CODEWITHSUNDEEP
c#multithreadingperformanceprocessparallel-processing
Mr. Blond
Mr. Blond

Reputation: 1177

Improve programs performance by splitting long running task into separate processes

I've read that when you run a long running operation in separate threads (in parallel) there will be noticeable improvement in performance when your PC has multiple processor cores. In C# there is a function Parallel.ForEach which, how I understand, splits the long running operation in separate threads which are ran on each processor core. Although, if there is only one core, this function will run as a simple Foreach loop (synchronous). Also, if you are creating more threads than there is processor cores, it may even have a negative impact.

I was thinking of creating a separate console application, pass args to it and run it's .exe on multiple processes like start -WindowStyle Hidden SeparateProcess.exe.

Edit:
I already created a console app like this

class Program
{
    static void Main(string[] args)
    {
        for (int i = 0; i < 100; i++)
        {
            Thread.Sleep(200);
            Console.WriteLine("" + i);
        }
        Console.WriteLine();
        Console.WriteLine(Guid.NewGuid() + ": Process finalized.");
    }
}

Then, using powershell executed it many times:

For ($i=0; $i -lt 10; $i++){
start -WindowStyle Normal SeparateProcess.exe 
}

Will this improve program performance when ran on single core?

Note: The long running task is Emgu CV image computing algorithm.

Upvotes: 0

Views: 517

Answers (1)

Lou Franco
Lou Franco

Reputation: 89242

You should check yourself by trying it, but ...

Is there a way to improve long running task performance even when there is only one core?

Yes, but obviously not by running it in parallel. You would use a performance analyzer on the code to find out which parts were taking the most time and try to rewrite it in a more optimized way.

EDIT based on comments:

If you are trying to do image-processing with CV and the thing you are doing is CPU-bound (maxes out the processor while running), then parallelizing it isn't going to help on a single core (they will compete with each other for CPU).

If the thing you are doing has to wait for I/O to complete, then parallelizing it will allow some of the work to go forward while other processes wait on the I/O.

It depends on what you are doing -- you should probably close this question and ask a much more specific one with details on exactly what you are trying to do and how CPU-intensive that task is.

(BTW, if you can, one way to make image processing go faster is to resize the image beforehand -- many things you might try to do with an image work fine on a smaller version -- for example: face detection).

Upvotes: 1

Related Questions