 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
.NET Forum / .NET Framework / New Users / November 2006How do I Sleep for a specified amount of time and relinquish contr
In the docs for Thread.Sleep() it appears unless I specify 0 my thread will still use the rest of its quanta (I guess doing nothing useful) as opposed to being suspended right away to allow other runable thread to execute. What's the best way to sleep for a specified amount of time and relinquish control right away as opposed to using up the rest of my quanta? Do I need to call the Win32 Sleep()? Do I need to call WaitOne() on a synchronization object?  SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com Hello nickdu, Use TimeSpan as param of the Thread.Sleep method For example Thread.Sleep (TimeSpan.FromSeconds (30)); //sleep for 30 secs for example n> In the docs for Thread.Sleep() it appears unless I specify 0 my n> thread will still use the rest of its quanta (I guess doing nothing n> useful) as opposed to being suspended right away to allow other n> runable thread to execute. n> n> What's the best way to sleep for a specified amount of time and n> relinquish control right away as opposed to using up the rest of my n> quanta? Do I need to call the Win32 Sleep()? Do I need to call n> WaitOne() on a synchronization object? n> n> nicknospamdu@community.nospam n> remove "nospam" change community. to msn.com --- WBR, Michael Nemtsev [C# MVP] :: blog: http://spaces.live.com/laflour "At times one remains faithful to a cause only because its opponents do not cease to be insipid." (c) Friedrich Nietzsche That's what I was doing which didn't appear to be working based on my observations. Yes, the thread that executed Sleep(<30 second timespan>) was stopped from doing anything useful for 30 seconds, but it also appeared to be consuming the rest of the thread's quanta. What I did instead, which appears to do what I want is: wait = new AutoResetEvent(false); while (true) { // do some work ... wait.WaitOne(thinkTime, true); } So I wait on an event that will never be signaled and I specify as a timeout the amount of time I want to sleep. Based on my observations I don't have the odd behavior I was seeing before which seems to be explained by the thread continuing to use its quanta when sleeping. I'm guessing calling the Windows Sleep() API via pinvoke would also solve my problem, but that's just a guess. SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com > Hello nickdu, > [quoted text clipped - 20 lines] > "At times one remains faithful to a cause only because its opponents do not > cease to be insipid." (c) Friedrich Nietzsche Hello nickdu, n> That's what I was doing which didn't appear to be working based on my n> observations. Yes, the thread that executed Sleep(<30 second n> timespan>) was stopped from doing anything useful for 30 seconds, but n> it also appeared to be consuming the rest of the thread's quanta. Don't u mix this with SpinWait method? When u waits or sleep a thread relinquish its allocation of CPU time and WaitSleepJoin is set to the ThreadState property --- WBR, Michael Nemtsev [C# MVP] :: blog: http://spaces.live.com/laflour "At times one remains faithful to a cause only because its opponents do not cease to be insipid." (c) Friedrich Nietzsche No, I'm not mixing Sleep() with SpinWait(). I believe Sleep(), as implemented by the .NET framework, is not relinquishing control to runnable threads. And it even mentions this in the docs. It say something to the effect of "specify 0 to relinquish control to runnable threads". And based on my observations it appears to be behaving that way, not relinquishing control as soon as I call Sleep(). It appears to be finishing using its quanta. And based on the fact that the behavior improved to what I was looking to achieve when I changed it to a WaitOne(), I would say that Sleep() with a non zero value is not really doing what I want, and maybe not doing what many others are thinking it should do. SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com > Hello nickdu, > [quoted text clipped - 13 lines] > "At times one remains faithful to a cause only because its opponents do not > cease to be insipid." (c) Friedrich Nietzsche Sleep(0) and Sleep(1) do have a few minor differences between then, aside from the obvious. The key is that Sleep(0) won't allow any lower priority threads to be scheduled, whereas Sleep(1) will. Lots of good detail by Joe Duffy at: http://www.bluebytesoftware.com/blog/PermaLink,guid,1c013d42-c983-4102-9233-ca54 b8f3d1a1.aspx -- Chris Mullins, MCSD.NET, MCPD:Enterprise http://www.coversant.net/blogs/cmullins > No, I'm not mixing Sleep() with SpinWait(). I believe Sleep(), as > implemented by the .NET framework, is not relinquishing control to [quoted text clipped - 28 lines] >> not >> cease to be insipid." (c) Friedrich Nietzsche > No, I'm not mixing Sleep() with SpinWait(). I believe Sleep(), as > implemented by the .NET framework, is not relinquishing control to > runnable > threads. Sleep() as implemented by the .NET Framework had better just be calling the Windows API function Sleep(). And that does indeed cause the calling thread to immediately be suspended, with the next runnable thread