C# - AsParallel Example - Dot Net Perls

AsParallel. Query expressions can be run in parallel. With the AsParallel extension method, we enable parallel threads to improve performance.

Extension

We show a method that becomes twice as fast on a dual-core machine when AsParallel is used. AsParallel can make certain queries much faster.

Example. We introduce 2 methods. The SumDefault() method (shown first) sums all the elements in the array with the Sum method alone.

And SumAsParallel first calls AsParallel on the array. It then calls Sum on the result of AsParallel. The 2 methods have the same result.

Sum

Result We can see that AsParallel makes the query twice as fast. This is the result on a dual-core machine under light load.

using System; using System.Diagnostics; using System.Linq; class Program { static int SumDefault(int[] array) { // Sum all numbers in the array. return array.Sum(); } static int SumAsParallel(int[] array) { // Enable parallelization and then sum. return array.AsParallel().Sum(); } static void Main() { // Generate array. int[] array = Enumerable.Range(0, short.MaxValue).ToArray(); // Test methods. Console.WriteLine(SumAsParallel(array)); Console.WriteLine(SumDefault(array)); const int m = 10000; var s1 = Stopwatch.StartNew(); for (int i = 0; i < m; i++) { SumDefault(array); } s1.Stop(); var s2 = Stopwatch.StartNew(); for (int i = 0; i < m; i++) { SumAsParallel(array); } s2.Stop(); Console.WriteLine(((double)(s1.Elapsed.TotalMilliseconds * 1000000) / m).ToString("0.00 ns")); Console.WriteLine(((double)(s2.Elapsed.TotalMilliseconds * 1000000) / m).ToString("0.00 ns")); } }

536821761 536821761 232450.53 ns 118515.85 ns

Discussion. AsParallel will not have great results on all kinds of queries. For small collections, AsParallel will be slower because of the extra method call.

Info For example, I changed the array to only have two elements with Enumerable.Range(0, 2).

Enumerable.Range

Result With two elements, AsParallel makes the query run over 200 times slower instead of two times faster.

1 1 48.84 ns 10914.27 ns

This program uses a static field in a Func implementation. It uses this Func in a parallel query (created by AsParallel()). The program is supposed to count the even numbers in the int array.

Func

int Array

However The AsParallel call causes multiple threads to access the static field. The threads use IsEvenCounter, a non-thread-safe method.

static

So The results are unpredictable. In the array with 500 even numbers, this program will count 478, 500, or 517 even numbers.

Odd, Even

Note If you remove the AsParallel() call from the statement, the result is always 500.

Tip This program proves that AsParallel is creating multiple threads. And this exposes an error in the program design.

using System; using System.Linq; class Program { static int _counter; static bool IsEvenCounter(int value) { // ... Return whether counter field is even. return Program._counter++ % 2 == 0; } static void Main() { Func<int, bool> func = Program.IsEvenCounter; int[] array = Enumerable.Range(0, 1000).ToArray(); // ... Use parallel query ten times on the array. // ... Write number of results matched by Func. for (int i = 0; i < 10; i++) { var result = array.AsParallel().Where(func); Console.WriteLine(result.Count()); } } }

517 514 500 509 500 478 500 508 500 501

It is tempting to think that AsParallel makes queries faster. This is true on certain, long-running queries, but it makes small and fast queries many times slower.