A port of the ForkJoinPool JSR-166 implementation used in Scala 2.11.x for usage in Scala 2.12.

In Scala 2.11.8 the scala.concurrent.forkjoin.ForkJoinPool implementation is a fork of the JSR-166 implementation by Doug Lea, in order to provide support for Java versions older than version 8. However in Scala 2.12.x the implementation is now an alias for java.util.concurrent.ForkJoinPool, given of its availability in Java 8 and the Scala 2.12 requirement to have Java 8 as a target.

Unfortunately these 2 implementations are NOT the same, as the old Scala 2.11 implementation has better throughput in testing.

The version of ForkJoinPool in 2.11 uses busy-waiting more aggressively which helps out in tests with more cores than tasks. However, this comes at the expense of other workloads on the process/machine, especially because the JVM does not let the busy waiting signal its intention to the CPU with a Spin Loop hint. This trade-off was found during testing of the ForkJoinPool as it was integrated into parallel java.util.stream and is discussed in JDK-8080623.

But in testing this implementation can have considerable better throughput and can prove useful for heavy CPU-bound tasks and the performance hit for some use-cases is quite noticeable.

Available for Scala 2.10, 2.11 and 2.12. Add the dependency in your SBT file:

libraryDependencies += "io.monix" %% "monix-forkjoin" % "1.0" 

And then:

import monix.forkJoin.ForkJoinExecutionContext

implicit val executionContext = 
  ForkJoinExecutionContext.createDynamic(
    parallelism = Runtime.getRuntime.availableProcessors(),
    maxThreads = 256,
    name = "forkJoin-dynamic",
    daemonic = true
  )

The monix-forkjoin project is licensed under the 3-Clause BSD license, the same license as Scala, see LICENSE.txt for details.

The ForkJoinPool implementation is copied from the Scala 2.11 repository, which itself was copied from Doug Lea's JSR-166 implementation and is under the public domain.