On Fri, Jan 09, 2009 at 01:02:36PM +0100, Roland Bock wrote:
Is there a way to determine the order in which the (shared) locks are acquired? I tried testing with the attached program and got rather weird and (seemingly) non-deterministic results.
Threads are by their nature non-deterministic; all ordering must be implemented manually. Even with a single, ordinary, lock, there are no guarantees on the order in which waiters are being scheduled upon unlock.
This I do not understand at all. The readers are using the shared mutex. Why is the second group waiting for the first to be finished?
Probably because the mutex implementation noticed that there have appeared new writers. This is just a wild guess, I haven't looked at your program. RW mutex implementation has to be careful not to starve readers nor writers.
And how can the last writer be handled in second place?
That's perfectly normal, mutexes are not FIFO. Now that you've mentioned RW-locks, I remembered a nice article on concurrency: http://queue.acm.org/detail.cfm?id=1454462