being scheduled for the CPU. > And it even mentions this in the docs. It say something to the > effect of "specify 0 to relinquish control to runnable threads". That's right. If you don't want to sleep for any specific amount of time, but still want to give up your timeslice to other runnable threads, you call Sleep() with the parameter of 0. That doesn't mean that calling it with some other value doesn't also relinquish the CPU. It just means that if *all* you want to do is relinquish the CPU, you can use the value of 0. > And based > on my observations it appears to be behaving that way, not relinquishing > control as soon as I call Sleep(). It appears to be finishing using its > quanta. Define "my observations". What makes you believe that after calling Sleep(), your thread continues to run out its timeslice? > And based on the fact that the behavior improved to what I was > looking to achieve when I changed it to a WaitOne(), I would say that > Sleep() > with a non zero value is not really doing what I want, and maybe not doing > what many others are thinking it should do. You haven't described your observations, so how do you justify your claim that calling WaitOne() with a timeout "behaves better" than simply calling Sleep() with the same timeout? Pete > Sleep() as implemented by the .NET Framework had better just be calling > the Windows API function Sleep(). And that does indeed cause the calling > thread to immediately be suspended, with the next runnable thread being > scheduled for the CPU. Well, it's not that simple. (It never is, is it?) On hyperthreaded processors, the behavior is slighly different. Jeff Richter goes into this a tiny bit at: http://msdn.microsoft.com/msdnmag/issues/05/10/ConcurrentAffairs/ http://msdn.microsoft.com/msdnmag/issues/05/10/ConcurrentAffairs/default.aspx?lo c=&side=true#a If I understood the issues here better, I would explain them. I'm hoping someone much smarter than me fills in all the issues here. > That doesn't mean that calling it with some other value doesn't also > relinquish the CPU. It just means that if *all* you want to do is > relinquish the CPU, you can use the value of 0. There are some good reasons no to use "0" as the value passed to sleep. Joe Duffy goes into the details at: http://www.bluebytesoftware.com/blog/PermaLink,guid,1c013d42-c983-4102-9233-ca54 b8f3d1a1.aspx > What makes you believe that after calling Sleep(), your thread continues > to run out its timeslice? Ya know, if I had to figure that out, and had full access to the machine with all the hardcores debuggers I could find, it would stil be VERY hard to figure out. I would be (pleasently) shocked to see an easy way to determine this. -- Chris Mullins, MCSD.NET, MCPD:Enterprise http://www.coversant.net/blogs/cmullins > [...] >> That doesn't mean that calling it with some other value doesn't also >> relinquish the CPU. It just means that if *all* you want to do is >> relinquish the CPU, you can use the value of 0. > > There are some good reasons no to use "0" as the value passed to sleep. This is true, depending on what your goals are. I did not intend my reply to be a fully detailed treatise on the exact behavior of Sleep(). However, the OP is claiming that even when he calls Sleep() with a *non-zero* value, the thread continues to run until the timeslice is finished. I hope we can agree that this is a *highly* unlikely scenario. Pete I take Sonata every night, just before bedtime, since once I get to sleep, I generally stay asleep. However, under stress, I occasionally add Lorazepam to the mix. Signature;-), Kevin Spencer Microsoft MVP Ministry of Software Development http://unclechutney.blogspot.com Any experience you can walk away from is a good one. >> [...] >>> That doesn't mean that calling it with some other value doesn't also [quoted text clipped - 12 lines] > > Pete From reading the docs I'm remember it saying something about continuing to use the thread's quanta. I don't think the .NET Thread.Sleep() simply calls the Win32 Sleep() API. You're correct that I didn't go into specifics on my observations which I will do now. I started looking into this because a load simulator I wrote to simulate a bunch of clients against our SQL database was exhibiting some weird behavior. I noticed that about every 20 seconds or so the output of my routine would pause for about 10 seconds. By the way, this was run on an 8 way machine and I started 100 threads each opening a connection to the database and executing queries with a 30 second think time. The think time I simulated with a Sleep() call. So I first figured that SQL might be having some problems which was causing my delay. I ran a SQL profile session and all the numbers looked pretty good, in the 10 to 20 ms range. So then I figured it was my client program. After reading the part about Sleep() continuing to use the thread's quanta I decided to use a wait function to see if that would produce any better results. And indeed it did. I no longer see the delay in my output. SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com > > No, I'm not mixing Sleep() with SpinWait(). I believe Sleep(), as > > implemented by the .NET framework, is not relinquishing control to [quoted text