> > One solution is to have the threads exit (via the pthread_join > mechanism, of course) when they have finished the current > request. > The way i understand the description of pthread_join, the thread which called pthread_create must call pthread_join. So that would mean that a thread who finished its job should somehow notify the main thread to go ahead with the join? Since my main thread is usually in a blocked state (accept()), a finishing thread should probably "sign in" into some (locked) variable which the main thread could inspect next time he wakes up. Would this be a correct way?
> > Another (very common) solution is not to create one thread per request, > but instead have a small set of worker threads, each one to handle > many requests over the life of the program. A "master" thread > accepts requests, and gives each one to one of the worker threads. > Only if there are no available worker threads does it create new > threads. > Here again, this would probably have to be done by some sign-in/sign-out mechanism... Thanks for the suggestions Jody > J > ------------------------------------------------------------------------------ This SF.net email is sponsored by: SourcForge Community SourceForge wants to tell your story. http://p.sf.net/sfu/sf-spreadtheword _______________________________________________ Valgrind-users mailing list Valgrind-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/valgrind-users
