org.apfloat.ParallelHelper Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of apfloat Show documentation
High performance arbitrary precision arithmetic library
The newest version!
/*
 * MIT License
 *
 * Copyright (c) 2002-2024 Mikko Tommila
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */
package org.apfloat;

import java.util.Arrays;
import java.util.List;
import java.util.ArrayList;
import java.util.Queue;
import java.util.PriorityQueue;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Future;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.atomic.AtomicInteger;

/**
 * Helper methods for parallel algorithms.
 *
 * @since 1.8.0
 * @version 1.14.0
 * @author Mikko Tommila
 */

class ParallelHelper
{
    @FunctionalInterface
    public interface ProductKernel
    {
        public void run(Queue heap);
    }

    private ParallelHelper()
    {
    }

    /**
     * Multiply two smallest elements in the heap and put the product back to the heap,
     * until only one element remains.
     *
     * @param x The elements to be multiplied.
     * @param heap The priority queue to use. Must be initially empty.
     * @param kernel The multiplication kernel callback.
     */

    // Thanks to Peter Luschny and Spiro Trikaliotis for the improved algorithm!
    public static  void parallelProduct(T[] x, Queue heap, ProductKernel kernel)
    {
        // If there are lots of numbers then use a fully parallel algorithm, for small products the overhead is not worth it
        ApfloatContext ctx = ApfloatContext.getContext();
        int numberOfProcessors = ctx.getNumberOfProcessors();
        if (x.length >= 1000 && numberOfProcessors > 1)
        {
            // First multiply small numbers in parallel;
            // number size would be so small that they would not be multiplied using a parallel algorithm (roughly dependent on cache L1 size)
            long maxSize = (long) (ctx.getCacheL1Size() * 2.5 / Math.log((double) ctx.getDefaultRadix()));
            // There is no efficient "ConcurrentPriorityQueue" data structure, so we just split the data
            List> subHeaps = new ArrayList<>();
            for (int i = 0; i < numberOfProcessors; i++)
            {
                subHeaps.add(new PriorityQueue<>(heap));
            }
            int i = 0;
            for (T a : x)
            {
                // Add only small numbers to the parallel heaps, big numbers go to the main heap directly
                (a.size() <= maxSize ? subHeaps.get(i++) : heap).add(a);
                i = (i == numberOfProcessors ? 0 : i);
            }
            AtomicInteger index = new AtomicInteger();
            Runnable runnable = () ->
            {
                Queue subHeap = subHeaps.get(index.getAndIncrement());
                long size = 0;
                // Multiply numbers as long as there are at least two and they are small enough
                // Note that with the heap we will start with the smallest numbers and size will grow
                while (subHeap.size() > 1 && size <= maxSize)
                {
                    kernel.run(subHeap);
                    size = subHeap.peek().size();
                }
                synchronized (heap)
                {
                    // Synchronize the adds; nothing must be reading the heap at the same time
                    heap.addAll(subHeap);
                }
            };

            // Run the runnable in multiple threads
            runParallel(runnable, numberOfProcessors - 1);
        }
        else
        {
            // Use the single-thread algorithm
            heap.addAll(Arrays.asList(x));
        }

        // Then just multiply the remaining numbers;
        // they will presumably be so big that the multiplications use the parallel algorithm anyway
        while (heap.size() > 1)
        {
            kernel.run(heap);
        }
    }

    public static  T getFuture(Future future)
    {
        try
        {
            return future.get();
        }
        catch (InterruptedException ie)
        {
            throw new ApfloatInterruptedException("Waiting for dispatched task to complete was interrupted", ie);
        }
        catch (ExecutionException ee)
        {
            throw new ApfloatRuntimeException("Task execution failed", ee);
        }
    }

    public static void runParallel(Runnable runnable)
    {
        ApfloatContext ctx = ApfloatContext.getContext();
        int numberOfThreads = ctx.getNumberOfProcessors() - 1;

        runParallel(runnable, numberOfThreads);
    }

    private static void runParallel(Runnable runnable, int numberOfThreads)
    {
        ApfloatContext ctx = ApfloatContext.getContext();
        ExecutorService executorService = ctx.getExecutorService();
        List> futures = new ArrayList>();

        // Dispatch other threads, if any
        for (int i = 0; i < numberOfThreads; i++)
        {
            futures.add(executorService.submit(runnable));
        }

        // Also run the Runnable in the current thread
        runnable.run();

        // Join the other threads, if any
        futures.forEach(ParallelHelper::getFuture);
    }
}