All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.jctools.queues.MpUnboundedXaddArrayQueue Maven / Gradle / Ivy

There is a newer version: 1.36.0
Show newest version
package org.jctools.queues;

import org.jctools.queues.IndexedQueueSizeUtil.IndexedQueue;
import org.jctools.util.PortableJvmInfo;
import org.jctools.util.Pow2;
import org.jctools.util.UnsafeAccess;

import java.util.AbstractQueue;
import java.util.Iterator;

import static org.jctools.queues.MpUnboundedXaddChunk.NOT_USED;
import static org.jctools.util.UnsafeAccess.UNSAFE;
import static org.jctools.util.UnsafeAccess.fieldOffset;

abstract class MpUnboundedXaddArrayQueuePad1 extends AbstractQueue implements IndexedQueue
{
    byte b000,b001,b002,b003,b004,b005,b006,b007;//  8b
    byte b010,b011,b012,b013,b014,b015,b016,b017;// 16b
    byte b020,b021,b022,b023,b024,b025,b026,b027;// 24b
    byte b030,b031,b032,b033,b034,b035,b036,b037;// 32b
    byte b040,b041,b042,b043,b044,b045,b046,b047;// 40b
    byte b050,b051,b052,b053,b054,b055,b056,b057;// 48b
    byte b060,b061,b062,b063,b064,b065,b066,b067;// 56b
    byte b070,b071,b072,b073,b074,b075,b076,b077;// 64b
    byte b100,b101,b102,b103,b104,b105,b106,b107;// 72b
    byte b110,b111,b112,b113,b114,b115,b116,b117;// 80b
    byte b120,b121,b122,b123,b124,b125,b126,b127;// 88b
    byte b130,b131,b132,b133,b134,b135,b136,b137;// 96b
    byte b140,b141,b142,b143,b144,b145,b146,b147;//104b
    byte b150,b151,b152,b153,b154,b155,b156,b157;//112b
    byte b160,b161,b162,b163,b164,b165,b166,b167;//120b
    byte b170,b171,b172,b173,b174,b175,b176,b177;//128b
}

// $gen:ordered-fields
abstract class MpUnboundedXaddArrayQueueProducerFields extends MpUnboundedXaddArrayQueuePad1
{
    private final static long P_INDEX_OFFSET =
        fieldOffset(MpUnboundedXaddArrayQueueProducerFields.class, "producerIndex");
    private volatile long producerIndex;

    @Override
    public final long lvProducerIndex()
    {
        return producerIndex;
    }

    final long getAndIncrementProducerIndex()
    {
        return UNSAFE.getAndAddLong(this, P_INDEX_OFFSET, 1);
    }

    final long getAndAddProducerIndex(long delta)
    {
        return UNSAFE.getAndAddLong(this, P_INDEX_OFFSET, delta);
    }
}

abstract class MpUnboundedXaddArrayQueuePad2 extends MpUnboundedXaddArrayQueueProducerFields
{
    byte b000,b001,b002,b003,b004,b005,b006,b007;//  8b
    byte b010,b011,b012,b013,b014,b015,b016,b017;// 16b
    byte b020,b021,b022,b023,b024,b025,b026,b027;// 24b
    byte b030,b031,b032,b033,b034,b035,b036,b037;// 32b
    byte b040,b041,b042,b043,b044,b045,b046,b047;// 40b
    byte b050,b051,b052,b053,b054,b055,b056,b057;// 48b
    byte b060,b061,b062,b063,b064,b065,b066,b067;// 56b
    byte b070,b071,b072,b073,b074,b075,b076,b077;// 64b
    byte b100,b101,b102,b103,b104,b105,b106,b107;// 72b
    byte b110,b111,b112,b113,b114,b115,b116,b117;// 80b
    byte b120,b121,b122,b123,b124,b125,b126,b127;// 88b
    byte b130,b131,b132,b133,b134,b135,b136,b137;// 96b
    byte b140,b141,b142,b143,b144,b145,b146,b147;//104b
    byte b150,b151,b152,b153,b154,b155,b156,b157;//112b
    byte b160,b161,b162,b163,b164,b165,b166,b167;//120b
    // byte b170,b171,b172,b173,b174,b175,b176,b177;//128b
}

