thread_do used to use pthread_exit(). However that resulted in
hard times of testing for memory leaks. The reason is that on pthread_exit()
not all memory is freed bt pthread (probably for future threads or false
belief that the application is terminating). For these reasons a simple return
pthread_exit() results in much more memory being allocated.
This is needed only in the tests. The reason is that if you call thpool_destroy and then exit immedietely, maybe the program will exit before all the threads had the time to deallocate. In that way it is impossible to check for memory leaks.
In production you don’t have to worry about this since if you call exit, immedietely after you destroyied the pool, the threads will be freed anyway by the OS. If you eitherway destroy the pool in the middle of your program it doesn’t matter again since the program will not exit immediately and thus threads will have more than enough time to terminate.
wait() will spike CPU usage to full when called. This is normal as long as it doesn’t last for more than 1 second. The reason this happens is that
wait() goes through various phases of polling (what is called smart polling).
wait() works in this way, is that the function is mostly used when someone wants to wait for some calculation to finish. So if the calculation is assumed to take a long time then we don’t want to poll too often. Still we want to poll fast in case the calculation is a simple one. To solve these two problems, this seemingly awkward behaviour is present.