Adjust a function for Parallel For in c#

Question

I just started to learn about the Parallel.For() in c# and I have a function that doing something and I don't know how to adjust it to be Parallel. Can any anybody show me an example of such similar case?

Update:

for (int i = 0; i < (samples32array.Length); i += 35) //Delimiter Samples Length = SampleRate*DelimiterTimeLength,( i.e: 44100*0.2=8820 samples)
{
    if ((i + (int)DelimiterSamplesCount - 1) > samples32array.Length)
    {
                    break;
    }
    double power = MyGoertzel.GoertzelFilter(samples32array, i, i + (int)DelimiterSamplesCount - 1, coeffDelimiter);
    if (power > max_power)
    {
        max_power = power;
        max_power_index = i;
    }
}

Update: this is my tryings but something is going wrong ...

private int delimiter_search_Parallel(float[] samples32array, int DelimiterSamplesCount, double DelimiterFreq)
    {
        double normalizedfreqDelimiter = DelimiterFreq / 44100;
        double coeffDelimiter = 2 * Math.Cos(2 * Math.PI * normalizedfreqDelimiter);
        double max_power = 0;
        int max_power_index = 0;
        List<double> powerParallelList = new List<double>();
        List<int> powerParallelListIndex = new List<int>();
        Parallel.For(0, samples32array.Length, (i, loopState) =>
            {
                if ((i + (int)DelimiterSamplesCount - 1) > samples32array.Length)
                {
                    //TODO -> Genereate a Windows with ERROR Exception
                    loopState.Stop();
                    return;
                }
                else
                {
                    double power = MyGoertzel.GoertzelFilter(samples32array, i, i + (int)DelimiterSamplesCount - 1, coeffDelimiter);
                    powerParallelList.Add(power);
                    powerParallelListIndex.Add(i);
                }
            }
            );
        max_power = powerParallelList.Max();
        max_power_index = powerParallelListIndex[powerParallelList.IndexOf(max_power)];
        return max_power_index;
    }

You can't parallelize that. Each iteration requires the answer of the previous iteration. — SimpleVar, Jun 07 '15 at 15:58
Hello @YoryeNathan, I updated the code. Sorry, I put on the wrong one — axcelenator, Jun 07 '15 at 16:02
Have an attempt at parallelizing it, and we'll help you where you struggle. Won't write code from scratch. — SimpleVar, Jun 07 '15 at 16:05
My problem is how to tell the Parallel.For to jump +35 for the i variable — axcelenator, Jun 07 '15 at 16:07
[This](http://stackoverflow.com/questions/7142446/parallel-for-step-size) is what you're looking for. Though it won't help much in this particular situation. — Yuval Itzchakov, Jun 07 '15 at 16:14
Just increment i by 1, and multiply it by 35 inside the loop. And obviously, you also need to divide the upper bounds by 35. — Bryce Wagner, Jun 07 '15 at 16:35
Hello do you mean i/35 for this line(upper bounds) Parallel.For(0, samples32array.Length, (i, loopState) => — axcelenator, Jun 07 '15 at 18:15

Alexander Bell · Answer 1 · 2015-06-07T18:44:56.887

For your purpose I would suggest to use a ConcurrentDictionary<int,double>dictPower similar to the following sample(re: https://msdn.microsoft.com/en-us/library/dd287191%28v=vs.110%29.aspx)

 // dictionary with the desired concurrencyLevel and initialCapacity
 ConcurrentDictionary<int, double> cd = new ConcurrentDictionary<int, double>(concurrencyLevel, initialCapacity);

(where concurrency level typically equal the number of CPU cores) and modify your code correspondingly: instead of

powerParallelList.Add(power);
powerParallelListIndex.Add(i);

use:

dictPower.Add(i,power)

Upon completion of that Parallel.For loop, just calculate the Max() of Items in dictPower (as per your requirement), or apply any other function upon necessity.

Hope this may help.

score 1 · Accepted Answer · answered Jun 07 '15 at 19:41

You should use Parallel.For overload that allows you to have thread local state, where you can keep your current maximum value.

private int delimiter_search_Parallel(float[] samples32array,int DelimiterSamplesCount,double DelimiterFreq) {
    if(samples32array.Length%35<DelimiterSamplesCount-1) {
        //TODO -> Genereate a Windows with ERROR Exception
    }
    double normalizedfreqDelimiter=DelimiterFreq/44100;
    double coeffDelimiter=2*Math.Cos(2*Math.PI*normalizedfreqDelimiter);
    object lockObject=new object();
    Tuple<double,int> result=null;
    Parallel.For(0,(samples32array.Length-(DelimiterSamplesCount-1))/35+1,() => (Tuple<double,int>)null,(i,loopState,max) => {
        i*=35;
        double power=MyGoertzel.GoertzelFilter(samples32array,i,i+(int)DelimiterSamplesCount-1,coeffDelimiter);
        return max==null||power>max.Item1?Tuple.Create(power,i):max;
    },max => {
        if(max!=null) {
            lock(lockObject) {
                if(result==null||max.Item1>result.Item1) {
                    result=max;
                }
            }
        }
    });
    return result.Item2;
}

Adjust a function for Parallel For in c#

2 Answers2