// $gen:ordered-fields
abstract class MpUnboundedXaddArrayQueueProducerChunk, E>
    extends MpUnboundedXaddArrayQueuePad2
{
    private static final long P_CHUNK_OFFSET =
        fieldOffset(MpUnboundedXaddArrayQueueProducerChunk.class, "producerChunk");
    private static final long P_CHUNK_INDEX_OFFSET =
        fieldOffset(MpUnboundedXaddArrayQueueProducerChunk.class, "producerChunkIndex");

    private volatile R producerChunk;
    private volatile long producerChunkIndex;


    final long lvProducerChunkIndex()
    {
        return producerChunkIndex;
    }

    final boolean casProducerChunkIndex(long expected, long value)
    {
        return UNSAFE.compareAndSwapLong(this, P_CHUNK_INDEX_OFFSET, expected, value);
    }

    final void soProducerChunkIndex(long value)
    {
        UNSAFE.putOrderedLong(this, P_CHUNK_INDEX_OFFSET, value);
    }

    final R lvProducerChunk()
    {
        return this.producerChunk;
    }

    final void soProducerChunk(R chunk)
    {
        UNSAFE.putOrderedObject(this, P_CHUNK_OFFSET, chunk);
    }
}

abstract class MpUnboundedXaddArrayQueuePad3, E>
    extends MpUnboundedXaddArrayQueueProducerChunk
{
    byte b000,b001,b002,b003,b004,b005,b006,b007;//  8b
    byte b010,b011,b012,b013,b014,b015,b016,b017;// 16b
    byte b020,b021,b022,b023,b024,b025,b026,b027;// 24b
    byte b030,b031,b032,b033,b034,b035,b036,b037;// 32b
    byte b040,b041,b042,b043,b044,b045,b046,b047;// 40b
    byte b050,b051,b052,b053,b054,b055,b056,b057;// 48b
    byte b060,b061,b062,b063,b064,b065,b066,b067;// 56b
    byte b070,b071,b072,b073,b074,b075,b076,b077;// 64b
    byte b100,b101,b102,b103,b104,b105,b106,b107;// 72b
    byte b110,b111,b112,b113,b114,b115,b116,b117;// 80b
    byte b120,b121,b122,b123,b124,b125,b126,b127;// 88b
    byte b130,b131,b132,b133,b134,b135,b136,b137;// 96b
    byte b140,b141,b142,b143,b144,b145,b146,b147;//104b
    byte b150,b151,b152,b153,b154,b155,b156,b157;//112b
    byte b160,b161,b162,b163,b164,b165,b166,b167;//120b
    // byte b170,b171,b172,b173,b174,b175,b176,b177;//128b
}

// $gen:ordered-fields
abstract class MpUnboundedXaddArrayQueueConsumerFields, E>
    extends MpUnboundedXaddArrayQueuePad3
{
    private final static long C_INDEX_OFFSET =
        fieldOffset(MpUnboundedXaddArrayQueueConsumerFields.class, "consumerIndex");
    private final static long C_CHUNK_OFFSET =
        fieldOffset(MpUnboundedXaddArrayQueueConsumerFields.class, "consumerChunk");

    private volatile long consumerIndex;
    private volatile R consumerChunk;

    @Override
    public final long lvConsumerIndex()
    {
        return consumerIndex;
    }

    final boolean casConsumerIndex(long expect, long newValue)
    {
        return UNSAFE.compareAndSwapLong(this, C_INDEX_OFFSET, expect, newValue);
    }

    final R lpConsumerChunk()
    {
        return (R) UNSAFE.getObject(this, C_CHUNK_OFFSET);
    }

    final R lvConsumerChunk()
    {
        return this.consumerChunk;
    }

    final void soConsumerChunk(R newValue)
    {
        UNSAFE.putOrderedObject(this, C_CHUNK_OFFSET, newValue);
    }

    final long lpConsumerIndex()
    {
        return UNSAFE.getLong(this, C_INDEX_OFFSET);
    }

    final void soConsumerIndex(long newValue)
    {
        UNSAFE.putOrderedLong(this, C_INDEX_OFFSET, newValue);
    }
}

