Larry Osterman, in his blog post(http://blogs.msdn.com/larryosterman/archive/2004/03/29/101329.aspx) talks of how threadpools that use critical sections to synchronize suffer from the boxcar problem and as a result,have scalability issues. He recommends using completion ports to achieve better performance.
Now, the implementation in win32threadpool.cpp seems to be quite similar to the one he points out. In particular, ThreadpoolMgr::WorkerThreadStart has a lot of similar code to what Larry writes - for e.g, the WorkRequestNotification event is similar (or atleast seems similar) to Larry's m_heventThreadPool in his first thread snippet. If this is the case,does this implementation suffer from the problems that Larry talks about? Thanks, Sriram =================================== This list is hosted by DevelopMentorŪ http://www.develop.com View archives and manage your subscription(s) at http://discuss.develop.com
