NumSharp is slower than normal for loop ?!?

[minhduc66532]

static void Main(string[] args)
{
    List<double> test2 = new();
    Random randNum = new Random();
    for (int i = 0; i < 1000000; i++)
    {
        test2.Add(randNum.NextDouble());
    }
    double[] ar = test2.ToArray();
    NDArray arn = new NDArray(ar);
    Stopwatch s = new();

    int count = 0;
    while (count < 5)
    {
        count++;

        double avg = 0;
        s.Start();
        for (int i = 0; i < ar.Length; i++)
        {
            avg += ar[i];
        }
        avg /= ar.Length;
        s.Stop();
        Console.WriteLine($"{s.ElapsedMilliseconds} {s.ElapsedTicks}");
        s.Reset();

        s.Start();
        arn.mean();
        s.Stop();
        Console.WriteLine($"{s.ElapsedMilliseconds} {s.ElapsedTicks}");
        s.Reset();
        Console.WriteLine();
    }
}

The result: Am I doing anything wrong because from the test result NumSharp is 3 times slower than normal for loop

[Nucs]

It is due to usage of Parallel.For internally by our DefaultEngine, we are considering to remove this feature in favor of regular tight for-loop by default where Parallel.For will be exposed via ParallelEngine that can be assigned to any NDArray. Any performance impact beyond that is the overhead of creating the NDArray, allocating the memory and handling dtype and shape behind the scenes.

[minhduc66532]

Ohh ok, so more improvements coming soon I guess. Thank you for the answer