abstract class MpUnboundedXaddArrayQueuePad5, E>
    extends MpUnboundedXaddArrayQueueConsumerFields
{
    byte b000,b001,b002,b003,b004,b005,b006,b007;//  8b
    byte b010,b011,b012,b013,b014,b015,b016,b017;// 16b
    byte b020,b021,b022,b023,b024,b025,b026,b027;// 24b
    byte b030,b031,b032,b033,b034,b035,b036,b037;// 32b
    byte b040,b041,b042,b043,b044,b045,b046,b047;// 40b
    byte b050,b051,b052,b053,b054,b055,b056,b057;// 48b
    byte b060,b061,b062,b063,b064,b065,b066,b067;// 56b
    byte b070,b071,b072,b073,b074,b075,b076,b077;// 64b
    byte b100,b101,b102,b103,b104,b105,b106,b107;// 72b
    byte b110,b111,b112,b113,b114,b115,b116,b117;// 80b
    byte b120,b121,b122,b123,b124,b125,b126,b127;// 88b
    byte b130,b131,b132,b133,b134,b135,b136,b137;// 96b
    byte b140,b141,b142,b143,b144,b145,b146,b147;//104b
    byte b150,b151,b152,b153,b154,b155,b156,b157;//112b
    byte b160,b161,b162,b163,b164,b165,b166,b167;//120b
    // byte b170,b171,b172,b173,b174,b175,b176,b177;//128b
}

/**
 * Common infrastructure for the XADD queues.
 *
 * @author https://github.com/franz1981
 */
abstract class MpUnboundedXaddArrayQueue, E>
    extends MpUnboundedXaddArrayQueuePad5
    implements MessagePassingQueue, QueueProgressIndicators
{
    // it must be != MpUnboundedXaddChunk.NOT_USED
    private static final long ROTATION = -2;
    final int chunkMask;
    final int chunkShift;
    final int maxPooledChunks;
    final SpscArrayQueue freeChunksPool;

    /**
     * @param chunkSize The buffer size to be used in each chunk of this queue
     * @param maxPooledChunks The maximum number of reused chunks kept around to avoid allocation, chunks are pre-allocated
     */
    MpUnboundedXaddArrayQueue(int chunkSize, int maxPooledChunks)
    {
        if (!UnsafeAccess.SUPPORTS_GET_AND_ADD_LONG)
        {
            throw new IllegalStateException("Unsafe::getAndAddLong support (JDK 8+) is required for this queue to work");
        }
        if (maxPooledChunks < 0)
        {
            throw new IllegalArgumentException("Expecting a positive maxPooledChunks, but got:"+maxPooledChunks);
        }
        chunkSize = Pow2.roundToPowerOfTwo(chunkSize);

        this.chunkMask = chunkSize - 1;
        this.chunkShift = Integer.numberOfTrailingZeros(chunkSize);
        freeChunksPool = new SpscArrayQueue(maxPooledChunks);

        final R first = newChunk(0, null, chunkSize, maxPooledChunks > 0);
        soProducerChunk(first);
        soProducerChunkIndex(0);
        soConsumerChunk(first);
        for (int i = 1; i < maxPooledChunks; i++)
        {
            freeChunksPool.offer(newChunk(NOT_USED, null, chunkSize, true));
        }
        this.maxPooledChunks = maxPooledChunks;
    }

    public final int chunkSize()
    {
        return chunkMask + 1;
    }

    public final int maxPooledChunks()
    {
        return maxPooledChunks;
    }

    abstract R newChunk(long index, R prev, int chunkSize, boolean pooled);

    @Override
    public long currentProducerIndex()
    {
        return lvProducerIndex();
    }

    @Override
    public long currentConsumerIndex()
    {
        return lvConsumerIndex();
    }

    /**
     * We're here because currentChunk.index doesn't match the expectedChunkIndex. To resolve we must now chase the linked
     * chunks to the appropriate chunk. More than one producer may end up racing to add or discover new chunks.
     *
     * @param initialChunk the starting point chunk, which does not match the required chunk index
     * @param requiredChunkIndex the chunk index we need
     * @return the chunk matching the required index
     */
    final R producerChunkForIndex(
        final R initialChunk,
        final long requiredChunkIndex)
    {
        R currentChunk = initialChunk;
        long jumpBackward;
        while (true)
        {
            if (currentChunk == null)
            {
                currentChunk = lvProducerChunk();
            }
            final long currentChunkIndex = currentChunk.lvIndex();
            assert currentChunkIndex != NOT_USED;
            // if the required chunk index is less than the current chunk index then we need to walk the linked list of
            // chunks back to the required index
            jumpBackward = currentChunkIndex - requiredChunkIndex;
            if (jumpBackward >= 0)
            {
                break;
            }
            // try validate against the last producer chunk index
            if (lvProducerChunkIndex() == currentChunkIndex)
            {
                currentChunk = appendNextChunks(currentChunk, currentChunkIndex, -jumpBackward);
            }
            else
            {
                currentChunk = null;
            }
        }
        for (long i = 0; i < jumpBackward; i++)
        {
            // prev cannot be null, because the consumer cannot null it without consuming the element for which we are
            // trying to get the chunk.
            currentChunk = currentChunk.lvPrev();
            assert currentChunk != null;
        }
        assert currentChunk.lvIndex() == requiredChunkIndex;
        return currentChunk;
    }

