Optimization program using ThreadPool. C#

Question

With this code vozhus very much in it for a long time tinkering that he worked steadily and rapidly. But can not seem to speed it up. Is slow ...

for (int jrCnt = rCnt; jrCnt <= arrayTable.GetUpperBound(0); jrCnt++)
{
    var prcI = new Price();

    /* here is the code search and add data to prcI */

    if ((!string.IsNullOrEmpty(prcI.name)) && (prcI.prc != 0))
    { // function add

        /* adding more information to prcI */

        ThreadPool.QueueUserWorkItem(delegate
        {
            if (!Accessor.AddProductUpdateProduct(prcI)) _updateCounter++;
            _countadd++;
        }); // I put the longest function in the streams
    }
}

Here we call the function. It runs for a long time, even with the threadpool.

public static bool AddProductUpdateProduct(Price price)
{
   using (var db = new PriceDataContext())
        {
            var matchedprod =
                db.Price.Where(x => x.name == price.name && x.company == price.company && x.date != price.date);

            if (matchedprod.Select(x=>x).Count() > 1)
            {
                db.Price.DeleteOnSubmit(matchedprod.First());
                db.SubmitChanges();
            }

            var matchedproduct = matchedprod.SingleOrDefault();

            if (matchedproduct != null)
            {
                matchedproduct.date = price.date;
                matchedproduct.prc = price.prc;

                db.SubmitChanges();
                return false;
            }
        }


/*here the code to add the product to the database.*/
return true;
}

Please tell me how to speed up the work with the threadpool?

Answer 1

Using a separate thread doesn't speed anything app. All you do is to move the processing from the primary thread to another one. You need to break the processing into smaller parts and move each part to a separate thread to get any performance gain.

However, it will not help in this case. It's your LINQ queries that are flawed. Enable debugging. Look at the generated SQL and fix them.

Secondly:

if (matchedprod.SingleOrDefault() != null)
{
    matchedprod.SingleOrDefault().date = price.date;
    matchedprod.SingleOrDefault().prc = price.prc;

    db.SubmitChanges();
    return false;
}

That will query the database three times. Once per call to SingleOrDefault. Do one query and store the result in a variable.

Third:

What's the difference between matchedprod and matchedprodDel? The queries for them are equal?

Forth:

This is easier to read:

var matchedprod = db.Price.Where(x => x.name == price.name && x.company == price.company && x.date != price.date)

Answer 2

It runs for a long time, even with the threadpool.

Delusion. If a method takes 10 seconds, it takes 10 seconds in and out of a thread. A thread does not magically improove the speed of a function.

What the threadpool does is that it allows you to run X method calls in parallel, but none of them gets faster in itself.

Now, I would start with the call - see, you have some SQL there that should not take more than a millisecond to execute, so if it is slow, may I suggest either overloaded hardware or some not so smart person not knowing what an index is, making the SQL statements very very slow?

Also note that a thread pool, unless preconfigured, is SLOWLY ramping up threads - like one per second I think. So if you queue a lot... you are dead.

Finally, why the heck do you use a thread pool, and not the tasks library? It is not that new.... and has a nicer API and more control over the threads.

And finally, use a profiler to find out where the time is spent. It is not that hard. Programming without a profiler is like trying to cook without heating anything - you are limited in what you doo. Professional cook? Use professional tools.

Answer 3

Usually speeding up code first involves using a so called "Profiler". This is a tool that actually measures how your code performs (down to the source code line level).

Before doing any optimization, this should always be your first step. My own experiences shown me that the pure guessing what is slow usually is wrong. Instead, profiling helped me to find the real bottlenecks to optimize.

So I suggest that you first use a profiler, personally I'm using ANTS Performance Profiler (14 day free trial), to detect your slow running code lines.