clipped - 36 lines] > > Pete > From reading the docs I'm remember it saying something about continuing to > use the thread's quanta. From reading which docs? It just makes no sense that any version of Sleep() would in fact do a spin wait until the timeslice is up. The moment you call Sleep() (.NET or Windows API) with a non-zero value, that thread should be suspended (and most of the time even with zero passed to it). I have a hard time believing you've seen any official documentation that says otherwise. The fact that using the wait-on-an-event method instead changes the behavior doesn't prove that Sleep() was using up the timeslice. It just means it does something different that changed the behavior. Pete I tried to go back and find the docs I'm referring to but can't. So the only conclusion I can come to is that I somehow read that into the docs based on seeing the part about specifying zero to relinquish control to other waiting threads. Though I don't think this caused me to see something in the behavior of my program after making the change that really isn't there. I no longer see this considerable lag in my output (which I'm doing via a Console.WriteLine()). Maybe that's explained by Jeffrey's post where he mentions that the OS will boost the priority of a thread after a Wait(). SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com > > From reading the docs I'm remember it saying something about continuing to > > use the thread's quanta. [quoted text clipped - 10 lines] > > Pete > [...] I no longer see > this considerable lag in my output (which I'm doing via a > Console.WriteLine()). Maybe that's explained by Jeffrey's post where he > mentions that the OS will boost the priority of a thread after a Wait(). Could be. Since you still have failed to provide any actual code or detailed description of what is actually happening, it's very hard for anyone else to say. All I'm saying is that I'm practically certain that when you call Sleep() with a non-zero value, your thread does not continue running. Pete Here is the code in question. try { portfolios = eom.Portfolios[Eom.Side.Client]; foreach(Eom.IPortfolio port in portfolios) { list.Add(port.Name); } names = (string[]) list.ToArray(typeof(string)); while (true) { // Thread.Sleep(this._thinkTime); wait.WaitOne(this._thinkTime, true); name = names[rand.Next(0, names.Length)]; portfolio = (Eom.IClientPortfolio) portfolios[name]; orders = portfolio.Orders; if (orders != null) { foreach(Eom.IOrder orer in orders) { } } marketPortfolios = portfolio.MarketPortfolios; foreach (Eom.IProperty prop in portfolio.Properties) { } // Set some properties // Set client portfolio property at client portfolio. property = portfolio.Properties[Eom.Properties.TradingInstructions]; if (property != null) property.Value = "Nick's trading instructions."; else portfolio.Properties.Add(Eom.Properties.TradingInstructions, "Nick's trading instructions."); // Set client order property at client portfolio. property = portfolio.Properties[Eom.Properties.Guaranteed]; if (property != null) property.Value = true; else portfolio.Properties.Add(Eom.Properties.Guaranteed, true); // Set client order property at client order. } } catch(Exception e) { Console.WriteLine("Worker encountered the following exception:\n\n{0}", e.ToString()); throw; } SignatureThanks, Nick nicknospamdu@community.nospam remove "nospam" change community. to msn.com Hi Nick, Thanks for your feedback. Most part of the code snippet you provided is the business code of your application which is not informative regarding the Sleep/Event.WaitOne delay. It is hard for us to provide any further information based on this code snippet. Thanks for your understanding. Can you provide more information regarding the "output rountine" in your application and thread modeling in your application? Also, what is the "better result" behavior as you described? Thanks. Best regards, Jeffrey Tan Microsoft Online Community Support ================================================== Get notification to my posts through email? Please refer to http://msdn.microsoft.com/subscriptions/managednewsgroups/default.aspx#notif ications. Note: The MSDN Managed Newsgroup support offering is for non-urgent issues where an initial response from the community or a Microsoft Support Engineer within 1 business day is acceptable. Please note that each follow up response may take approximately 2 business days as the support professional working with you may need further investigation to reach the most efficient resolution. The offering is not appropriate for situations that require urgent, real-time or phone-based interactions or complex project analysis and dump analysis issues. Issues of this nature are best handled working with a dedicated Microsoft Support Engineer by contacting Microsoft Customer Support Services (CSS) at http://msdn.microsoft.com/subscriptions/support/default.aspx. ================================================== This posting is provided "AS IS" with no warranties, and confers no rights. Hi Nick, It costs me some time to read all your discussions with the community before I can provide a reply to you. I am not sure if I understand you completely. Based on my understanding, you