    protected final R appendNextChunks(
        R currentChunk,
        long currentChunkIndex,
        long chunksToAppend)
    {
        assert currentChunkIndex != NOT_USED;
        // prevent other concurrent attempts on appendNextChunk
        if (!casProducerChunkIndex(currentChunkIndex, ROTATION))
        {
            return null;
        }
        /* LOCKED FOR APPEND */
        {
            // it is valid for the currentChunk to be consumed while appending is in flight, but it's not valid for the
            // current chunk ordering to change otherwise.
            assert currentChunkIndex == currentChunk.lvIndex();

            for (long i = 1; i <= chunksToAppend; i++)
            {
                R newChunk = newOrPooledChunk(currentChunk, currentChunkIndex + i);
                soProducerChunk(newChunk);
                //link the next chunk only when finished
                currentChunk.soNext(newChunk);
                currentChunk = newChunk;
            }

            // release appending
            soProducerChunkIndex(currentChunkIndex + chunksToAppend);
        }
        /* UNLOCKED FOR APPEND */
        return currentChunk;
    }

    private R newOrPooledChunk(R prevChunk, long nextChunkIndex)
    {
        R newChunk = freeChunksPool.poll();
        if (newChunk != null)
        {
            // single-writer: prevChunk::index == nextChunkIndex is protecting it
            assert newChunk.lvIndex() < prevChunk.lvIndex();
            newChunk.soPrev(prevChunk);
            // index set is releasing prev, allowing other pending offers to continue
            newChunk.soIndex(nextChunkIndex);
        }
        else
        {
            newChunk = newChunk(nextChunkIndex, prevChunk, chunkMask + 1, false);
        }
        return newChunk;
    }


    /**
     * Does not null out the first element of `next`, callers must do that
     */
    final void moveToNextConsumerChunk(R cChunk, R next)
    {
        // avoid GC nepotism
        cChunk.soNext(null);
        next.soPrev(null);
        // no need to cChunk.soIndex(NOT_USED)
        if (cChunk.isPooled())
        {
            final boolean pooled = freeChunksPool.offer(cChunk);
            assert pooled;
        }
        this.soConsumerChunk(next);
        // MC case:
        // from now on the code is not single-threaded anymore and
        // other consumers can move forward consumerIndex
    }

    @Override
    public Iterator iterator()
    {
        throw new UnsupportedOperationException();
    }

    @Override
    public int size()
    {
        return IndexedQueueSizeUtil.size(this);
    }

    @Override
    public boolean isEmpty()
    {
        return IndexedQueueSizeUtil.isEmpty(this);
    }

    @Override
    public int capacity()
    {
        return MessagePassingQueue.UNBOUNDED_CAPACITY;
    }

    @Override
    public boolean relaxedOffer(E e)
    {
        return offer(e);
    }

    @Override
    public int drain(Consumer c)
    {
        return MessagePassingQueueUtil.drain(this, c);
    }

    @Override
    public int fill(Supplier s)
    {
        final int chunkCapacity = chunkMask + 1;
        final int offerBatch = Math.min(PortableJvmInfo.RECOMENDED_OFFER_BATCH, chunkCapacity);
        return MessagePassingQueueUtil.fillInBatchesToLimit(this, s, offerBatch, chunkCapacity);
    }

    @Override
    public int drain(Consumer c, int limit)
    {
        return MessagePassingQueueUtil.drain(this, c, limit);
    }

    @Override
    public void drain(Consumer c, WaitStrategy wait, ExitCondition exit)
    {
        MessagePassingQueueUtil.drain(this, c, wait, exit);
    }

    @Override
    public void fill(Supplier s, WaitStrategy wait, ExitCondition exit)
    {
        MessagePassingQueueUtil.fill(this, s, wait, exit);
    }

    @Override
    public String toString()
    {
        return this.getClass().getName();
    }
}




© 2015 - 2024 Weber Informatics LLC | Privacy Policy