I’m currently working on my upcoming ASP.NET 4.5 Expert Skills book, and it struck me that there are now at least 4 different options available to the programmer when implementing multi-threading. If you’re not already familiar with these classes, choosing between them is likely to be extremely difficult.<\/p>\n
Let’s take some simple pseudo-code as an example:<\/p>\n
foreach (var FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 EncryptFile(FileToEncrypt);\r\n}<\/pre>\nIn this example, let’s imagine that the EncryptFile<\/em> method is very processor-intensive, so we’d like to split the calls to EncryptFile<\/em> into separate threads.<\/p>\nUsing the Thread class<\/strong><\/span><\/p>\nThe Thread<\/em> class has been around since the original release of the .NET framework, so\u00a0 it’s likely that existing multi-threading code will use this class.<\/em><\/p>\nThe Thread<\/em> class is found in the System.Threading<\/em> namespace.<\/p>\nHere’s how the multi-threaded code would look using the Thread<\/em> class:<\/p>\nList<Thread> MyThreads = new List<Thread>();\r\nforeach (var FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 Thread NewThread = new Thread(new ParameterizedThreadStart(FileToEncrypt));\r\n\u00a0\u00a0\u00a0 NewThread.Start(FileToEncrypt);\r\n}<\/pre>\nThis looks fine as long as it is all that your program needs to do. But let’s assume that we need to do something with the encrypted files after the threads have finished encrypting them. We’ll need to add some code that waits until all of the threads have finished before continuing:<\/p>\n
List<Thread> Threads = new List<Thread>();\r\nforeach (string FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 Thread NewThread = new Thread(new ParameterizedThreadStart(FileToEncrypt));\r\n\u00a0\u00a0\u00a0 Threads.Add(NewThread);\r\n\u00a0\u00a0\u00a0 NewThread.Start();\r\n\u00a0\u00a0\u00a0 Stuff(FileToEncrypt);\r\n}\r\nbool AllThreadsFinished = false;\r\nwhile (!AllThreadsFinished)\r\n{\r\n\u00a0\u00a0\u00a0 AllThreadsFinished = true;\r\n\u00a0\u00a0\u00a0 foreach (Thread MyThread in Threads)\r\n\u00a0\u00a0\u00a0 {\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 if (MyThread.IsAlive) AllThreadsFinished = false;\r\n\u00a0\u00a0\u00a0 }\r\n\u00a0\u00a0\u00a0 Thread.Sleep(1000);\r\n}<\/pre>\nWe’ve had to add quite a lot of code to achieve it, but this code will now wait for all of the threads to finish. This isn’t the only way to achieve this using the Thread<\/em> class, but it’s the easiest to understand and uses the least code.<\/p>\nThe Thread<\/em> class does a good job of splitting code into multiple threads, and will even benefit from the improvements to threading that were added to ASP.NET 4.5. If your existing code uses the Thread<\/em> class, there’s really no need to change it.<\/p>\nUsing the ThreadPool class<\/strong><\/span><\/p>\nThe ThreadPool<\/em> class is another class that has been around since the first version of the .NET framework. ThreadPool<\/em> offers more control over how threads are distributed to the computers processor or processors, but it is much harder to implement the ‘wait’ code that we used for the Thread<\/em> class.<\/p>\nThe ThreadPool<\/em> class is found in the System.Threading<\/em> namespace.<\/p>\nHere’s how our threading code would look using the ThreadPool<\/em> class:<\/p>\nforeach (string FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 ThreadPool.QueueUserWorkItem(new WaitCallback(EncryptFile), FileToEncrypt);\r\n}<\/pre>\nThis looks much easier, but the ThreadPool<\/em> class doesn’t give us any way to check on the status of the thread that we just created, so how can we tell when it has finished?<\/p>\nThere are some very complicated ways of extracting the status of the thread, but an alternative option is to maintain a counter that is incremented when the EncryptFile<\/em> method finishes. We’ll call it: NumberOfFilesEncrypted<\/strong><\/p>\n<\/strong>We can then wait for all of the threads to finish by using the following code:<\/p>\nforeach (string FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 ThreadPool.QueueUserWorkItem(new WaitCallback(EncryptFile), FileToEncrypt);\r\n}\r\nwhile (NumberOfFilesEncrypted < FilesToEncrypt.Count)\r\n{\r\n\u00a0\u00a0\u00a0 Thread.Sleep(1000);\r\n}<\/pre>\nOf course, this is far from ideal. If one of the threads failed for some reason, the code would wait forever for the last thread to finish. This could be worked around with error handling and timeouts, but it’s really better to avoid the ThreadPool<\/em> class if possible.<\/p>\nUsing the new Task class<\/strong><\/span><\/p>\nThe Task<\/em> class was introduced in .NET 4.0, but has only really been finalized in the .NET 4.5 release. The Task<\/em> class combines the functionality of the Thread<\/em> and ThreadPool<\/em> classes to make threading easier and more efficient.<\/p>\nThe Task<\/em> class is found in the new System.Threading.Tasks<\/em> namespace.<\/p>\nHere’s how our code would look using the Task<\/em> class:<\/p>\nforeach (string FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 Task.Factory.StartNew(Argument => EncryptFile(Argument), FileToEncrypt);\r\n}<\/pre>\nOnce again, of course, we need to check whether all of the threads have finished before continuing. We can do this in a very similar way to the Thread<\/em> class, but with much less code:<\/p>\nList<Task> MyTasks = new List<Task>();\r\nforeach (string FileToEncrypt in FilesToEncrypt)\r\n{\r\n\u00a0\u00a0\u00a0 MyTasks.Add(Task.Factory.StartNew(Argument => EncryptFile(Argument), FileToEncrypt));\r\n}\r\nTask.WaitAll(MyTasks.ToArray());<\/pre>\nThe Task.WaitAll<\/em> method makes everything much easier! If you have the choice between the Thread<\/em> and Task<\/em> classes, I’d recommend the Task<\/em> class every time.<\/p>\nHowever, I’ve saved the best for last. The Parallel <\/em>class enables us to carry out the task with even less code.<\/p>\nUsing the new Parallel class<\/strong><\/span><\/p>\nThe Parallel<\/em> class enables you to change any for<\/em> or foreach<\/em> loop into multi-threaded code. Every iteration of the loop will be intelligently assigned to a separate thread and will be executed in the most efficient way possible.<\/p>\nThe Parallel<\/em> class can be found in the System.Threading.Tasks<\/em> namespace.<\/p>\nHere’s how our code would look using the Parallel<\/em> class:<\/p>\nParallel.ForEach(FilesToEncrypt, FileToEncrypt =>\r\n{\r\n\u00a0\u00a0\u00a0 EncryptFile(FileToEncrypt);\r\n});<\/pre>\nYou might be wondering how we wait for all of the threads to finish. In fact, the Parallel<\/em> class handles this automatically! The foreach<\/em> loop won’t finish until all of the threads have finished running.<\/p>\nIf the code that you want to be multi-threaded can be expressed as a for<\/em> or foreach<\/em> loop, there’s little doubt that you want to use the Parallel<\/em> class.<\/p>\nIf you’re using ASP.NET 4.5, I would highly recommend using the Task<\/em> class and Parallel<\/em> classes and ignoring the older Thread<\/em> and ThreadPool<\/em> classes.<\/p>\n","protected":false},"excerpt":{"rendered":"I’m currently working on my upcoming ASP.NET 4.5 Expert Skills book, and it struck me that there are now at least 4 different options available to the programmer when implementing multi-threading. If you’re not already familiar with these classes, choosing … Continue reading