Comparing Java Fibers vs ForkJoinPool - performance of FJP is 60% faster than Fibers are currently and use less memory

App.java

package com.sampullara.fibers;

import java.util.concurrent.LinkedBlockingDeque;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.LongAdder;

public class App {

  public static final int NUM = 100_000_000;

  public static void main(String[] args) throws InterruptedException {
    long start = System.currentTimeMillis();
    LinkedBlockingDeque<Fiber> q = new LinkedBlockingDeque<>();
    LongAdder la = new LongAdder();
    Fiber.execute(
        () -> {
          for (int i = 0; i < NUM; i++) {
            q.add(Fiber.execute(la::increment));
          }
        });
    for (int i = 0; i < NUM; i++) {
      q.poll(1, TimeUnit.MINUTES).await();
    }
    long diff = System.currentTimeMillis() - start;
    System.out.println(la.longValue() / diff + " increments per ms");
  }
}

App2.java

package com.sampullara.fibers;

import java.util.concurrent.*;
import java.util.concurrent.atomic.LongAdder;

public class App2 {

  public static final int NUM = 100_000_000;

  public static void main(String[] args) throws InterruptedException, ExecutionException {
    long start = System.currentTimeMillis();
    LinkedBlockingDeque<ForkJoinTask> q = new LinkedBlockingDeque<>();
    LongAdder la = new LongAdder();
    ForkJoinPool.commonPool().execute(() -> {
      for (int i = 0; i < NUM; i++) {
        q.add(ForkJoinPool.commonPool().submit(la::increment));
      }
    });
    for (int i = 0; i < NUM; i++) {
      q.poll(1, TimeUnit.MINUTES).get();
    }
    long diff = System.currentTimeMillis() - start;
    System.out.println(la.longValue() / diff + " increments per ms");
  }
}