are writting a stress simulator application for SQL Server. After openning a connection to the SQL Server, the thread uses Thread.Sleep() method to simulate 30 second *processing time*(think time). However, every 20 seconds or so the output of your routine would pause for about 10 seconds, so you wanted to find out the root cause of this problem. If I have misunderstood you, please feel free to tell me, thanks. The point that I do not understand well is what "output of your routine pause for 10 seconds" means. Can you tell me what is this "routine"? How does the routine generate output? Another point: >Sleep() continuing to use the thread's quanta I decided to use a wait function to see >if that would produce any better results I think I do not understand your expected behavior completely. Can you telll me what is the "better result" behavior? Then I will judge to see why Thread.Sleep() does not meet your need. Thanks. Anyway, I wanted to share some information with you: 1. .Net2.0 System.Threading.Thread.Sleep() method internally calls Win32 SleepEx() API. I can prove this by setting a breakpoint in Kernel32!SleepEx API in windbg for a .Net2.0 Winform application. Below is the call stack I got while I call Thread.Sleep() method in .Net: 0:000> !dumpstack Current frame: KERNEL32!SleepEx ChildEBP RetAddr Caller,Callee 0013f230 792144fc mscorwks!Thread::UserSleep+0x93, calling KERNEL32!SleepEx 0013f250 79214574 mscorwks!ThreadNative::Sleep+0x30, calling mscorwks!Thread::UserSleep 0013f260 011a03a6 011a03a6 0013f28c 79bbaeb0 (stub for System.Threading.Thread.Sleep), calling 011a037c 0013f290 0111026f (MethodDesc 0xc253c8 +0x17 winformtest.Form1.button1_Click), calling 79bbaeab (MethodDesc 0x79bbaeb0 System.Threading.Thread.Sleep) 0013f2a0 7b881fa4 (MethodDesc 0x7b9e5108 +0x54 System.Windows.Forms.Control.OnClick) If you want to know more detailed information, you may download SSCLI2.0 project to view the sample source code of ThreadNative::Sleep, Thread::UserSleep of CLR. "Shared Source Common Language Infrastructure 2.0 Release" http://www.microsoft.com/downloads/details.aspx?FamilyId=8C09FD61-3F26-4555- AE17-3121B4F51D4D&displaylang=en 2. In Windows thread schedulling, any wait operation will be decreased 1 point of quantum unit. And the default thread quantum has about 3 quantum units to run. So after Sleep(30000) calling, the calling thread will go into wait state for 30 seconds, then its quantum will reduce 1 unit. Note: once in wait state, this thread can not be scheduled for execution, other thread will execute. Also, after wakeup, the calling thread will not run immediately, it will be placed in the read thread queue for scheduling. For event/semaphore, it will normally behaves the same as Thread.Sleep(). However, after wakeup, Windows will provide a special priority boost to the calling thread, so that the wakeup thread can run immediately. I am not sure if this difference is the root cause of your problem. 3. If you think your simulator application is hanging for 10 seconds, the best technology to troubshoot is using "Process Explorer" to examine what each thread is hanging about. You may just examine each thread call stack during this 10 seconds hang. The call stack will reveal what's going on. The following article may not be very relevant to your question, but it demonstrates how to troubleshoot the hang/delay style problem: http://www.codeproject.com/system/NoDeleteDelay.asp Finally, the Windows thread scheduling is not a trivial topic, it is covered best in Chapter 6 of Mark E. Russinovich's wonderful book <Windows Internals>. More specificly, "Thread Scheduling" section is dedicated for this topic. You may give it a reading to understand more background information. Hope this helps. Best regards, Jeffrey Tan Microsoft Online Community Support ================================================== Get notification to my posts through email? Please refer to http://msdn.microsoft.com/subscriptions/managednewsgroups/default.aspx#notif ications. Note: The MSDN Managed Newsgroup support offering is for non-urgent issues where an initial response from the community or a Microsoft Support Engineer within 1 business day is acceptable. Please note that each follow up response may take approximately 2 business days as the support professional working with you may need further investigation to reach the most efficient resolution. The offering is not appropriate for situations that require urgent, real-time or phone-based interactions or complex project analysis and dump analysis issues. Issues of this nature are best handled working with a dedicated Microsoft Support Engineer by contacting Microsoft Customer Support Services (CSS) at http://msdn.microsoft.com/subscriptions/support/default.aspx. ================================================== This posting is provided "AS IS" with no warranties, and confers no rights. Free MagazinesGet these publications absolutely FREE for up to 12 months. There are no hidden fees and no obligation. Simply choose a title, complete the application form and submit it. Read more ...
|
Reply to this Message
|
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2008 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.