pro: 21.0.1-graal * `sdk use java 21.0.1-graal` * */ public class CalculateAverage_royvanrijn { private static final String FILE = "./measurements.txt"; // mutable state now instead of records, ugh, less instantiation. static final class Measurement { int min, max, count; long sum; public Measurement() { this.min = 1000; this.max = -1000; } public Measurement updateWith(int measurement) { min = min(min, measurement); max = max(max, measurement); sum += measurement; count++; return this; } public Measurement updateWith(Measurement measurement) { min = min(min, measurement.min); max = max(max, measurement.max); sum += measurement.sum; count += measurement.count; return this; } public String toString() { return round(min) + "/" + round((1.0 * sum) / count) + "/" + round(max); } private double round(double value) { return Math.round(value) / 10.0; } } public static void main(String[] args) throws Exception { new CalculateAverage_royvanrijn().run(); // new CalculateAverage_royvanrijn().runTests(); } private void testInput(final String inputString, final int start, final boolean bigEndian, final int[] expectedDelimiterAndHash, final long[] expectedCityNameLong) { byte[] input = inputString.getBytes(StandardCharsets.UTF_8); ByteBuffer buffer = ByteBuffer.wrap(input).order(bigEndian ? ByteOrder.BIG_ENDIAN : ByteOrder.LITTLE_ENDIAN); int[] output = new int[2]; long[] cityName = new long[128]; findNextDelimiterAndCalculateHash(buffer, SEPARATOR_PATTERN, start, buffer.limit(), output, cityName, bigEndian); if (!Arrays.equals(output, expectedDelimiterAndHash)) { System.out.println("Error in delimiter or hash"); System.out.println("Expected: " + Arrays.toString(expectedDelimiterAndHash)); System.out.println("Received: " + Arrays.toString(output)); } int amountLong = 1 + ((output[0] - start) >>> 3); if (!Arrays.equals(cityName, 0, amountLong, expectedCityNameLong, 0, amountLong)) { System.out.println("Error in long array"); System.out.println("Expected: " + Arrays.toString(expectedCityNameLong)); System.out.println("Received: " + Arrays.toString(cityName)); } } private void run() throws Exception { var results = getFileSegments(new File(FILE)).stream().map(segment -> { long segmentEnd = segment.end(); try (var fileChannel = (FileChannel) Files.newByteChannel(Path.of(FILE), StandardOpenOption.READ)) { var bb = fileChannel.map(FileChannel.MapMode.READ_ONLY, segment.start(), segmentEnd - segment.start()); // Work with any UTF-8 city name, up to 100 in length: var buffer = new byte[106]; // 100 + ; + -XX.X + \n var cityNameAsLongArray = new long[13]; // 13*8=104=kenough. var delimiterPointerAndHash = new int[2]; // Calculate using native ordering (fastest?): bb.order(ByteOrder.nativeOrder()); // Record the order it is and calculate accordingly: final boolean bufferIsBigEndian = bb.order().equals(ByteOrder.BIG_ENDIAN); MeasurementRepository measurements = new MeasurementRepository(); int startPointer; int limit = bb.limit(); while ((startPointer = bb.position()) < limit) { // SWAR method to find delimiter *and* record the cityname as long[] *and* calculate a hash: findNextDelimiterAndCalculateHash(bb, SEPARATOR_PATTERN, startPointer, limit, delimiterPointerAndHash, cityNameAsLongArray, bufferIsBigEndian); int delimiterPointer = delimiterPointerAndHash[0]; // Simple lookup is faster for '\n' (just three options) int endPointer; if (delimiterPointer >= limit) { bb.position(limit); // skip to next line. return measurements; } if (bb.get(delimiterPointer + 4) == '\n') { endPointer = delimiterPointer + 4; } else if (bb.get(delimiterPointer + 5) == '\n') { endPointer = delimiterPointer + 5; } else { endPointer = delimiterPointer + 6; } // Read the entry in a single get(): bb.get(buffer, 0, endPointer - startPointer); bb.position(endPointer + 1); // skip to next line. // Extract the measurement value (10x): final int cityNameLength = delimiterPointer - startPointer; final int measuredValueLength = endPointer - delimiterPointer - 1; final int measuredValue = branchlessParseInt(buffer, cityNameLength + 1, measuredValueLength); // Store everything in a custom hashtable: measurements.update(buffer, cityNameLength, delimiterPointerAndHash[1], cityNameAsLongArray).updateWith(measuredValue); } return measurements; } catch (IOException e) { throw new RuntimeException(e); } }).parallel() .flatMap(v -> v.values.stream()) .collect(Collectors.toMap(e -> e.cityName, MeasurementRepository.Entry::measurement, Measurement::updateWith, TreeMap::new)); System.out.println(results); } /** * -------- This section contains SWAR code (SIMD Within A Register) which processes a bytebuffer as longs to find values: */ private static final long SEPARATOR_PATTERN = compilePattern((byte) ';'); private static final long[] PARTIAL_INDEX_MASKS = new long[]{ 0L, 255L, 65535L, 16777215L, 4294967295L, 1099511627775L, 281474976710655L, 72057594037927935L }; public void runTests() { // Method used for debugging purposes, easy to make mistakes with all the bit hacking. // These all have the same hashes: testInput("Delft;-12.4", 0, true, new int[]{ 5, 1718384401 }, new long[]{ 499934586180L }); testInput("aDelft;-12.4", 1, true, new int[]{ 6, 1718384401 }, new long[]{ 499934586180L }); testInput("Delft;-12.4", 0, false, new int[]{ 5, 1718384401 }, new long[]{ 499934586180L }); testInput("aDelft;-12.4", 1, false, new int[]{ 6, 1718384401 }, new long[]{ 499934586180L }); testInput("Rotterdam;-12.4", 0, true, new int[]{ 9, -784321989 }, new long[]{ 7017859899421126482L, 109L }); testInput("abcdefghijklmnpoqrstuvwxyzRotterdam;-12.4", 26, true, new int[]{ 35, -784321989 }, new long[]{ 7017859899421126482L, 109L }); testInput("abcdefghijklmnpoqrstuvwxyzARotterdam;-12.4", 27, true, new int[]{ 36, -784321989 }, new long[]{ 7017859899421126482L, 109L }); testInput("Rotterdam;-12.4", 0, false, new int[]{ 9, -784321989 }, new long[]{ 7017859899421126482L, 109L }); testInput("abcdefghijklmnpoqrstuvwxyzRotterdam;-12.4", 26, false, new int[]{ 35, -784321989 }, new long[]{ 7017859899421126482L, 109L }); testInput("abcdefghijklmnpoqrstuvwxyzARotterdam;-12.4", 27, false, new int[]{ 36, -784321989 }, new long[]{ 7017859899421126482L, 109L }); // These have different hashes from the strings above: testInput("abcdefghijklmnpoqrstuvwxyzAROtterdam;-12.4", 27, true, new int[]{ 36, -792194501 }, new long[]{ 7017859899421118290L, 109L }); testInput("abcdefghijklmnpoqrstuvwxyzAROtterdam;-12.4", 27, false, new int[]{ 36, -792194501 }, new long[]{ 7017859899421118290L, 109L }); MeasurementRepository repository = new MeasurementRepository(); // Simulate adding two entries with the same hash: byte[] b1 = "City1;10.0".getBytes(); byte[] b2 = "City2;41.1".getBytes(); repository.update(b1, 5, 1234, new long[]{ 1234L }); repository.update(b2, 5, 1234, new long[]{ 4321L }); // And update the same record shouldn't add a third (this happened): repository.update(b1, 5, 1234, new long[]{ 1234L }); if (repository.values.size() != 2) { System.out.println("Error, should have two entries:"); System.out.println(repository.values); } MeasurementRepository.Entry firstInserted = repository.values.getFirst(); if (!firstInserted.cityName.equals("City1")) { System.out.println("Error, should have correct name: " + firstInserted.cityName); } } /** * Already looping the longs here, lets shoehorn in making a hash */ private void findNextDelimiterAndCalculateHash(final ByteBuffer bb, final long pattern, final int start, final int limit, final int[] output, final long[] asLong, final boolean bufferBigEndian) { int hash = 1; int i; int lCnt = 0; for (i = start; i <= limit - 8; i += 8) { long word = bb.getLong(i); if (bufferBigEndian) word = Long.reverseBytes(word); // Reversing the bytes is the cheapest way to do this int index = firstAnyPattern(word, pattern); if (index < Long.BYTES) { final long partialHash = word & PARTIAL_INDEX_MASKS[index]; asLong[lCnt] = partialHash; hash = 961 * hash + 31 * (int) (partialHash >>> 32) + (int) partialHash; output[0] = (i + index); output[1] = hash; return; } asLong[lCnt++] = word; hash = 961 * hash + 31 * (int) (word >>> 32) + (int) word; } // Handle remaining bytes long partialHash = 0; for (; i < limit; i++) { byte read; if ((read = bb.get(i)) == (byte) pattern) { asLong[lCnt] = partialHash; hash = 961 * hash + 31 * (int) (partialHash >>> 32) + (int) partialHash; output[0] = i; output[1] = hash; return; } partialHash = partialHash << 8 | read; } output[0] = limit; // delimiter not found output[1] = hash; } private static long compilePattern(final byte value) { return ((long) value << 56) | ((long) value << 48) | ((long) value << 40) | ((long) value << 32) | ((long) value << 24) | ((long) value << 16) | ((long) value << 8) | (long) value; } private static int firstAnyPattern(final long word, final long pattern) { final long match = word ^ pattern; long mask = match - 0x0101010101010101L; mask &= ~match; mask &= 0x8080808080808080L; return Long.numberOfTrailingZeros(mask) >> 3; } record FileSegment(long start, long end) { } /** Using this way to segment the file is much prettier, from spullara */ private static List<FileSegment> getFileSegments(final File file) throws IOException { final int numberOfSegments = Runtime.getRuntime().availableProcessors(); final long fileSize = file.length(); final long segmentSize = fileSize / numberOfSegments; final List<FileSegment> segments = new ArrayList<>(); if (segmentSize < 1000) { segments.add(new FileSegment(0, fileSize)); return segments; } try (RandomAccessFile randomAccessFile = new RandomAccessFile(file, "r")) { for (int i = 0; i < numberOfSegments; i++) { long segStart = i * segmentSize; long segEnd = (i == numberOfSegments - 1) ? fileSize : segStart + segmentSize; segStart = findSegment(i, 0, randomAccessFile, segStart, segEnd); segEnd = findSegment(i, numberOfSegments - 1, randomAccessFile, segEnd, fileSize); segments.add(new FileSegment(segStart, segEnd)); } } return segments; } private static long findSegment(final int i, final int skipSegment, RandomAccessFile raf, long location, final long fileSize) throws IOException { if (i != skipSegment) { raf.seek(location); while (location < fileSize) { location++; if (raf.read() == '\n') return location; } } return location; } /** * Branchless parser, goes from String to int (10x): * "-1.2" to -12 * "40.1" to 401 * etc. * * @param input * @return int value x10 */ private static int branchlessParseInt(final byte[] input, final int start, final int length) { // 0 if positive, 1 if negative final int negative = ~(input[start] >> 4) & 1; // 0 if nr length is 3, 1 if length is 4 final int has4 = ((length - negative) >> 2) & 1; final int digit1 = input[start + negative] - '0'; final int digit2 = input[start + negative + has4]; final int digit3 = input[start + negative + has4 + 2]; return (-negative ^ (has4 * (digit1 * 100) + digit2 * 10 + digit3 - 528) - negative); // 528 == ('0' * 10 + '0') } // branchless max (unprecise for large numbers, but good enough) static int max(final int a, final int b) { final int diff = a - b; final int dsgn = diff >> 31; return a - (diff & dsgn); } // branchless min (unprecise for large numbers, but good enough) static int min(final int a, final int b) { final int diff = a - b; final int dsgn = diff >> 31; return b + (diff & dsgn); } /** * A normal Java HashMap does all these safety things like boundary checks... we don't need that, we need speeeed. * * So I've written an extremely simple linear probing hashmap that should work well enough. */ class MeasurementRepository { private int size = 16384;// 16384; // Much larger than the number of cities, needs power of two private int[] indices = new int[size]; // Hashtable is just an int[] MeasurementRepository() { populateEmptyIndices(indices); } private void populateEmptyIndices(int[] array) { // Optimized fill with -1, fastest method: int len = array.length; array[0] = -1; // Value of i will be [1, 2, 4, 8, 16, 32, ..., len] for (int i = 1; i < len; i += i) { System.arraycopy(array, 0, array, i, i); } } private final List<Entry> values = new ArrayList<>(512); record Entry(int hash, long[] cityNameAsLong, String cityName, Measurement measurement) { @Override public String toString() { return cityName + "=" + measurement; } } public Measurement update(byte[] buffer, int length, int calculatedHash, long[] cityNameAsLongArray) { final int cityNameAsLongLength = 1 + (length >>> 3); // amount of longs that captures this cityname int hashtableIndex = (size - 1) & calculatedHash; int valueIndex; Entry retrievedEntry = null; while (true) { // search for the right spot if ((valueIndex = indices[hashtableIndex]) == -1) { break; // Empty slot found, stop the loop } else { // Non-empty slot, retrieve entry if ((retrievedEntry = values.get(valueIndex)).hash == calculatedHash && arrayEquals(retrievedEntry.cityNameAsLong, cityNameAsLongArray, cityNameAsLongLength)) { break; // Both hash and cityname match, stop the loop } } // Move to the next index hashtableIndex = (hashtableIndex + 1) % size; } if (valueIndex >= 0) { return retrievedEntry.measurement; } // --- This is a brand new entry, insert into the hashtable and do the extra calculations (once!) // Keep the already processed longs for fast equals: long[] cityNameAsLongArrayCopy = new long[cityNameAsLongLength]; System.arraycopy(cityNameAsLongArray, 0, cityNameAsLongArrayCopy, 0, cityNameAsLongLength); Entry toAdd = new Entry(calculatedHash, cityNameAsLongArrayCopy, new String(buffer, 0, length), new Measurement()); // Code to regrow (if we get more unique entries): (not needed/not optimized yet) // if (values.size() > size / 2) { // // We probably don't want this... // // int newSize = size << 1; // int[] newIndices = new int[newSize]; // populateEmptyIndices(newIndices); // for (int i = 0; i < values.size(); i++) { // Entry e = values.get(i); // int updatedIndex = (newSize - 1) & e.hash; // newIndices[updatedIndex] = i; // } // indices = newIndices; // size = newSize; // } indices[hashtableIndex] = values.size(); values.add(toAdd); return toAdd.measurement; } } /** * For case multiple hashes are equal (however unlikely) check the actual key (using longs) */ private boolean arrayEquals(final long[] a, final long[] b, final int length) { for (int i = 0; i < length; i++) { if (a[i] != b[i]) return false; } return true; } }