How to combine Thread-local variables and speed up small loop bodies?

Sun, 14 Jun 2009 17:16:17 Z

Hi,

I was trying to combine the two How-To´s from msdn:
Speed Up Small Loop Bodies
Write a Parallel.For Loop with Thread-Local Variables

The example with the parallel thread-local variables was slower then the corresponding sequential implementation.
So I want to create partitions based on the cpu count and parallelize just this (like the first example).

Of course the following code isn´t correct because the variable "total" isn´t summed up thread-safe:

 private static IEnumerable<Tuple<int, int>> CreateRanges(int fromInclusive, int toExclusive)
    {
        // argument checking omitted
        int rangeSize = (toExclusive - fromInclusive) / System.Environment.ProcessorCount;
        for (int i = fromInclusive; i < toExclusive; i += rangeSize)
        {
            int from = i, to = Math.Min(i + rangeSize, toExclusive);
            yield return Tuple.Create(from, to);
        }
    }

    static void Main()
    {
        int[] arr2 = Enumerable.Range(0, 1000000).ToArray();
        long total = 0;
        Parallel.ForEach(CreateRanges(0, arr2.Length), range =>
        {
            for (int i = range.Item1; i < range.Item2; i++)
            {
                // Small body.
                total += arr2[i];
            }
        });

        Console.WriteLine(total);
        Console.ReadKey();
    }

But I am really stuck cause I don´t have a glue how to parallelize the partitions and get the correct results.
Can anybode help me?

Thanks,
Alex

How to combine Thread-local variables and speed up small loop bodies?

Sun, 14 Jun 2009 20:09:52 Z

ForEach like this might do the job:

 Parallel.ForEach <Tuple<int, int>, long>(CreateRanges(0, arr2.Length),
                () => 0,
                (range, loop, subtotal) =>
                {
                    for (int i = range.Item1; i < range.Item2; i++)
                    {
                        // Small body.
                        subtotal += arr2[i];
                    }
                    return subtotal;
                },
                subtotal => Interlocked.Add(ref total, subtotal));

Miha Markic [MVP C#] http://blog.rthand.com