performance issue with Nondeterminism

[Tiger Chan]

Hi, could anyone explain why concurrent process with nondeterminism is slower than sequential approach? attached below is my demo code:

def seqFib(n: Int): Task[Int] =  n match {

  case 0 | 1 => Task now  1

  case n => {

    for {

      x <- seqFib(n-1)

      y <- seqFib(n-2)

    } yield x + y

  }

 }                                                //> seqFib: (n: Int)scalaz.concurrent.Task[Int]

 //concurrent approach

 def parFib(n: Int): Task[Int] = n match {

    case 0 | 1 => Task now 1

    case n => {

       val ND = Nondeterminism[Task]

       for {

         pair <- ND.both(parFib(n-1), parFib(n-2))

         (x,y) = pair

       } yield x + y

    }

 }                                                //> parFib: (n: Int)scalaz.concurrent.Task[Int]

 def msFib(n: Int, fib: Int => Task[Int]) = for {

   b <- Task now { System.currentTimeMillis() }

   a <- fib(n)

   e <- Task now { System.currentTimeMillis() }

 } yield (a, (e-b))                               //> msFib: (n: Int, fib: Int => scalaz.concurrent.Task[Int])scalaz.concurrent.T

                                                  //| ask[(Int, Long)]

 

 msFib(20, parFib).unsafePerformSync              //> res3: (Int, Long) = (10946,315)

 msFib(20, seqFib).unsafePerformSync              //> res4: (Int, Long) = (10946,16)

thanks!