Optimizing CPU usage by performing tasks in parallel with different priorities in Java

A small limitation of Java CompletableFuture is the inability to choose the priority of the thread they are assigned to, thus making it impossible to assign the desired workload to the cpu.

In this situation, the BackgroundExecutor component of Burningwave Core library comes to our aid, by giving us the ability to launch different functional interfaces in parallel by setting the priority of the thread they will be assigned. There is also the option to wait for them start or finish. This component is also used by the IterableObjectHelper component to iterate collections or arrays in parallel.

To use the BackgroundExecutor you should simply add the following to your projects dependencies:

For obtaining threads the BackgroundExecutor uses the ThreadSupplier component which can be configured in the burningwave.static.properties file through the following properties:

thread-supplier.default-daemon-flag-value=true
thread-supplier.default-thread-priority=5
thread-supplier.max-detached-thread-count=${thread-supplier.max-poolable-thread-count}
thread-supplier.max-detached-thread-count.elapsed-time-threshold-from-last-increase-for-gradual-decreasing-to-initial-value=30000
thread-supplier.max-detached-thread-count.increasing-step=autodetect
thread-supplier.max-poolable-thread-count=autodetect
thread-supplier.poolable-thread-request-timeout=6000

The ThreadSupplier provides a fixed number of reusable threads indicated by the thread-supplier.max-poolable-thread-count property and, if these threads have already been assigned, new non-reusable threads will be created whose quantity maximum is indicated by the thread-supplier.max-detached-thread-count property. Once this limit is reached if the request for a new thread exceeds the waiting time indicated by the thread-supplier.poolable-thread-request-timeout property, the ThreadSupplier will proceed to increase the limit indicated by the thread-supplier.max-detached-thread-count property for the quantity indicated by the thread-supplier.max-detached-thread-count.increasing-step property. Resetting the thread-supplier.max-detached-thread-count property to its initial value, will occur gradually only when there have been no more waits on thread requests for an amount of time indicated by the thread-supplier.max-detached-thread-count.elapsed-time-threshold-from-last-increase-for-gradual-decreasing-to-initial-value property.

Let’s see now an example on how to create tasks:

import static org.burningwave.core.assembler.StaticComponentContainer.BackgroundExecutor;

import org.burningwave.core.ManagedLogger;
import org.burningwave.core.concurrent.QueuedTasksExecutor.ProducerTask;
import org.burningwave.core.concurrent.QueuedTasksExecutor.Task;


public class TaskLauncher implements ManagedLogger {

    public void launch() {
        ProducerTask<Long> taskOne = BackgroundExecutor.createProducerTask(task -> {
            Long startTime = System.currentTimeMillis();
            logInfo("task one started");
            synchronized (this) {                
                wait(5000);
            }
            Task internalTask = BackgroundExecutor.createTask(tsk -> {
                logInfo("internal task started");    
                synchronized (this) {                
                    wait(5000);
                }
                logInfo("internal task finished");    
            }, Thread.MAX_PRIORITY).submit();
            internalTask.waitForFinish();
            logInfo("task one finished");
            return startTime;
        }, Thread.MAX_PRIORITY);
        taskOne.submit();
        Task taskTwo = BackgroundExecutor.createTask(task -> {
            logInfo("task two started and wait for task one finishing");
            taskOne.waitForFinish();
            logInfo("task two finished");    
        }, Thread.NORM_PRIORITY);
        taskTwo.submit();
        ProducerTask<Long> taskThree = BackgroundExecutor.createProducerTask(task -> {
            logInfo("task three started and wait for task two finishing");
            taskTwo.waitForFinish();
            logInfo("task two finished");
            return System.currentTimeMillis();
        }, Thread.MIN_PRIORITY);
        taskThree.submit();
        taskThree.waitForFinish();
        logInfo("Elapsed time: {}ms", taskThree.join() - taskOne.join());
    }

    public static void main(String[] args) {
        new TaskLauncher().launch();
    }

}